Patrick Collison on Innovation and Scientific Progress
Jan 28 2019

instacart-300x200.jpg Patrick Collison, co-founder and CEO of Stripe, talks with EconTalk host Russ Roberts about the pace of innovation. Collison argues that despite enormous increases in the numbers of scientists and researchers, the pace of progress in scientific and technological understanding does not seem to be increasing accordingly. The conversation looks at the challenge of measuring innovation and whether the pace of innovation should be a matter of concern and if so, what might be done about it.

Tyler Cowen on Stubborn Attachments, Prosperity, and the Good Society
Tyler Cowen of George Mason University and the co-host of the blog Marginal Revolution talks with EconTalk host Russ Roberts about Stubborn Attachments, his book-length treatment of how to think about public policy. Cowen argues that economic growth--properly defined--is the...
Graham on Start-ups, Innovation, and Creativity
Paul Graham, essayist, programmer and partner in the y-combinator talks with EconTalk host Russ Roberts about start-ups, innovation, and creativity. Graham draws on his experience as entrepreneur and investor to discuss the current state of the start-up world and how...
Explore audio highlights, further reading that will help you delve deeper into this week’s episode, and vigorous conversations in the form of our comments section below.


John K
Jan 28 2019 at 7:57am

As much hand wringing there is about innovation, I don’t think most companies are really looking for it seriously. I know that has been the case with 90% of the places I have worked. The focus today is on making more money, not on doing something special. This is the case in the music business, film and now technology. When I started in tech everyone was trying to do something special. Then the business starting hiring MBAs and it started putting up barriers to being hired.  Backed with tech was getting started we took people who had passion that seemed smart. Steve Jobs’s resume (before Apple became successful) would have never even make it through HR screening. Today, everyone is trying to make more money. Innovation doesn’t happen without creativity and creativity cannot willed into existence or controlled. Today’s companies are all about control and control and creativity are not compatible.

Jan 28 2019 at 10:18am

Most of  my career has been involved in scientific research, so this episode has personal interest to me.

Patrick Collison: (my emphasis added) … Ben Jones and Bruce Weinberg have done some kind of neat analysis of just the profiles of Nobel Prize winners. And in particular they observe that over the course of the Nobel Prize, in the early days, the average age of a scientist who won was about 37 years old. Whereas, today, across, essentially all fields, all Nobel Prizes, that has grown by 10 years to about 47. And again, this is kind of thought-provoking. Right? … it is of course also very consistent with a picture of: It is just getting harder to discover things. It’s taking longer. It takes more time to get to the frontier of knowledge. You have to sort of accumulate more comprehension, understanding, of, you know, more different fields. And so I think that kind of data is thought-provoking.

Consider what would be required to become a “Renaissance man” with significant understanding of many different fields.  Compare that challenge today with what it would have been hundreds of years ago.

If you imagine accumulated scientific knowledge as an expanding sphere, in earlier days one individual could get their arms around large parts of that sphere.  Yet, as it grows, that is inevitably less possible and any one individual can manage deep knowledge that touches the frontier for a smaller and smaller part of the whole.  That isn’t a “problem” to solve, simply a reality that per person grasp of the frontiers will inevitably shrink in proportion to the whole.

Yet in a given domain — especially where our current dominant paradigm has gotten it wrong — there will still be opportunities for someone to have the insights that contribute to a paradigm shift.

Later in the episode, Collison talks about the significant difference in returns for the Howard Hughes Medical Institute (HHMI) grants, which fund a person rather than a specific grant proposal.

I would suggest that on that point there may be a confusion about what is cause or effect.  If an organization funds grant proposals, they have to assess what they imagine will be productive investment.  When HHMI supports a person, they are selecting those people who seem most promising.  The method of funding is not causing those people to be promising.  Rather, it is because they are promising researchers that they are being chosen.

Even if there is some real benefit to giving some researchers unconstrained freedom to explore, if we vastly increased the number of researchers receiving such grants, there would certainly be diminishing returns as more grants are given to less promising individuals.

On investing in new companies (around 1:05:00), there was a great earlier episode with a relevant fascination discussion of the decision process of investors (e.g. would you have invested in Google?)   See Marc Andreessen on Venture Capital and the Digital Future

Jan 30 2019 at 8:55pm

I am also a researcher. The fallacy about the HHMI has been on Econtalk before this episode. The HHMI chooses to fund the top 1% of scientists – basically the leaders in every field. They are extremely selective. The amount of money they receive reflects their past performance. And the HHMI does not renew their 5-year grant if you do not publish groundbreaking research.

I describe the HHMI funding as being similar to the NFL. The NFL pay lots of money to the best players. Giving a large amount of money to a player does not cause them to be the best. The amount of money they receive simply reflects their past performance. And there is a strong correlation between past and future performance.

I would add that this guest is simply wrong. One quick example: CRISPR/Cas9 is a recent innovation that is the biggest advance in Biology since PCR.



Dr Golabki
Feb 1 2019 at 11:05am

Brian –


CRISPR is also a great example of the benefit of having so many scientist. The people who discovered CRISPR weren’t a bunch of polymaths thinking profound thoughts about the biggest questions of biology. They were scientists working weird bacteria. Most big discoveries (at least in Biology) are weird accidents where you thinking you’re working on a very narrow problem but it turns out to have surprising importance. You need a wide net.


I general, I think the Nobel Prize analysis is mistaken. There’s many more writers now than at any point in human history, so does that factor that many of the great works of literature are from hundreds of years ago mean that writing is wildly less efficient? Does the existence of Babe Ruth and Bill Russell prove that baseball and basketball play is much less efficient? The data is telling us about the stories we tell about science (or arts or sports), not about the fundamental value of the science itself.


Also there are certainly many areas of science that are much more efficient. Genome sequencing is an obvious example in Biology. But there’s lots of examples like that. In 2009 (when I was still a research scientist) I routinely ran 96 parallel experiments on plates over the course of a day. In 1999, in the same amount of time I could have run 1 similar experiments. In 1989 running 1 of these experiments might have taken a month. In 1979 those experiments would have been literally impossible. Now you can run >1,000 of these experiments in parrellel.


Another example is journal research. In 2019, if I want to find the 20 most important articles on a scientific topic, I can now do that with Google Scholar in less than an hour. In 1999 it would have taken a couple days and you’d almost certainly have to be part of a university system. In 1979 you’d better know someone who already knew the field, because if you didn’t you were likely to be in the basement of a library for months.

Feb 5 2019 at 10:17am

This view of CRSPR is also wrong.  The brilliant scientist realizes he has something amazing.  The average scientists throws out the data as noise.  This is a recurring scientific theme where lesser scientists often don’t even realize they have a potential breakthrough.  In retrospect we credit the genius with serendipity.

Adam Wildavsky
Jan 28 2019 at 2:05pm

I suggest that research today is on average less productive than in the past because the vast majority of it is funded by governments. We should note not the total amount of funding but its source. I remember an Econtalk episode a few years ago where we were encouraged to consider, not whether government-funded research had produced useful innovations, but the productivity of government-funded research compared to privately funded research such as that by HHMI and for-profit firms.

Ayn Rand addressed this issue in her 1972 essay “The Establishing of an Establishment”, excerpted at

How would Washington bureaucrats—or Congressmen, for that matter—know which scientist to encourage, particularly in so controversial a field as social science? The safest method is to choose men who have achieved some sort of reputation. Whether their reputation is deserved or not, whether their achievements are valid or not, whether they rose by merit, pull, publicity or accident, are questions which the awarders do not and cannot consider. When personal judgment is inoperative (or forbidden), men’s first concern is not how to choose, but how to justify their choice. This will necessarily prompt committee members, bureaucrats and politicians to gravitate toward “prestigious names.” The result is to help establish those already established—i.e., to entrench the status quo.

The worst part of it is the fact that this method of selection is not confined to the cowardly or the corrupt, that the honest official is obliged to use it. The method is forced on him by the terms of the situation. To pass an informed, independent judgment on the value of every applicant or project in every field of science, an official would have to be a universal scholar. If he consults “experts” in the field, the dilemma remains: either he has to be a scholar who knows which experts to consult—or he has to surrender his judgment to men trained by the very professors he is supposed to judge. The awarding of grants to famous “leaders,” therefore, appears to him as the only fair policy—on the premise that “somebody made them famous, somebody knows, even if I don’t.”

Mort Dubois
Jan 29 2019 at 6:39pm

I would think that Rand’s method would be quite efficient, as it eliminates all women out of hand.

Dr Golabki
Jan 30 2019 at 1:41pm

But the period of rapid technological growth (according to Patrick’s analysis) was also heavily supported by government funding. The system is certainly not perfect but iIt’s facile to just blame the government.


When I was a research scientist my funding came from (A) the private university where I worked, (B) grants from private institutions interested in supporting science (including but not limited to HHMI), (C) grants from the defense department, (D) grants from for-profit drug companies, (E) grants from government healthcare groups (e.g. NIH).


This is a system that has emerged in the US and is widely considered to be the best in the world. And as Russ pointed out, it’s not at all obvious what the “right” way to fix it is.

Ruth Fisher
Jan 28 2019 at 2:42pm

The productivity of innovation has slowed because society has become much more complex than it used to be. Returns to innovation require the funding and development of new technologies, together with their adoption by society. All aspects of this process have become much more complex.

New innovations today require both incredible depth as well as breadth of understanding. Hence the increasing size of research teams. Technologies themselves have become increasingly complex. For example, the average car today contains roughly 30,000 parts. Compare that to Ford’s Model T, manufactured during 1908 – 1927, which contained roughly 1,481 parts. Alexander Graham Bell’s large box telephone contained 4 critical parts. The average smartphone today has over 300 parts and is covered by 250,000 patents. Imagine what it takes to coordinate the production of a new model automobile or a new model phone!

Then, once a technology has been developed, it must be adopted by society. This generally requires “plugging” the technology into an existing ecosystem of other technologies and processes and behaviors. This is very complex, both from a technical standpoint, as well as from a social standpoint.

There is evidence that frontier firms have been innovating at a nice pace, but that their innovations have not been diffusing throughout society as quickly as they used to. This suggests that the social aspect of adopting new technologies has become problematic.

Richard Fulmer
Jan 28 2019 at 4:01pm

The “impact review” of Nobel Prizes is unconvincing.  A scientist will likely have a far better grasp of the importance of a 100 year-old discovery than that of a year-old discovery.

James Lohner
Jan 28 2019 at 4:21pm

Does anyone know where he gets the figure from for Self reported happiness vs GDP per capita?

J Scheppers
Jan 28 2019 at 7:21pm

Innovation is hard. Everyone likes nice answers that innovators find after they are in the mainstream, but almost no likes change. Even those that go through the crucible and make meaningful change find pain, hard work, and constant dismissal from a majority.

Want more innovation? You need arbitrage players that will bet their own resources and form teams that feel deep materially important goals that make teammate work synergistically.  Built in with the difficult truth more often than not the pain givers and rejecters are going to be right 75% of the time.  To make it even more difficult it is only the rejecters and pain givers who can validate your innovation arbitrage in the market.

Consumers choose how much innovation they desire by how frequently the change or experiment with new potentially valuable products.

JK Brown
Jan 28 2019 at 10:38pm

It is rather naive to think that the innovation that came to a zenith in the early 20th century would continue, given the multitude of built up ideas that were finally manifest due to technological development.  New forms of power (electricity), new sources of power (oil, then nuclear), new prime movers (Internal Combustion Engine), new chemicals and compounds, cheaper building material (steel),new control mechanisms (transistors), etc.

There is no massive overhang of newly penetrated natural phenomena that is just waiting for technology to make their confirmation possible.  We are in an incremental phase.

Add to that the harsh reality of the decline in individual liberty over the 20th century.  After increasing individual liberty and appreciation of unapproved thoughts, through the 17th to late 19th century, the 20th century was a time of the increase in state power and the licensing, regulating and control of the individual inhibiting unapproved advancements.  Innovative minds chaff at the bureaucrats red tape and look elsewhere for a fruitful life.  We saw part of this in 1980-2000, when the best minds could run free in cyberspace while avoiding the state interventionism in the old industrial areas

All mankind’s progress has been achieved as a result of the initiative of a small minority that began to deviate from the ideas and customs of the majority until their example finally moved the others to accept the innovation themselves. To give the majority the right to dictate to the minority what it is to think, to read, and to do is to put a stop to progress once and for all.

Mises, Ludwig von (1927). Liberalism (p. 54).

Then we should consider that productivity growth is naturally lower in the Service Economy.  The nature of services is personalized service that can only be done in person over a set amount of time.  One innovator can’t take 10 cents off a products cost and then have that multiplied to the billions spread over the work of hundreds as can be done in manufacturing.

But all is not lost.  There is quite a lot of innovation down in the dull infrastructure.  Switch mode power supplies get smaller and cheaper, motor controls are advancing, miniature components are facilitating advancements in medicine, etc.  As for Moore’s Law, the limits of the silicon is being reached, but the advancements continue off the one chip to the whole system resulting in speed increases.

One might consider it ironic that our entire modern digital communications are governed by equations developed in the 19th century to model telegraph lines.  They didn’t even teach time-domain transmission lines when I was an undergraduate in the ’80s, nor according to the Georgia Tech professor whose course lectures I listened to, in the 1990s.  We learned time-harmonic transmission line theory for things like radio antenna cables and power lines.

Jan 28 2019 at 11:43pm

i for one do not believe we are as innovative as we think we are. If we were truly innovative we would have eliminated anxiety fear greed envy and the likes by now. And not thru a pharmaceutical.

Jan 29 2019 at 10:37am

Maybe a small portion of the decrease in the rate could be due to crowding.  Say that 1 cook in the kitchen can produce meals for 10 people.  Maybe 5 cooks in the kitchen can specialize and produce meals for 100 people.  Now imagine another 10x increase of 50 cooks in the same kitchen.

Jan 29 2019 at 10:38am

One explanation as to why we’re seeing a diminished “return of investment” for each individual, modern-day scholar relates to the nature of knowledge and innovation itself. When we increase our knowledge of a subject, our awareness of the complexity of that subject also increases. That complexity prompts people in that subject field to specialize and subspecialize to gain greater mastery. If science was an assembly line, the more we produce, the larger the assembly line becomes.

For example, I work as a physician at a major cancer center. Primary cancer care (at least for solid tumors) is coordinated between medical oncologists, radiation oncologist and surgeons. If a person develops neurological toxicity to treatment, they will see a neurologist specializing in cancer care — breathing issues a cancer pulmonologist, mental health issues a cancer psychiatrist, etc. Each specialist has their own set of nurses and support staff who also specialize in that field. There’s also dedicated research staff in every department enrolling patients in studies and tracking them (some performing a dual clinical role, but most specializing in pure research).

Decades ago, much of the care and research would be managed by a single oncologist or surgeon. A century ago, it all may have been managed by a primary care physician. You could argue that we’re seeing a diminished return with providers, but we just realize that cancer care is more complex.

This would also explain the bias around Nobel prizes. A century ago, a Nobel prize may have gone to a discovery that opened up (hypothetically) five new fields. That next generation will each produce a Nobel prize winning discovery in their respective field, and so forth. So while that first Nobel prize is credited for growth at an exponential, each subsequent individual field doesn’t appear as expansive, but that doesn’t mean innovation overall is dropping, overall.

Jan 29 2019 at 11:35am

This interview reminds me of Solzhenitsyn’s “In The First Circle” Chapter 54, character Kagan when knowing that bad things were going to happen to him writes to the Council of Ministers that he would create a “system of remote control for torpedo boats.”  He has no competency to do, but creates a team in which they accomplish nothing and yet he survives disappears in the Gulag Marfino.  Measuring innovation is like measuring air.  All these measures are useless because of capitalism.  Capitalism is the greatest system in the world.  It is a conversation between the producer and consumer.  As an inventor myself, to measure my innovation with a number like GDP or TFP, etc is a non starter.  It takes a long time for the dance of capitalism to register into sales.  In the internet of all things, it is easier.  Why?  You are not producing anything.  You are not creating anything.  All you are doing is acting as a conduit for easier and easier conversation processes.  Measuring innovation and productivity in the way it is described in the interview only works in Large organizations and academic institutions.  But these two groups only care about hand-outs from the government or the internal cash flow of their companies.  And they buy innovation from others.  Neither group is engaged with the customer, the real customer.  Because their customer is peer reviews, or government agencies or some process that stops innovation to keep the current monopolies in place.  To me people like Patrick Collison have no skin in the game of humanity and innovation.  They just find places in the conversation of commerce which is ineffective to make it more effective.  They are rent seekers/the new middlemen of society.  As for Nobel Prizes, they stopped being practical when they became political and peer review.   Ideas are not getting harder to find.  That is silly.  If the government and large institutions were to get out of the way, there are plenty of ideas big and small.

Thank you Mr. Roberts for all that you have taught me in these discussions.  I am reading books that I never would have thought were for me.

Nick Ronalds
Jan 29 2019 at 3:03pm

Very thought-provoking. I wonder whether part of the explanation for the slowdown in innovation is the Mancur Olson thesis, that mature economies get more sclerotic because of growing power of vested interests (rent seeking by “distributional coalitions”). And BTW, could it be that the age of grant-getters is increasing because the grant-givers, whether governments or private sources, are getting more risk averse?

Patrick raised some tantalizing possibilities, such as the astonishing success rate when grants are given to individuals rather than for specific projects. Perhaps the best Rx would be more diversity and entrepreneurship in the grant-giving process itself.

Todd Kreider
Jan 29 2019 at 9:20pm

“as I’m sure most of your listeners know, that kind of the story of factor productivity growth in the United States over the course of the last, call it, 150 years is one of sort of lowish growth, and then sort of, you know, a real growth spurt over the course of the middle of the 20th century. It’s been a much lower level since perhaps 1970 or so–slight kind of peak around, you know, 2000, hypothesized to be because of the quote-unquote “ICT [information and communications technology] Revolution.” And now back to, again, quite historically low levels,…”

Looking at TFP statistics from Robert Gordon and FRED:

1890-1900 0.6%

1900-1910 0.3%

1910-1920 0.7%

1920-1930 1.3%

1930-1940 1.8%

1940-1950 3.4%

1950-1960 1.6%

1960-1970 1.9%

1970-1980 0.4%

1980-1990 0.8%

1990-2000 0.8%

2000-2014 0.7%

But FRED shows nothing unusual around the year 2000.


Dr Golabki
Jan 30 2019 at 10:03am

An alternative story about why accelerating innovation is hard… if you have 100 Einsteins you don’t get 100x the theories of relativity. You get a little bit more physics innovation a little bit earlier. In other words, if Einstein had not existed it’s not that we never would have figured out relativity, it would have taken a few more years, but someone else would have done it.

This view is that the limiting factor for new breakthrough ideas isn’t the number of smart scientists, it’s the overall state of science being ready to have someone make the breakthrough.

This answers the puzzle of why so many of the great scientific and math breakthroughs seem to happen independently at roughly the same time. An example that I think Russ has mentioned on few times is Leibniz and Newton both discovering calculous simultaneously, but there as similar examples in physics, biology and chemistry.

This runs counter to the idea that innovation comes of the white hot burning mind of a 1 in a trillion genius… but I think that idea is not very helpful in understanding how science actually works and much more the product of human storytelling and our own mythology about ourselves.

Andy Walker
Jan 30 2019 at 10:40am

I was surprised there was no mention in the discussion or suggested reading of at least two books:

The Structure of Scientific Revolutions by Thomas Kuhn (1962)
The Innovator’s Dilemma by Clayton Christensen (1997)

Both books cover much of what was discussed in the interview i.e. that there is rapid progress in a given endeavor that peaks at some point and then faces diminishing returns for any number of reasons. A third author, Joel Barker, took Thomas Kuhn’s ideas and popularized them in books such as Future Edge (1992).

Jan 30 2019 at 11:24am

What’s the world going to look like in a hundred years? In a thousand years? Right? I think, you know, arguably the single most important input into, sort of, that prediction, is: What is the aggregate rate of progress in science going to be, between now and then? Right?

But why limit yourself to a single most important input?
Because nobody knows the future, wouldn’t it be best to hold a zero prejudice in regards to any field of input? Perhaps the institutions that educate scientists are not cultivating imagination and creativity so what you end up with is a
homogenous scientific population that can’t see the world as well as they should, let alone interact, and understand it.

Have Brian Eno on this show and have him explain to you what produced great innovators. His story of Kadinsky has an insight to how things happen.

Russ, thx for calling out…But I’m not quite as convinced as you are that something obviously is happening that we understand.

Carter Ferrell
Feb 2 2019 at 1:53pm

There is another side of the conversation surrounding the increasing age of Nobel Prize recipients that Russ nearly latches on to, but misses. He mentions the reputation cost of getting it “wrong” as an explanation of growing conservatism at Nobel, but possibly even more of a problem is the structural issue that an annual prize of recognition faces when addressing an entire field of study.

A lot of valuable innovation goes unrecognized every year. For every Nobel there may be a dozen more compelling cases that would have sufficed; and as each field develops its own pool of talent, so too does the pressure build every year to recognize “X” while he or she still draws breath. Every year we fall further and further behind.

Feb 3 2019 at 2:43pm

What’s the deal with criticizing Mazzucato at the end? At 1:11 you suggest we rethink how institutions like the NIH work, and that’s exactly what she was saying. One of her main points was that we need to find ways to build funding agencies that can iteratively learn from the past while operating under uncertainty. That approach would be right at home in this conversation, but then you feel the need to criticize, which I think says quite a lot about your baseline ideology.

Tyler Larson
Feb 4 2019 at 2:04am

The cost of discovery is enormous and growing; this was mentioned in the interview, but it doesn’t seem that either the host or the guest had a grasp of the scale involved. It is so extreme that it’s not unreasonable to attribute the entire loss in efficiency to this fact alone. No, really, I’m serious —

Start out with this video: — go on, it’s only a couple of minutes, an excerpt from a Q&A session from a few days ago. The idea is that thousands of man-years of science and engineering has to get expended just to design a single bit of technical infrastructure that you never knew existed, that you never knew was necessary, but without which all the visible science would be impossible. Replicate that same story millions of times over, year after year, and you get an idea where all our scientists are.

There’s some obvious examples in the world of physics: The Higgs Boson took a staggering amount of infrastructure to discover – we’re all familiar with the size of the LHC at CERN; we famously got the world-wide-web out of it merely as an optimization for sharing data between scientists. But the infrastructure for collecting and analyzing the data they used — for example the entire careers of some of the most skilled researches in the world have gone into nothing but figuring out how to capture data fast enough for the LHC to operate at all, and there are tens, maybe hundreds of thousands of problems like that to solve.

But is that an isolated example? Take the recent nobel for gravitational waves, then — discovered by measuring the stretching of space itself. Not measuring the size of objects, but rather detecting tiny waves in the spacetime that our world inhabits. Each segment of their apparatus uses two mirrors 4km apart, and has to detect changes in the distance between them of less than the width of a proton. Again, the science and innovation that had to happen just to conduct this experiment is beyond comprehension. But that all pales in comparison to what they have to do for the next discovery. To get the accuracy they need, the next version of this setup will have to be placed in orbit around the sun. Putting things in space makes everything infinitely more complicated, but earth is apparently just too small. They’re launching in about a decade.

It’s not that we’re running out of things to discover, it’s that we’re running out of cheap things to discover, in terms of time and complexity. Science and Innovation themselves aren’t slowing down, they’re just being consumed as invisible prerequisites to the “real” science and innovation that everyone measures.

Feb 5 2019 at 10:34am

Wow the comments are just exceptional.  Thanks for all the responses.

I am not too worried about the rate of innovation because for many reasons I think we are doing decent.  Maybe not super well but decent.  More experimentation from funding agencies and private agencies would be a good idea.  I sat in an intimate meeting in graduate school with a recent Nobel prize winner who funding his nobel prize research with grants for other things because the funding agencies would never fund his real research.  Stories like this abound.

Why are R01 grant recipients getting older?  A scientists in the biological sciences needs to graduate HS, go to college, go to grad school, be a post-doc, maybe do another post-doc, and then get a position and work to get funding on his own lab.  Thats 4+4+2-3+2-3 years before they get their own lab.  Those steps are not merely hoops (although they could be).  Usually at each step they are learning general knowledge followed by  a new technique to answer their specific research questions.  All this takes time.

I think the criticisms of the HH grants made above are all correct – but it would be great if we could identify the top scientists earlier.  Can we identify them in HS and skip all this long process?  How about early in college?  I think finding and risking money on scientists early in their careers would be one avenue.  We would get a lot of duds, but the current system also produces duds.  The benefit of funding people instead of grants is that they might be able to focus less on grant writing (a considerable time investment) and more on science earlier in their careers.  I doubt that is the “answer” but I do think it would be great if more billionaires spent less on humanitarian aims and more on innovative ways to try and fund science.  If they develop a Y-combinator of science that may be one of the most important innovations in our lifetimes.

Feb 9 2019 at 11:15am

Moving from extensive schooling toward a greater focus on education seems highly valuable.  Not only because a lot of schooling is a waste of time but more importantly because the whole schooling process is a kind of conditioning / programming process that heavily biases you toward status quo thinking (some or even much of which is false).

Take economics for example, there are various incompatible ideas under this heading including modern neo-Marxist economics, Keynesian economics, Austrian Economics, etc.  Which one do you learn in school?  The one that’s true (if any), or the one that fits the status quo and the culture of the school?  Imagine how many economists are incredibly well-schooled in their set of ideas and will spend their whole lives writing papers and books and the like advancing faulty ideas and training students with bad ideas and implicitly doing harm to the advancement of the science by their highly-schooled and well-intentioned efforts.

To me this kind of problem cries out for a greater knowledge of epistemology.  Before we can say which ideas are true we have to have a solid grasp of the machinery of truth.  To learn which school(s) of economics have sound ideas you need to be good judge of truth.  Imagine, one of the most popular economic textbooks circa 1989 thought the USSR would overtake the US in economic output.  The fact that he author believed that should be incredibly embarrassing.  The fact that this textbook was used to school generations of economists should be terrifying.  And that’s the world we live in.  Rather than focus on schooling and status quo thinking, we need to think more about how to figure out what’s true and why.

David Nickum
Feb 5 2019 at 3:18pm

I really enjoyed this podcast, Dr. Roberts you were at your best, exploring the subject from so many different angle. I really enjoy your program.

We had extraordinary tools created during the industrial revolution (e.g., trains, planes, cars, cranes, etc). We are just now starting to build the tools necessary (e.g., CRISPR, Deep Learning, etc) for the next revolution in plant science, chemical science, designer drugs, etc. The future’s so bright.

Electricity was more straightforward to put in place for the Industrial Revolution. Data is the infrastructure needed for the next revolution. Even though it’s taking longer to put in place, the revolution is still coming.

Feb 9 2019 at 11:42am

Various kinds of economic freedom indexes seem to support the notion of economic prosperity being related to freedom.  And there are many theoretical reasons to believe this too, such as the Haykenian knowledge problem or the Misesian calculation problem, etc.

If the goal is a greater rate of innovation then surely we should pursue greater freedom.

It costs something like $1b to bring a drug to market in the US because there is very little freedom as it relates to producing and selling drugs.  This is a clear example of a lack of freedom (“regulation”) functioning to constrain innovation.

The nice thing about increasing freedom is not just that it helps to increase productivity and innovation, but more importantly it does so without someone putting their thumb on the scale.  In today’s world of heavy regulation and taxation and the government funding of schools and science there are many thumbs on many scales, and it’s imposed on us from the top-down, not voluntarily chosen by all participants.

I know it’s unfashionable to say that we would be better off without wise technocratic overlords putting their thumbs on the scale.  I blame schools for this because they encourage obedience to authority / conformity over critical thinking or valuable knowledge generally.

If we want to live in a better world with more innovation we need to make people smarter.  So far it doesn’t look like IQs can be improved much with existing technology, but surely there is a huge potential for gain by just doing a better job of filling people’s heads with better ideas.  Less school, more education.

C Farquharson
Feb 15 2019 at 10:59am

I wondered throughout this talk why Collison was chosen to discuss scientific innovation. It was still a thought provoking discussion. This is the first time I have come to the site and the conversations spawned by this episode are every bit as informative as the talk itself. I second the comment above that EconTalk has let me to read many other authors that I had not considered. Great podcast.

Feb 16 2019 at 12:37am

More mediocre scientists do not just do mediocre work;  they mostly tear down exceptional work and scientists.

Marilyne Tolle
Feb 17 2019 at 1:52pm

On the structure of funding, this February 2019 paper “Large teams develop and small teams disrupt science and technology” (U. of Chicago and Northwestern) suggests that funding should be reallocated to small teams from large ones.

Abstract (bold parts are mine):

One of the most universal trends in science and technology today is the growth of large teams in all areas, as solitary researchers and small teams diminish in prevalence. Increases in team size have been attributed to the specialization of scientific activities, improvements in communication technology, or the complexity of modern problems that require interdisciplinary solutions. This shift in team size raises the question of whether and how the character of the science and technology produced by large teams differs from that of small teams. Here we analyse more than 65 million papers, patents and software products that span the period 1954–2014, and demonstrate that across this period smaller teams have tended to disrupt science and technology with new ideas and opportunities, whereas larger teams have tended to develop existing ones. Work from larger teams builds on more-recent and popular developments, and attention to their work comes immediately. By contrast, contributions by smaller teams search more deeply into the past, are viewed as disruptive to science and technology and succeed further into the future—if at all. Observed differences between small and large teams are magnified for higher-impact work, with small teams known for disruptive work and large teams for developing work. Differences in topic and research design account for a small part of the relationship between team size and disruption; most of the effect occurs at the level of the individual, as people move between smaller and larger teams. These results demonstrate that both small and large teams are essential to a flourishing ecology of science and technology, and suggest that, to achieve this, science policies should aim to support a diversity of team sizes.


required, not displayed
required, not displayed

This site uses Akismet to reduce spam. Learn how your comment data is processed.


This week's guest:

This week's focus:

Additional ideas and people mentioned in this podcast episode:

A few more readings and background resources:

A few more EconTalk podcast episodes:

TimePodcast Episode Highlights

Intro. [Recording date: December 14, 2018.]

Russ Roberts: I want to remind listeners to go to and in the upper lefthand corner you'll find a link to our annual survey, where you can vote for your favorite episodes of the year and tell us about yourself and your listening experience.


Russ Roberts: And now for today's guest, Patrick Collison.... Our topic for today is innovation, based on an article your wrote in The Atlantic with Michael Nielsen. But, I want to start with what Stripe is, for listeners who don't know.

Patrick Collison: Sure. So, Stripe is an infrastructure company that builds--well, we started out building kind of payment services for internet businesses, technology companies. And so, if they want to sell their products over the Internet, we build the APIs [Application Program Interfaces] and tools for them to do that. And we've kind of expanded kind of these initial payment services to now provide a kind of slightly broader set of, you know, economic or kind of financial tools and services, to businesses and so forth. So, for example, we help them with fraud detection; or we help them run a marketplace--you know, something like Lyft or Airbnb, those kinds of products. And, we provide an incorporation service, so that for founders--and especially founders not in the United States, for whom sort of getting access to the kinds of opportunities that we have here might be more difficult--we also help them do that. And so we kind of describe Stripe as building economic infrastructure for the Internet, and the kind of, the core idea, is we want to make the flow of money and the ability to transact kind of a true universal. We think there is a kind of missing layer of Internet infrastructure that sort of, kind of surprisingly, has not been built over the course of the Internet's history to date.

Russ Roberts: And why wasn't PayPal and other solutions like that, that solution?

Patrick Collison: Well, I think it's a very interesting question. And we certainly wondered about that when we were starting Stripe. And I think the--I think [?] PayPal kind of had the right aspiration. I mean, if you kind of read contemporaneous materials from kind of when they started--you know, their vision was slightly different: it was more about consumers, more about building a new kind of maybe online bank, and so on. So the focus was a bit different. But I do think they had the kind of right central idea, that, again, there's kind of a missing layer of infrastructure. And I think, really, the thing that kind of stymied them was their acquisition by eBay. There was sort of an outflux of talent--really immediately. And I think the--this is a--accomplishments of the company in subsequent 5 years, subsequent 10 years--you know, really just did not compare at all to that which they'd accomplished so far. And they--of course, that's a pretty common story. Right? I mean, that's not, I think, something kind of specific, perhaps, to the PayPal/eBay case. That's potentially a common pattern.

Russ Roberts: And, I have to ask this, just to let listeners know what we're talking about: Stripe's done pretty well.

Patrick Collison: Um, we feel very lucky, in terms of the adoption we've had. John and I--John's my brother and co-founder--we were developers ourselves. And when we started Stripe, we were in college. And we went to build Stripe for people kind of like us--in the sense of, as developers, we were just kind of mystified that it was so difficult to move money around online. Especially since so many other things, like cloud hosting and so on, were becoming so straightforward, and so easy to manage from a developer's standpoint. And so, we built what we felt was a kind of oddly missing component of sort of the building blocks of the Internet. But the thing that we did not anticipate--and again, we've been sort of very fortunate around, is, while Stripe was indeed adopted by, kind of people like, quote-unquote, "like us"--you know, individual developers, very early-stage startups and so on, and kind of those really were where our initial adoption came from--the kind of breadth has really expanded over the kind of 7 years since we launched. And so, you know, Stripe is now used by many such startups; but also by, you know, some of the largest companies in the world. You know, including Facebook and Amazon. And so, we did not expect that breadth of adoption. And I think it [?] just speaks to the, again, the extent to which there is a layer of stuff that's really needed which just was not being built.


Russ Roberts: So, the idea for this episode came from a piece we'll link to, as I mention, in The Atlantic. And the focus of the piece is that science isn't progressing at a particularly fast rate. Innovation seems to be slowing down. And, it's, perhaps something to be concerned about. And you start with an argument that I'd like to talk about first--which, of course, is very close to my heart. Which is: measuring this is quite difficult. It's very hard to know whether science is doing great, speeding up, slowing down. Different people in different corners of a field, different fields, are wholly different parts of the elephant: They don't know what's going on outside their field and they might have a perception that's very field-centric, or personal, from their own perspective. How do we even begin to get at this question? How did you end up, with Michael Nielsen, trying to get at this question?

Patrick Collison: Well, [?] clarify one thing with the argument to start, which is: We don't really take a stand on whether the rate of progress is slowing down, per se. We're--we kind of see arguments either way there. But, we're somewhat agnostic on it. What we are much more confident of--and I think the, sort of the hypothesis that the data very strongly supports, um, is that the rate of progress is declining very rapidly on a per-person, per-hour, or a per-dollar basis. Right? And so, perhaps because of the enormous increase in our investment, in our efforts, you know, perhaps we are in fact still sustaining, you know, constant or, you know, approximately similar overall aggregate progress, but again, on a sort of on an individual basis. That our per-person productivity seems much, much lower. In terms of kind of how we got to thinking of measuring it, I guess it's always seemed a bit strange to me that we just don't collectively obsess over this question more. Right? I mean, with Stride, we think a lot about, sort of rates of growth. We describe Stride's mission as trying to increase the GDP [Gross Domestic Product] of the Internet. And, you know, that's obviously kind of a trite phrasing. But the sort of core idea is we do want to sort of take a broad view of the Internet economy, of technological progress, how the company got started, how well they fare, and so on; and try to figure out [?] really broad-based measures and interventions that sort of can help create, that might help nudge that rate of growth to be somewhat higher. But, of course, technology, and especially kind of software technology and so on--that's a kind of application. Right? We're sort of in the translation field of sort of taking the results of science, of sort of fundamental knowledge and discovery, and sort of deploying them to the rest of the world. We think about the sort of aggregate, you know, progress of humanity--there is that really important applied aspect. But there is also, obviously, the kind of, the fundamental underpinnings of our understanding of nature and basic knowledge--you know, research furthers not a particular destination in mind; we might be really surprised at what it is that you discover. And I think it just as what we think as people. We think of, as a civilization, you know: What's the world going to look like in a hundred years? In a thousand years? Right? I think, you know, arguably the single most important input into, sort of, that prediction, is: What is the aggregate rate of progress in science going to be, between now and then? Right? Maybe some tail, really bad things are going to happen. But if they don't happen, I think that the single biggest determinant will be the aggregate rate of progress in science. And so I think the kind of first-order thing that should surprise us is that this is not something that we are collectively obsessed with. Right? These type of questions [?] about what the actual rate is--I think we should all be able to agree that the rate is super-important. And that we should really care a lot about it. And so, we kind of initially came to this, and in terms of just wondering about the rate itself. And then, kind of as we thought more about it and how one might measure it and what kind of other correlates might be, and so forth: Then we really became pretty convinced of--again, the case made in the article--that the rate per person is declining very sharply. And the reason, of course, that should be concerning, even if the aggregate rate today is, you know, constant-ish, is that, of course, we cannot sustain these--I say, exponentially in a qualitative sense, not in a sort of rigorous quantitative sense--where we cannot sustain these qualitatively exponentially increasing inputs indefinitely.


Russ Roberts: I interviewed Tyler Cowen, earlier on the program a number of times, but about his recent work on growth and the book he has now, called Stubborn Attachments.

Patrick Collison: Yes.

Russ Roberts: It's a very provocative and interesting book--

Patrick Collison: Published by Stripe Press.

Russ Roberts: I did not know that. That is so cool. And I was--well, first, I found his case very persuasive, in the abstract. And it certainly would have been my view 5 years ago--lately, as listeners know, I've become a little bit more skeptical about the value of growth and the value of the advance of technology. I mean, if we look back to 1900 and we look to today--if we've been in 1900 and say, you raise the question, 'What's the world going to look like in 100 years?' it's a really good question. In 1900, we would have been rather clueless about it. And we certainly--if we could go back to 1900, bring that person to the present and say, 'How does this look to you?' there would be many, many things that that person would be wildly exuberant about. Those things would include longer life expectancy, fewer women dying in childbirth, huge reductions in infant mortality; much more pleasant daily life. So, there are a lot of glorious things that have happened. And, if that person had been black, or a woman, or gay, there would be other things to be happy about as well--social things. And I think those--I've started to wonder and worry a little bit about whether we underrate those social things. So, I happened to mention some good things that have happened since 1900. There have been some not-so-good things that have happened since 1900 in daily life. People worry, as I recently discussed with Sebastian Junger--and this episode hasn't aired yet, Patrick, so you haven't heard it--but, we seem to be quite a bit more isolated. We seem a little bit more lonely. There's some dysfunctionality about our society. And science and technology, I don't think, solve those problems. And they might even make them worse. So, what are your thoughts on that? When you said, 'We ought to be obsessed about it,' maybe there are other things we ought to be more obsessed about, and we pick science because we have some idea of how it happens.

Patrick Collison: Um, you know, I'm certainly sympathetic to aspects of that view. Economic or scientific growth are certainly not everything. But I really want to stand up for growth here. The caveats I think do not diminish its central importance. And so, if you just look at GDP per capita around the world, I mean the scatterplot of that against self-reported wellbeing, the R2 on that regression is 0.68. It's astonishing. As you know, it's very rare in sort of anything around human psychology and sociology and so on that you get a correlation that's that strong. You look at it visually, it's really pretty amazing. And even--again, I don't want to sound like an absolutist here. There certainly are some things that have gotten worse about the world since, for example, 1900. But even if you take for example the case of isolation loneliness, while, yes, I think there are aspects or communities that have degraded or where there have been unintended and unfortunate kind of second-order consequences, be it the automobile or single-family zoning and all the rest, at the same time, there does seem to be some aspect where isolation is a kind of consumption good that people tend to purchase as they become more affluent. And so it's not entirely clear to me whether all of that observed isolation is in fact undesirable--

Russ Roberts: Yeah, I agree with you--

Patrick Collison: and [?] people's underlying revealed preferences. And so again, not an absolutist here, but I do think the central case for growth remains extremely strong.

Russ Roberts: Yeah. I'm sympathetic to that also, of course. I just wonder--I think there's an issue of focus. And, I think it's much easier to focus on technological change and the things we can count, like GDP. It's ironic a little bit, because we're talking about something that's hard to count--knowledge. Eventually that knowledge gets translated, we hope, into innovation; and we hope that innovation translates into a better experience for human beings. But it does get complicated, each piece of that chain.


Russ Roberts: Let's move to the evidence. So, on the surface, how would you know--as you point out, we're not obsessed with it, so there's no index that most people would agree is the right way to think about it. What are some ways that you and Michael Nielsen looked at it and that others have done?

Patrick Collison: Right. So, I think there are maybe 4 or so key ways you can--that we've seen so far--that you can take a look at this question. And, obviously, we are and would be kind of super-interested to read or to study any other analyses of this question. So, I think the most obvious one is actually to just look at the macroeconomic statistics; and in particular I think, for some obvious reasons, TFP [Total Factor Productivity], kind of productivity growth, is sort of the most relevant variable here, where, if science is working super-well--if we're discovering tons of new knowledge, things that enable us to produce economic output, economy value, more efficiently, we should really see that in the TFP statistics. And, you know, perhaps [?]--

Russ Roberts: Total Factor Productivity--

Patrick Collison: Exactly. Yeah. And so, perhaps not every single discovery manifests here, but sort of overall, over time, it really should be visible. And, of course, you know, as I'm sure most of your listeners know, that kind of the story of factor productivity growth in the United States over the course of the last, call it, 150 years is one of sort of lowish growth, and then sort of, you know, a real growth spurt over the course of the middle of the 20th century. It's been a much lower level since perhaps 1970 or so--slight kind of peak around, you know, 2000, hypothesized to be because of the quote-unquote "ICT [information and communications technology] Revolution." And now back to, again, quite historically low levels, at least if we take our window as being the last 150 years. And so, I mean, that is--that chart, that picture, is very consistent with one where sort of mid-century science worked really well: we discovered a lot of amazing things, but somehow the rate of discovery seems to be lower since then. So, that's one way of looking at the question. Another way of looking at the question is to try to do that kind of TFP calculation, but within fields. Within specific domains. Because, you know, you might say, 'Well, the economy is so big and so many other things going on, and there's so many confounders[?], and all the rest, it's really hard to conclude a whole lot when you look at something as coarse as the total economy of the United States.' So, you know, perhaps you could choose some sort of specific domains. Perhaps you could choose drug discovery. Or maybe you could choose crop production techniques. Or maybe you could choose semiconductor production. Right? And within any of those fields you can try to do some more specific, granular output measures in that: Maybe it's drug discovery; maybe it's semiconductor density; maybe it's [?] of volume of wheat production and so on. And if you do that--and this, in particular, was done by a really fantastic paper from Michael Webb, Nick Bloom, [?] Jones--

Russ Roberts: Charles Jones and John Van Reenen--the paper is called "Are Ideas Getting Harder to Find?" And we'll link to that.

Patrick Collison: Yeah. "Are Ideas Getting Harder to Find?" And I really do encourage any of your listeners who are interested in this to go read it. It's a very kind of readable paper. And the kind of central conclusion is that not only do they find that kind of per person or per dollar productivity is declining within these specific fields and industries, but they find that it is in fact declining exponentially. And at surprisingly sort of consistent rates, in a way that really should be, I think, at the very least, thought-provoking for us. And moreover, quite concerning. And since that paper came out--I think the first pre-print was probably about two years ago or so, perhaps even a bit more--there's been no strong refutation that I've seen. And so, kind of this picture of, again, exponentially-defined productivity growth across a sort of quite broad variety of industries, is definitely striking. Thirdly--

Russ Roberts: And just--one sec. Hold that thought. I just want to--for listeners who are puzzled by that: The point is that, you say 'declining productivity'--it's because innovation is continuing. We're improving our knowledge of various things. And it's a great example in that paper about Moore's Law. So, we have a lot more capacity on our chips, our computer chips. The problem is, is that it takes an enormously larger number of workers and researchers and scholars to produce that improvement--

Patrick Collison: That's right--

Russ Roberts: and so the per-scholar rate is what's going down. Not the overall level of innovation, say, in certain fields. And the question, then, is why? We'll come back to that. So, continue on. So that's two types of evidence.

Patrick Collison: Yep. Third one you go look at is, kind of micro-data within sort of the shape of the production of scientific knowledge. Um, and so, for example, Ben Jones and Bruce Weinberg have done some kind of neat analysis of just the profiles of Nobel Prize winners. And in particular they observe that over the course of the Nobel Prize, in the early days, the average age of a scientist who won was about 37 years old. Whereas, today, across, essentially all fields, all Nobel Prizes, that has grown by 10 years to about 47. And again, this is kind of thought-provoking. Right? In that, I mean, you can of course imagine explanations that sort of have nothing to do with sort of rates of progress, and so on. Like, you know, maybe something has changed about the kind of structure of academic institutions or something. Um, but, you know, it is of course also very consistent with a picture of: It is just getting harder to discover things. It's taking longer. It takes more time to get to the frontier of knowledge. You have to sort of accumulate more comprehension, understanding, of, you know, more different fields. And so I think that kind of data is thought-provoking. In a similar vein the sort of size of scientific teams, the sort of extent of kind of co-authorship and the number of authors in publications and so on--you know, that's also sort of really growing very substantially. And again, it's like this picture where it takes bigger groups, more work, larger teams to create, you know, this new knowledge and these breakthroughs. So, I say that's kind of a third category, these kind of analyses of, so that the structure of knowledge production and science. And then the last one--and, the--as far as we're aware, at least, a kind of new one that Michael and I sort of introduce in this article--is to try to come up with some kind of aggregate measure of the sort of significance of scientific breakthroughs--because that is in some sense the relevant output measure here: Are we discussing kind of--are we discovering important new knowledge? And so what we decided to do is we took sort of pairwise comparisons of different Nobel Prize winning discoveries, and we asked scientists to kind of, in these comparisons, to choose that which they considered more significant. And then to use, sort of, all those kind of partial orderings to produce kind of total scores for different Nobel Prizes. And then we can kind of buckle[?] a little bit and just look at, for those scores of individual prizes, we can maybe generalize to a decade and so on, and get some kind of trend line for the kind of, you know, generally-perceived significance for Nobel Prize winning work that's occurred. Now, that, of course, is very subjective in very obvious ways. And therefore, I think, quite imperfect. But, I think it's not--we should not totally dismiss it. In the sense that, I think it's hard to dismiss this methodology without also dismissing the Nobel Prize itself. Because the whole conceit of the Nobel is that you can assess a very broad array of work, and you can sort of choose that which is most significant, most prize-worthy. And so if you can do it, kind of, in order to issue a single prize, I think you should also be able to do it across Nobel prizes. And, hey[?], when you kind of construct that dataset--and we did it for three Nobel Prizes: physics, medicine, and chemistry--what you see is that physics kind of has this picture that kind of accords in some way with that of TFP in the United States for kind of--really did well in kind of early and middle parts of 20th century; has really kind of declined somewhat since then, though the error bars are relatively wide. And then, chemistry and, you know, medicine/physiology--it's kind of high variance, but it looks relatively flat. And so, anyway, I think this is, again, quite thought-provoking and striking, because when you look at the number of researchers or the number of papers, or the, you know you calculate sort of the number of hours invested in these fields, these are growing enormously. Like, it, it's--

Russ Roberts: Hmmm--

Patrick Collison: it's really worth kind of looking at these charts to see just how much. Right? And so, if you kind of have, you know, 50 X [meaning 50 times?--Econlib Ed.], say, more researchers in one of these fields, but as judged by scientists themselves were producing significant discoveries at a nearly constant rate, you know, on some level, on some just very naive level, that does suggest that researchers--like an individual researcher--is becoming kind of 50 X less likely to produce, you know, this kind of really significant work. Um, and so, that I'd say is the fourth category. And it--looking across the board, I don't think that any one of these, um, sort of, um--I don't think any one category is itself knock down. Any of these ways of looking at the question, you know, has its sort of limitations and flaws. Um, and, you know, I have no enormous confidence in any of them. But, I don't see a reason as to why they should all be biased in the same direction. Right? In that, if this, you know, if this hypothesis that the productivity of science is not declining, that hypothesis is not true. I think we'd expect to see some kind of, some kind of contradictory data, when we look at different kinds of analyses. Whereas, we kind of see all of these analyses pointing in the same direction. And so it's when you take all of them together--and they are all sort of pointing the same way--that is something I think we should really take seriously.


Russ Roberts: I don't disagree. But I'm not quite as convinced as you are that something obviously is happening that we understand. And I know you don't pretend to understand it thoroughly, either. But, I just think a silly counter-example, published work in the sciences has increased dramatically. And I'm guessing--I could be wrong about that, but I'm guessing.

Patrick Collison: You mean the amount of published work?

Russ Roberts: Yeah. The number of articles--

Patrick Collison: Absolutely. Yeah, yeah. It's increased enormously.

Russ Roberts: And that's not--it doesn't tell us anything, particularly, except that there are more journals. And we know why there are more journals. It's that there's an enormous return to publishing an article. So, academics have innovated to increase the supply of that precious thing in response to those incentives. And I think--and the same is true of collaborative work: I don't know how much more collaborative work is really--how much more collaborative work is. Because, there's a big incentive to help people get more publications. Of course, you water down your own credit to some extent, but maybe not. I don't know. But, all these things--and, just to pick on you just for a second, Patrick: You know, the 'happiness data'--maybe people in rich countries think they are supposed to be happier, and so their self-purported happiness is listed to be higher. I just--we have--a lot of these kind of data are so complicated. So, just to take the--I love what you did by asking scientists about previous discoveries and sort of ranking them, Nobel Prizes: What are the great Nobel Prizes? It reminds me a little bit about baseball. You say, 'Who are the greatest baseball players of all time?' This is an age-old argument. People say, 'Well, obviously, there's nobody as good as Babe Ruth, or Ted Williams.' And there's a good argument to be made. There's also an argument to be made that, 'Oh, my gosh, Babe Ruth or Ted Williams today wouldn't be very good at all.' And present players are much, much better. It's so obvious. But, there's a lot of romance about Albert Einstein, and a lot of romance about Rutherford, and a lot of romance about Niels Bohr. And maybe that's why they rated the 1920s and 1930s so much higher. So, it's just hard to know. Right?

Patrick Collison: So, it is certainly hard to know. And again, I do want to emphasize that I do not think that any single argument here is sort of dispositive and definitive. Right?

Russ Roberts: Yep.

Patrick Collison: But, I do kind of want to push back on that degree of agnosticism, in the sense that, like there are some things we really do know. We do know that the number of working scientists has increased enormously. Right? And so I think it's kind of a very reasonable and valid question, 'Well, okay: For that increase,' and again, it is vast, you know, 'Have we gotten a proportional increase in the rate of discovery?' Right? And like, that should not be a hard question to answer. If the number of scientists have increased by, say, 20%, you know, perhaps the kind of error bars and all this stuff are so wide that we can't tell whether it's a 20% increase or not. But we are not talking about, um, sort of a 20% increase in our inputs. Depending on the measure you take, it varies. What we are talking about is increases of between sort of 10 X and 100 X. And if we are investing 10 X to 100 X more, it should really not be a difficult question: Are we getting 10X to 100X more output than we were previously. And--

Russ Roberts: So--

Patrick Collison: And so the fact that it's so hard for us to find that kind of vast improvement in our output--like, the mere fact that the answer to the question is not blindingly obvious, I think is itself suggestive.


Russ Roberts: So, an analogy, I guess would be, well--it's hard to figure out what the analogy is. I'm thinking when we have a lot more workers in some areas. Of course, most fields we have a lot fewer workers and more productivity. Because--

Patrick Collison: Absolutely--

Russ Roberts: which is interesting in and of itself.

Patrick Collison: That's what we should be squaring[?] to.

Russ Roberts: Right. So let's shift, let's take what you've claimed as true. And now let's think about two things: What might be the explanations? Of course there are many. And, what might be done about it if we wanted to change that? Because, it's a great example of how I think the disease--if you want to cure the disease, you'd better know what the disease is, rather than just, say, 'We need more science, education, and school.' Which is a kind of mindless corollary people draw, which I think is wrong--although we might need more science education in school or better science education--

Patrick Collison: Yes--

Russ Roberts: But, so, we can think about, I think, different things. One answer, of course, which you talk about and which we'll go into is that this is the nature of reality. The lowest-hanging fruit's been picked. We might also argue that the lowest-hanging scientists have already been chosen. And as we expand into the population with more and more people in these fields, we'd expect the gains to be smaller. So, that's one possibility. The second possibility would be, we organize the search for knowledge poorly. Universities and innovation labs and Silicon Valley need to be overhauled, because we are not getting the return that we are getting. And the third would be we need to change the rules of the game in the larger society. Obviously, you can lower your rate of total factor productivity by passing some really bad legislation--

Patrick Collison: yes--

Russ Roberts: that hampers people. And, of course, legislation and the rules of the game are changing all the time, as are cultural norms. And, all those things, all those three things get tangled up together. So--

Patrick Collison: yes--

Russ Roberts: So, talk to me about what you see as those possible explanations, and if you think there's any reason to think one of them is maybe more compelling than the other.

Patrick Collison: Yeah. So, I think you are absolutely right to kind of want avoid that sort of politician syllogism of, you know, 'We've got to do something and this is something therefore we've got to do that with, you know, science education in schools,' or whatever the case might be, even though, on that specific intervention, who knows? Maybe it does work. And so it's one thing to really emphasize at the outset of this, is, Michael and I, we really want more science, and more sort of scientific discovery, and more knowledge, and all the rest. And so, kind of, you know, one might in some sense kind of take this claim or this article or something to be kind of pessimistic about science. We're in some ways almost the opposite, where, um, we think it could be so much better. And we think it is so important that, if it is possible for it to be better for it to kind of get there, we really think we've got to engage with this question. Um, and, I think you are kind of dead right with that framing, in terms of kind of separating sort of contingent institutional or kind of sociological explanations for this line from those that are kind of about the metastructure of knowledge.

Russ Roberts: Yeah. The nature of reality, versus, I would say, the rules of the game. And I want to mention, by the way, that one of the more obvious implications of your work, if you take it just barefaced, is that we are spending too much money subsidizing graduate education and science; and we have too many scientists. That would be--I don't think that's, I know that's not your goal, but you could easily draw that conclusion as well.

Patrick Collison: Yeah. Well, I really think such a conclusion would be premature. To be very clear. That is not the case we are making. But, yes: I do think there should be, um, a more, sort of robust discussion about how it is we should be allocating our efforts and how all that input should be structured. On sort of this core question of like, okay, is this about the nature of reality or is it kind of just the nature of, you know, how we're doing things? I really think that that's very difficult to know. And people have been sort of making the case--you know, people like Sean Carroll have made the case in physics that, you know, we, there is a sense in which we've kind of gotten there and really explained a very large fraction of, you know, that which is to be explained. And, of course, you can always push back on that. And, you know, there are quotes, and you know, real, some apocryphal, of people who sort of thought we kind of reached the end of frontier in the past. But, you know, even if those assessments are wrong in the past, it doesn't mean they are necessarily wrong today. And obviously--so, I think there is kind of some sort of some credence to be ascribed to that argument. However, Michael's and my view is that, even if it is the case--or, well, I think it is assuredly the case that, depending on the field, that some of the story is kind of possibly of low-hanging fruit, and some of the story is sort of these kind of more contingent factors--and, we're pretty convinced, I guess, that a nontrivial part of the story entails these, sort of, again, sociological, institutional considerations. And so I think you sort of don't actually need to answer the question of, 'Well, is it 80-20 this way,' or the other way. Even if it is only 20% institutional factors, like the return to science is so enormous that still it's sort of overwhelmingly worth fixing those. And then, as we fix them, and as we experiment there and so on, perhaps we will learn more about which it kind of truly is. And, you know, I think you really--it's not hard to kind of discover that, or to, I've come to realize, that how we sort of produce knowledge today, how kind of our scientific industrial complex works, is not optimal. I have yet to meet a scientist who has even called it, you know, pretty good. Whether that's kind of the, you know, conservatism of funding mechanisms and apparatus, or the kind of assessment criteria and so on, or the time horizon of them, or the kind of rigidity with which fields and sort of career tracks are prescribed and so on--there's, I think there really is a lot that holds scientists back. And to kind of--it's been sort of widely reported that NIH [National Institutes of Health] R01 grants, NIH of course being the largest funder of science in the United States. The average age of kind of first receipt of NIH R01 grants has been kind of steadily increasing. Sort of, so, indicating, in kind of some sense it's getting even worse to be a sort of new scientist arriving on the scene. And there's very suggestive evidence that kind of perturbations in, you know, some of our institutional mechanisms, can really yield higher returns. There's a--a great paper on the Howard Hughes Medical Institute, HHMI, which is a really neat funder of science in the United States. And, you know, they kind of fund science somewhat differently than NIH. They give longer grants. They give grants to people rather than for kind of a specific work and project and so on. And so, anyway, this paper from Pierre Azoulay looked at sort of as you tried to take two populations that, you know, by most observable characteristics seem pretty similar; and as you look at those who, you know, received it, HTMI[?HMMI?] funding and those who did not, Azoulay concluded that, or showed, that HHMI grant recipients were 98% more likely to produce work that is in the top 1% by citations. And so, really not a small effect size. And, of course, you know, they are going to--actually perturbing that much; you are changing how funding works. But, you know, these people have sort of been trained in the same way; they still have the same kind of institutions. They are still, you know, working with the same kinds of colleagues and so on. And the fact that you can get such a large effect size just by changing one variable suggests to me that--it would be very surprising if, sort of perturbations and, you know, shifts and occurrences and so on, with other variables--you know, it was not possible to kind of create such of other, such improvements. And it's easy to forget, of course, how, kind of, in some way how accidental it is, where we've ended up today. Kind of, how the sort of U.S. post-War structure of science came to be. I mean, you know, as with anything like that, it's the product of, you know, all sorts of semi-random, you know, political, human factor and considerations. And, you know, obviously it's working pretty well on some level. You know, we've made more progress in the last, you know, hundred years than we've made in any prior century. But, I think we have very little reason to believe that it's close to optimal. And so, I guess Michael's and my core, strongest view, is that we really should be experimenting more here: the return is seen likely to be very high.


Russ Roberts: Yeah. That's a--well, we know one thing that post-War institutional setup is really good at, and that's producing more scientists. Which is not our ultimate goal.

Patrick Collison: Right.

Russ Roberts: But I do think you raise--you made me think of some really interesting cultural norms and responses that the current, complex way that different factors interact might be handicapped. Just going to think some thoughts out loud; and you can respond to it. You mentioned that Nobel Prize winners are getting older. The whole idea--and you mentioned that, you know, the Nobel Prize has to inevitably try to rank the quality of innovations. Well, maybe it's not very good at it. And, in particular, it's conservative. It's cautious. It's prudent. In the early days, not so much. Didn't know what they were doing. Didn't matter so much. Didn't have much of a brand name. Now, the idea that the Nobel Prize would be awarded to something that turned to be, like, wrong, would be very embarrassing. And so, as things--there's a sclerosis, it seems to me, about scientific knowledge as it's become more organized and more institutionalized. And I think there's a general lesson there that's really important. Because, people tend to talk about, assume, 'Well, the more organized, the better. Because then we can make it do stuff.' But it couldn't be--because of groupthink and caution and risk. As things get more advanced, and as we get richer--as society and as reputation has longer and bigger impacts--that a lot of the creativity has been, you know, wrung out of the system. And I'm very drawn to that for a bunch of ideological priors I have, so I have to be careful about my biases here. But it does seem to me--

Patrick Collison: But, but, but--no, I think you are right. I mean, you read the early history of sort of the great academic institutions in the United States, or the biographies of many of those who made some of the most important breakthroughs. And there is a kind of free-feeling nature to them, that does seem far less prevalent today. I was having this conversation recently with David Deutsch, the physicist--you know, some really significant work in quantum computing. Foundational work. And, you know, we were kind of reflecting on and sort of chatting about his career. And he was very firmly of the view that he could not have had the career he did, and could not have done the work that he did, had he been starting out today. Because, he didn't fit in any neat box. He wasn't exactly kind of, you know, fish or fowl and so on, and kind of flitted a little bit between different fields. And, you know, of course, that is also his great strength: he's such a sort of a deep and original thinker. And, as things have gotten a little more, kind of more striated, a little bit more structured and a little bit more--perhaps not a little bit more, perhaps a lot more--but, at least more, kind of formalized, I think there is kind of a real question as to the degree to which we are losing some of the, sort of, you know, very intrinsically necessary creativity and sort of--you know, discovering[?] new knowledge necessarily involves kind of breaking from existing models. And you know, perhaps in some way we are somehow disincentivizing that.

Russ Roberts: Yeah. I worry about groupthink. I know that NIH, having had family and friends in that research world, there are a lot of fads. There are things--there's political correctness. Not of the, necessarily the kind that people think about in, say, campus life or day-to-day life. But, within ideology--within fields, within science, there's political correctness about what's okay to think about and work on.

Patrick Collison: Yep. And it's striking, I think, how we've shifted a little--well, in particular, as we've kind of formulized funding mechanisms. How, I think, the dynamics of those mechanisms have really changed. Where, you know, I was just recently reading this biography of Gammal[?] and Delbrück, who kind of became a molecular biologist and also Gammal[?], the physicist. And it's, it's very striking how sort of they really relied upon sort of specific interventions by particular people to sort of sustain their careers and keep them going. It wasn't that they kind of hopped through the hoops of, again, standardized funding and [?]-making apparatus. It's not really clear to me--and I guess the kind of implicit thesis of the book is that they probably would not have. They, again, did not fit these standard molds. And so I do wonder a lot whether--I mean, this is of course always been group-think. Right? But perhaps the groupthink has become a little bit more potent, you know, over the past couple of decades [?].


Russ Roberts: So, let me suggest a hypothesis, then. And then maybe when we're done with this we'll think about some other--we focused on one piece of this puzzle as a possible main problem. But, let's imagine the following. Einstein did his innovations while he was a patent clerk. At night. Or probably some of the day, some, too--let's be honest. There was work on the job, even in the early 1900s. I mean, leisure on the job. The opportunity to do things other than your explicit tasks. But he didn't need a large lab. He didn't need a nuclear, an atomic accelerator, a SLAC [Stanford Linear Accelerator Center], which I incidentally see at Stanford, Stanford Linear Accelerator Center, whatever SLC stands for I don't know. But it's obviously--one thing that's obvious is that a lot of scientific innovation today requires capital. And, once it requires capital, it requires funding. And once it requires funding, it requires wealth. We have a lot of it, which is a glorious thing. The question is: How do you channel that wealth in ways that are going to be the most productive? A lot of it we channel through the political process--through NSF [National Science Foundation] and NIH. A lot of it is taking place even though, obviously, a lot of this work is more applied, in the corporate side; there is still real innovation taking place in the corporate sector. But, these fundamental science things take--not all of them, but a lot of them take money. And this sclerosis--and I would call it sort of the maturing of any industry--it gets a little less eager to take risks. It's not any one person doing that, although it does happen also at the personal level. I think, you know, it's not surprising that Bill Gates has not done a lot of innovation in computing lately. He also became a little more staid and a little more cautious and little more eager to maintain his current position. So, it's hard to break out of that. And, the challenge is: What kind of changes would we think about that would let us get to that more innovative, wild west frontier?

Patrick Collison: Yeah. I mean, it's, obviously, I come to this question from my kind of own perspective and biases, and no doubt highly incomplete view of the world. But one thing that is striking from sort of technology is obviously sort of the, certainly in this domain, the very deep truth of what you just said. Where, there is not a chance that the kind of innovation we see in kind of software and IT [Information Technology] and kind of, electronics and so on, there's not a chance that that would have occurred over the past couple of decades had we been reliant on a small handful of companies or institutions to produce them. Or even if we'd been reliant on one or two major funders. Because, of course, within the funders there's kind of biases and, you know, group think, and all of the rest. And so you actually have sort of--you have multiple layers of kind of competition, right? Competition among companies; and then competition among funders. Um, of course all of, coming to approach the world with sort of, you know, quite varying and substantially different theses about, you know, what even yields innovation. Some of them are very founder-oriented. Some of them are very market-oriented. Some of them are really, you know, want really robust and rigorous execution plan and all the rest. And, there is no right answer here, right? For some kinds of companies, for some kinds of founders, whatever, some approaches are going to work better than others. Right? And, so the fact that--well, and of course, that is indeed closer to kind of what science looked like sort of before it was professionalized. And before kind of institutions formed with kind of the same potency that they have today. And so, yeah. I guess the big question that I wonder about is: How do we--it's not that I have, or I think really anyone can have a strong view of kind of what the right funding mechanisms are, what the right institutional mechanisms are, and all the rest. It's more that it seems very surprising to me that the kind of structural monoculture is as strong as it is. And I think that's really the thing that we should be trying to break out of. And yeah, so, for example, an intervention that I think we need to try is having more scientists allocate more dollars directly themselves. Right? In that, let's say we carve out, you know, $100 million dollars, or a billion dollars a year; and we somehow, through whatever mechanism, choose, you know, really good scientists, and have them each allocate--maybe it's just a million dollars every year. Right? And, but they can kind of unilaterally direct that to whoever they think is doing kind of important work, where, you know, the marginal returns will be very high. And, you know, you kind of begin, talk to top scientists--they almost all know top people who they think are doing really important work not likely to get funded through current mechanisms but they would really, they think it's important to see make progress.

Russ Roberts: Well, if my claim--

Patrick Collison: Well, I guess my--

Russ Roberts: If my claim is correct about this cultural risk aversion, then, having that process be anonymous rather than named--

Patrick Collison: Sure--

Russ Roberts: would be an interesting model. Normally, we'd say, 'Well, we want to know who said this; we should spend a minimum [?] on x,' because they'll be more responsible. But maybe they'd be--we ¬know they'd be more cautious.

Patrick Collison: Right. Maybe it should only be revealed if it works. Or something. Right?

Russ Roberts: Yeah--

Patrick Collison: And then they should be showered with credit, after great risk[?].

Russ Roberts: There you go.

Patrick Collison: But, I guess my point in raising this is not to claim, again, that sort of, I think this is the sort of quote-unquote "right" intervention, or that this is kind of the model where we should kind of shift to, or I think that. It's more that I think we should have a portfolio of approaches that we're kind of experimenting with and observing the results of. I had dinner earlier this week with Howard Chang, who is a fabulous and indeed HHMI-funded professor at Stanford, and, you know, he was making the point while we were discussing this question that, we really should have some people who are becoming PIs, Principal Investigators, leading their own labs at much earlier ages. Some people are just ready for that in their early 1920s. But, again, kind of our current institutional structural mechanisms don't support that. And: How good an idea is that? How great would the returns be? I, of course, don't know. But I think that sounds like a great idea to try. Right? And so, I think we should be assembling our portfolio of bets here and really experimenting more.


Russ Roberts: I want to raise a different angle on this that we haven't talked about yet. I want to think about Nassim Nicholas Taleb, who has been a guest on the program many times, and he is famous, among other things for the idea of a Black Swan--a rare, catastrophic event that comes out of the far lefthand tail of a distribution that is not normally distributed; and that much of our intuition about life comes from the normal distribution. And I think some of what we're talking about here implicitly is making that error: We are talking about White Swans--

Patrick Collison: Yes.

Russ Roberts: It's not the right word: Black Swans are rare; White Swans are everywhere. But, by White Swan I mean a wildly pleasant unexpected event.

Patrick Collison: A positive catastrophic event. Yes.

Russ Roberts: And by our current--you and I are both talking about this, and I think the way most people think about it is, 'Well, if we had a lot more scientists, we'd have more people in the righthand tail, in a normal distribution.' But maybe it's actually--that's the wrong way to think about it. I want to read a quote from Scott Alexander, who--we'll link to this post he had. He attended a conference that you and Michael Nielson created, and he has some very thoughtful reflections on the issues we are talking about right now. But he said the following. He said

Shakespearean England had 1% of the population of the modern Anglosphere, and presumably even fewer than 1% of the artists. Yet it gave us Shakespeare. Are there a hundred Shakespeare-equivalents around today?

Meaning, my interpolation here, meaning, 'We have so many more people; we should have a lot more Shakespeares.' And Scott Alexander continues:

This is a harder problem than it seems--Shakespeare has become so venerable with historical hindsight that maybe nobody would acknowledge a Shakespeare-level master today even if they existed--but still, a hundred Shakespeares? If we look at some measure of great works of art per era, we find past eras giving us far more than we would predict from their population relative to our own. This is very hard to judge, and I would hate to be the guy who has to decide whether Harry Potter is better or worse than the Aeneid [Vergil's classic work--Russ Roberts]. But still? A hundred Shakespeares?

End of quote. And so, one way to think about this challenge that you are articulating is that we need to be thinking about Shakespeares. The Einsteins. The one in a million--and realize that if you have one in a million, if you have ten million you don't necessarily get ten of them. And how do we, culturally, create the opportunity for that very outside-the-box mind to flourish?

Patrick Collison: Yeah. So, I guess two points to that. So, first, kind of just to what you were saying previously, I think you're kind of exactly right that in these kind of, in these domains with these kind of convex returns where, or some power law returns, whatever you want to call it, I think you're going to be exactly right that what you want to do is you want to increase the number of samples and you want to kind of push towards more variance. Right?

Russ Roberts: Yeah.

Patrick Collison: And again, I think kind of another way of articulating that, which we're talking about, is that kind of there are all these mechanisms that actually seem not only not designed to increase the variance, but in many ways they kind of structurally attenuate us. And that just is really bad when we get these positive catastrophes--to abuse the [?] term. And on the a hundred Shakespeares point--I think that's actually a quite interesting example, because it's certainly not obvious that we do in fact have a hundred Shakespeares today, or that even in 500 years people will think that we had sort of a hundred people who were kind of sort of at the same level. And so, perhaps on some level, within drama or theater, however we want to define Shakespeare's field, perhaps we're seeing diminishing returns. Right? But, when we take a kind of broader view of, Shakespeare was also an entertainer, or in some way part of the entertainment production edifice, it's not clear to me that we don't have a hundred Shakespeares.

Russ Roberts: Absolutely. I agree.

Patrick Collison: Like, let's think about YouTube stars--

Russ Roberts: Yep.

Patrick Collison: or funny Twitter accounts. Digital arts.

Russ Roberts: Netflix.

Patrick Collison: Netflix. Video game designers. Right? And, of course, most of these things don't nearly have the kind of luster and status that sort of Shakespeare now holds; but that takes time. Status always lags. And so I think if we take a broader conception of the set of eligible fields, you know, again, it's really not clear to me that have seen any such decline.

Russ Roberts: Yeah. But there were probably a hundred Shakespeares available to be productive during Shakespeare's time, but most of them died young or spent their time hungry and didn't burn brightly enough with ambition or passion for their work to spend their time writing plays all day. And today we have the luxury that thousands and thousands of people can be in the entertainment industry. And, I have said before on this program--as soon as I say we are in the Golden Age of television, the Golden Age of movies: I think we are in the Golden Age of story-telling. YouTube is one example that you mentioned; and I mentioned Netflix. But the quality of the mini-series--and they are miniseries. So, something like, a show like The Americans which I'm in the middle of--no spoilers, please--but is really extraordinarily good, and is sustained over 6 seasons--I hope; I'm in Season 2, or 3--Season 3. But 6 seasons of 13 episodes each--that's just, that's 72--did the math wrong--84--what's 6 times 13? 78. It's 78 movies that this person--they're short; they're only 40 minutes long--but the creativity and the quality of the film-making, there or in The Crown or in The Wire or in--there's so many. So, there's a lot of Shakespeares out there in a certain dimension. That, yes, it's true that they don't have quite the--they don't have the luster because that's almost--it would be by definition. But the quality of riveting entertainment in our time: There's at least a hundred.

Patrick Collison: Totally agree. And, on your point that in Shakespeare's time there were probably a hundred more who sort of did not get access or the same kind of opportunities, that's something we think a lot about at Stripe. It's kind of part of our core thesis, in that we sit around and talk about this notion that talent is approximately evenly distributed, but sort of opportunity, of course, is so much more uneven. And I think the returns to fixing that, you know, we really should presume would be very high. And it's why we do things like Atlas; it's why we care so much about global expansion. And it's even why Stripe has invested in some sort of semi-adjacent companies, some of which you might be familiar with, like Pioneer or Lambda School--efforts like this. And I think that it's going to be easy to sort of structurally underestimate, I think, the returns to expansion of opportunity, because you can't really measure it ex ante. You have to sort of take it on some amount of faith. But, sort of, our intuition is that those returns are likely to be very high.


Russ Roberts: So, I want to move to a slightly different focus. We've talked about some of the cultural shortcomings, perhaps, and the way the funding is organized. You're an interesting case. You are--if listeners haven't noticed, you have a lilt in your voice because you grew up in Ireland. I think you came to the United States for college, at a young age--

Patrick Collison: That's right.

Russ Roberts: And, if you had been born in--we can name a lot of places where if you had been born, you wouldn't be here having this conversation and Stripe wouldn't exist, along with your brother. And fortunately, you were able to come to the United States. You were smart enough and risk-taking enough, I think, to drop out of college. Is that correct?

Patrick Collison: Yes.

Russ Roberts: Dropped out of college. Found a place--the West Coast of the United States, California, where there were really wealthy people willing to take a chance on you--and if that isn't true in your particular story it doesn't matter because it's true in lots of stories. And that strange confluence of opportunity, which is: The opportunity to move--we let you in; you had a chance to use your gifts in a way that required an enormous, some investment--it may be a big investment; and it came through--most of them don't--we know that. That system is really powerful. And, of course, some of that's at risk all the time. It's at risk all the time. Every piece of it. There's worries about we have too many immigrants. There's worries that we talked about, that Silicon Valley has become a little bit sclerotic and less innovative because of certain cultural or, because, we don't know, but all the good stuff has been discovered. But, I know you and I don't believe that. I don't think that's the problem that's holding us back. And so, what do you think of the--well, talk from your own story. What's your--what allowed you to flourish fully? Or somewhat fully--I don't mean to suggest that you've fulfilled your potential, because obviously, you are young--you're very young: you're 26, I think.

Patrick Collison: Oh, no, no, no. I've now aged a bit. I'm now a far older 30.

Russ Roberts: Oh my gosh.

Patrick Collison: I'm now old by Silicon Valley standards. But, yeah--I think you're really right in bringing this up, in that I think that a lot of, I think my perspective does come from, perhaps as with all of us, the kind of particulars of the background I had. And, I think kind of your characterization of it is, you know, exactly right. And, yeah. I conclude very strongly from kind of my upbringing and the fact that I had the opportunity to come to Silicon Valley, that there must be many more people like me, but who, for whatever reasons, sort of the, in sort of the Pachinko Machine of outcomes, it didn't all kind of all kind of fall out the right way. And, for me, there was, like: I think if you kind of ran the experiment multiple times, that plausibly in most outcomes I didn't make it here. And it's because of so much random happenstance. I [?] was really interested in this programming language called LISP. Because of that I decided to go to a conference at Stanford when I was 16. I discovered--so it was my first time in the United States--I kind of discovered college in the United States. And had I not have gone to that conference, I wouldn't have applied to school in the United States. And I'm pretty sure if I hadn't applied to school in the United States, I wouldn't have dropped out to go start a company; because when I got to the United States, I sort of discovered--I got to know the folks working on Reddit, and I was like, 'Oh, cool: starting a company seems pretty neat.' And so, there was so much kind of happenstance in my upbringing that led to where I am today that it really feels kind of necessarily the case, again, that there must be people who are every bit as capable as me--or much more capable than me, but, again, things have not come together quite as they did in my case. The other thing that is very striking from my story that again kind of gets back to some of these sort of science questions is, not only is it sort of potentially the case that some kind of committee might not have been willing to take a bet on Stripe or to fund Stripe: it was the case that most such committees that we approached, most VC [Venture Capital] firms, you know, were extremely unenthusiastic about the prospects of two near-teenagers going and sort of entering the financial industry and told us in no uncertain terms that they thought the idea was pretty foundationally ill-conceived. And, it was only because of a couple of specific individuals, among them Paul Graham and Peter Thiel, who were willing to, themselves, place that bet. If that had not happened, there would be no Stripe. And I'm really pretty convinced that sort of it was sufficiently--our success was sufficiently implausible that there was almost no committee that would ever have come to unanimity on the worthiness of a bet on Stripe. Just, if you take that for any one person there's a 1 in 10 chance they would believe in us--no need to--one in ten to any power starts to rapidly become a very small number. And so that's really very striking. And then of course as we work on Stripe itself, I mean, we're in the business of working with and sort of aiding the success of, again, these positive catastrophes--these, like, amazing companies that start out as two people and become these enormous, you know, forces in the world. Like, it's so vividly clear to us how fragile and delicate and implausible they are at the larval stage. And we see so tangibly the kind of potency of mechanisms to shepherd them to provide them with kind of role models, with inspiration, with cultural capital, with education and so on, that, again, I really feel like this must be case in other domains besides.


Russ Roberts: So, I want to think about a particular area for a minute, and I think it will pull a lot of the factors we've been talking about together. I just--I listen to you talk, and I have a lot of romance about innovation and entrepreneurship. I find it actually quite moving--not just the reality but your awareness of the tenuousness of success and achievement; and it's really an extraordinary thing. But as you were talking I kept thinking how everyone's response to these original conversation, piece of our conversation, about lagging productivity is, 'Well, we need to find the right way to structure funding science.' Or, 'We need to find the right way to reform scientific education, science education at the K-12 [Kindergarten through 12th grade] or at the college level.' And one of the things I wish--I think I've learned; and I think it's true: I'd like to get your response--is: that whole idea is just a dead end--

Patrick Collison: That's right--

Russ Roberts: The whole idea of finding the right way. And that most of the time, the right way is something we haven't even thought about at all. So, the idea of doing what we do better is almost certainly the wrong way to think about it. It comes up a lot in education: You know, 'We just need--what's the right curriculum?' And so much of the great education takes place on the ground, at the 1-on-1 level, the teacher-classroom level--we don't fully understand that interaction well at all. And we need to think about how to make that better. And we know how--I am pretty sure I know how not to do it. Which is: I don't want a national curriculum, and I don't want a national education requirement that people have, say, a Master's Degree in Education. All the standard things are--the only thing I know is that most of the standard things are wrong. So, we need new things. So, when I think about, say, one particular area, which is health, where there is so much potential. Forget all of our previous conversation about whether we are going to solve a theory of everything in physics that will be like Einstein's. We're just thinking about the ways we might innovate in health: So much of it is so focused on the current structure. And it would seem to me that there is such potential for someone like you, and the people you know well, who have, fortunately, have very large sums of money--you have been extremely successful. And, do you think that that success can be risked? And done in ways that are not traditional right now? That are--you mentioned Paul Graham. You were in the Y Combinator [YC]--is that correct?

Patrick Collison: Um, yeah. YC invested in Stripe, yes.

Russ Roberts: So, Paul Graham, we've had him on talking about the Y Combinator. Y Combinator is a pretty cool innovation, [?] quite like it--

Patrick Collison: It's an amazing innovation.

Russ Roberts: So we need, it seems to me, we need people with wherewithal, like you, and others, to get together and say, 'We've got to be brave enough to try to fail.' And it's ironic, because everybody knows in Silicon Valley that failure is acceptable; and it's even a badge of honor. And yet I worry that as people get more and more successful, they worry about their reputation. They worry about losing what they already have--

Patrick Collison: yes--

Russ Roberts: which is the human impulse. But, it would be great if you guys took a few more chances.

Patrick Collison: I agree with you very strongly. I think that--I'm very optimistic about sort of individual human agency; and I'm very optimistic about sort of the width of the right tail. And I'm very optimistic about the sort of transmissibility of new knowledge. And I'm sort of pessimistic about the ability of groups of humans to engage in effective, ex ante, top-down design. Again, I'm sure most of your listeners are familiar with the book Seeing Like a State, the sort of great screed against sort of monoculture in some sense. And so I think, you know, oriented that way we quite clearly need exactly what you're describing--which is mechanisms to break out of the monoculture. It just does not serve us well.

Russ Roberts: And this is, maybe--I don't know if it's the right thing to say, but, it's not my money. Easy for me to say to you, 'Patrick, you've done so well. Now take a chance.' So, that's none of my business; and I do think there is an enormous challenge, an opportunity, for some of the more successful people in Silicon Valley to do something outside the narrowest of spaces that are profit-oriented. And I've talked to a number of folks in that situation. And they all have an urge to change the world beyond the commercial. They have an urge to change the world beyond the rate of return. They don't just care about creating the next unicorn. They care about making a difference. But I guess I'd start by saying: Although that's not my money, the part that is my money, it's being taxed to fund NIH [National Institutes of Health] and NSF [National Science Foundation] and public education at the state level, the state university level. We just might re-think that; and then let other flowers blossom outside that way that we've been doing it for so long.

Patrick Collison: Yeah. One thing that's striking about HHMI [Howard Hughes Medical Institute] is they are very thoughtful about how they re-allocate their dollars. As in, they think that, 'Well, what are our lowest productivity dollars, and what can we do with them that might move the needle more?' Because they have--not exactly fixed, but, you know, fixed-ish endowment. And so their budget doesn't change a whole lot, year to year. And so, because of trying to maximize the impact of their work and the extent to which they move the needle on, you know, biomedical sciences, they, again, they really think hard about this question of, 'How should we change what we are doing?' And I think that a lot of the sort of individual incentives and institutional structure of some of our centralized funding mechanisms kind of militate against this sort of reconsideration, this kind of reapportioning and reallocation. And to kind of get back to an earlier point that you made, which is that one could construe the argument we made as a case for reducing funding in science. I really wouldn't consider that for a second. I think the questions we should be asking ourselves are much more about reapportioning and reallocation.

Russ Roberts: Yeah: Who makes those decisions, and--you haven't heard this episode, either; it's also been recorded recently with Mariana Mazzucato. She wants the government to be actively innovating--not just funding, but actually doing the innovation within government. And--obviously, I'm not in agreement with her, but that's another approach. But I think--obviously I agree with you, that the secret is to--we don't necessarily need to spend more money; we do need to spend it differently probably than we are spending it now.

Patrick Collison: Yeh.


Russ Roberts: Anything you want to add, Patrick? Anything we didn't talk about that you wish we had?

Russ Roberts: No, I think--I'm delighted we had the chance to have this conversation. Again, as we discussed at the outset, I really think that this is--I mean, it sounds kind of hyperbolic and overstated to say it, but I think this is kind of one the few things that is truly sort of a question or set of questions of kind of civilizational importance. And so, I guess my goal in talking about them and thinking about them and making kind of the very small contribution we did is hopefully kind of elevate their prominence somewhat. Like, if somebody else sort produced an analysis that showed that we are completely wrong, and actually the way to kind of improve the metastructure and institutional aspects of science is something totally different, I'd be delighted. That's great. Whatever makes the whole thing work better is what I'm in favor of. And, again, I think it's just something we should focus more on, take more seriously, and again, hopefully break out of the status-quo monoculture.

More EconTalk Episodes

Search Econlib