WEBVTT
00:00:07.679 --> 00:00:09.439
Welcome to Geoconomic Competition.
00:00:09.519 --> 00:00:15.119
I'm Francesca Guerti, your host, and in every episode I dive into the intricacies of geoconomics with other experts.
00:00:15.279 --> 00:00:20.559
You might have noticed we haven't really done an episode on tech or on AI yet.
00:00:20.719 --> 00:00:33.359
So today I have Carl Benedict Frey, Dieter Schwartz Associate Professor of AI and work at the Oxford Internet Institute, but also the author of The Technology Trap and of How Progress Ends.
00:00:33.600 --> 00:00:37.200
Two beautiful, fantastic books that I recommend you go by and read.
00:00:37.359 --> 00:00:39.119
Thank you for being here with me today, Carl.
00:00:39.359 --> 00:00:40.079
Good to be with you.
00:00:40.320 --> 00:00:43.039
There are many very interesting things in your latest book.
00:00:43.200 --> 00:00:51.359
And one of the most interesting framings, at least for me, is that society needs two engines that often are treated as somewhat competing.
00:00:51.439 --> 00:00:56.880
And one is the decentralized exploration to generate new ideas, we could call it somewhat innovation.
00:00:57.039 --> 00:01:03.759
And the other one is centralized exploitation to scale them, adopt them, diffuse them, you you pick the term.
00:01:03.920 --> 00:01:11.840
Can you walk us through how that tension works and especially why getting the balance wrong has been fatal for so many great powers?
00:01:12.239 --> 00:01:23.200
So if you run an assembly line or if you run a factory and every single worker on that assembly line questions every single procedure, you're not going to produce too much, right?
00:01:23.439 --> 00:01:28.560
But if nobody questions anything, you're also not going to invent anything new.
00:01:28.719 --> 00:01:32.959
And sort of that's the tension I'm I'm trying to explore in the book.
00:01:33.120 --> 00:01:48.319
And the key point, as you mentioned, is that if you want to explore new things, a certain degree of autonomy, different people exploring different technological trajectories is usually important to come up with new ideas, new technologies.
00:01:48.480 --> 00:02:09.039
But when you then settled on what it is you want to do, and when you want to produce and make your goods or services available to as many people as possible, you want to cut costs, you want to exploit new technologies like factories, that then leads you towards managerial hierarchies and greater centralization.
00:02:09.120 --> 00:02:15.439
And so both are needed for progress, but different stages of the technology life cycle, if you like.
00:02:15.840 --> 00:02:20.960
And we get to what stage we're in and who's doing these best in in a moment.
00:02:21.199 --> 00:02:36.159
Part of your book is tracing the history and the experiences of some of these, let's call them actors' great power, whether it is some dynasty in China, the Dutch Republic, Victorian Britain, Soviet Union, and so on and so forth, but also modern Japan, and we'll get to that in a moment.
00:02:36.479 --> 00:02:41.439
And you notice how many of these surged ahead and then stalled, right?
00:02:41.599 --> 00:02:52.400
So is there a common pattern in how the transition from the rise to stagnation plays out and perhaps some lessons learned or elements we should keep our eyes open for?
00:02:52.719 --> 00:03:00.960
So the case I make in the book is that moving between decentralization and centralization means that you need some institutional flexibility.
00:03:01.039 --> 00:03:12.800
And that is hard to do over the long run because you need to change your institutions and you often have incumbent interests that have a stake in the exist existing order of things.
00:03:12.960 --> 00:03:17.520
They are often going to resist reform and new technologies as well.
00:03:17.759 --> 00:03:24.479
But you can grow quite quickly for a fairly long period of time without being at the forefront of technology.
00:03:24.560 --> 00:03:39.520
And so if you take the Soviet Union, for example, we tend to look back at it as a failure now, and obviously in many ways it was, but it was also growing at six percent per year over a period of around four decades or so, which is quite respectable.
00:03:39.680 --> 00:04:00.960
And it did so basically by taking technology invented elsewhere and scaling it domestically, and so it took ample advantage of the Ford Motor Company's open door policy, sent delegations to Detroit to look at uh the Ford factories, learn from that, then build a vehicle industry in the Soviet Union.
00:04:01.039 --> 00:04:10.960
And in the age of mass production, the authorities could could hold factory uh managers accountable by simply benchmarking a performance across plants.
00:04:11.120 --> 00:04:20.399
And so it could that way determine if some factories were slacking off and not producing as much as maybe should compared to its peers.
00:04:20.560 --> 00:04:24.319
But at the frontiers of technology, that becomes much harder, right?
00:04:24.560 --> 00:04:28.720
Because when something is new, what do you benchmark against?
00:04:28.959 --> 00:04:40.079
In addition to that, what what the Soviet Union needed for further growth as the mass production system petered out globally was something new.
00:04:40.240 --> 00:04:45.920
That something new was the computer revolution, to which Soviet contributions were essentially none.
00:04:46.000 --> 00:04:54.000
And a big part of that was that if you were an inventor in the Soviet Union, you could go to, let's say, the Red Army and ask for funding.
00:04:54.160 --> 00:04:57.279
If they declined, maybe two or three other options.
00:04:57.439 --> 00:05:01.680
If they declined, then basically your your idea would die with you.
00:05:01.839 --> 00:05:13.120
That's quite different from the American system of much more decentralized finance where Besma Venture famously declined to invest in Google back in 1999 and probably regret it today.
00:05:13.279 --> 00:05:18.800
But it also illustrates that in in the late 90s Google was not a safe bet, right?
00:05:18.959 --> 00:05:27.680
Alta Vista and Yahoo were dominating such, and so somebody needed to take the risk to invest in order to know if Google would catch on.
00:05:27.759 --> 00:05:32.319
And the fact that Bessmer didn't invest didn't mean the end of Google because others stepped in.
00:05:32.399 --> 00:05:40.079
And so in a more decentralized system, you have different players that can take different bets, you can explore more technological pathways that way.
00:05:40.240 --> 00:05:50.720
And so that's more important at the technological frontier when it comes to pushing innovation and where the opportunities for simply adopting technologies invented invented elsewhere diminishes.
00:05:51.040 --> 00:05:51.439
Alright.
00:05:51.600 --> 00:06:00.639
And the other side of the spectrum is sort of the adoption part, which you know you've already mentioned in relation to the Soviet Union, you've already mentioned the limits.
00:06:00.800 --> 00:06:12.720
Perhaps in authoritarian states, other limits could be sort of the political purge and the political cycles where then the same people who were praised 10 years ago that then put aside or even imprisoned.
00:06:12.800 --> 00:06:16.800
Or we've seen that in the Soviet Union, we've seen that in China as well most recently.
00:06:17.040 --> 00:06:20.720
But if I can ask you a question is so where are we now?
00:06:20.959 --> 00:06:27.680
What are we in the phase where we should be betting into innovation or should we bet towards diffusion?
00:06:27.839 --> 00:06:31.839
Or is never really a single phase and there's a bit of both every time?
00:06:32.319 --> 00:06:35.199
I think the answer is there's a bit of both every time.
00:06:35.439 --> 00:06:37.680
A second question is who are we, right?
00:06:37.759 --> 00:06:47.680
So uh if if the focus is on you know artificial intelligence, for example, and the European Union, you could argue that we have some catching up to do.
00:06:47.839 --> 00:06:57.600
At the same time, we could also make the case that AI, impressive as it is, may not be quite as transformative yet to have a material impact on the economy, right?
00:06:57.680 --> 00:07:05.360
And so we're not seeing much of an impact on artificial intelligence either in the United States or, although it's harder to tell, in China either.
00:07:05.439 --> 00:07:07.519
So both are having productivity problems.
00:07:07.680 --> 00:07:12.319
So my view is that we need more innovation in AI to push forward.
00:07:12.480 --> 00:07:34.079
I tend to think that artificial intelligence needs its separate condenser moment, by which I mean that you know, if you go back to the first industrial revolution and the proliferation of steam power, in the early days, you know, uh steam engines were tremendously energy inefficient, they were only used to drain coal mines, and even that they didn't do particularly well.
00:07:34.240 --> 00:07:47.120
And so for steam to be applied to transportation, to rail, to steamships and so on, you needed further innovation and you needed the separate condenser in particular for um energy efficiency.
00:07:47.279 --> 00:07:57.600
With AI today, I think we need greater data and compute efficiency, and so humans are very good at picking up patterns from just a few examples.
00:07:57.759 --> 00:07:59.680
We don't need millions of trials.
00:07:59.839 --> 00:08:09.519
And so if the world was just a you know static distribution of events, then you could you know probably brute force AI, just throw more data and compute at the problem and you will get there, right?
00:08:09.600 --> 00:08:12.000
But the world is changing every time uh every day.
00:08:12.240 --> 00:08:20.639
And so the question always emerges: even if AI was accurate yesterday, is it's going to be accurate accurate today?
00:08:20.720 --> 00:08:24.720
And I think the the key example here is um Stuart Russell in his lab.
00:08:24.879 --> 00:08:28.639
Everybody knows about AlphaGo beating the world champion back in 2016.
00:08:28.800 --> 00:08:40.960
What few people know is that actually human amateurs using standard computers exposing the best Go programs to positions that they would not have encountered in training actually beat those programs uh relatively easily.
00:08:41.120 --> 00:08:58.720
And so that answers puts forward some some fundamental questions in the sense that even if AI has achieved superhuman performance in a domain, when exposed to new situations, novel circumstances that might no longer hold, and I think that resilience challenge uh challenge still persists.
00:08:59.039 --> 00:09:00.639
I have three questions.
00:09:00.799 --> 00:09:06.480
I'm gonna I'm gonna anticipate all three of them, but we're gonna go through one by one very, very slowly.
00:09:06.720 --> 00:09:11.519
So the first question and is linked to what you've just said, and it is about the tech competition, right?
00:09:11.679 --> 00:09:20.879
So is innovation that important in the sense of having the breakthrough in your country, let's say the United States rather than China, is that fundamental?
00:09:21.039 --> 00:09:36.240
Or at the end of the day, if indeed knowledge flows, and this is some is also part of your book, part of the conversation in your book, if knowledge flows, then where the breakthrough happens is not as important as how you diffuse or you adopt the technology.
00:09:36.399 --> 00:09:50.240
And I'm asking you specifically because there is this sort of framework going around that the United States is innovating, innovating, innovating, but China is thinking a lot more about industrial implementation and how to turn AI specifically into productivity.
00:09:50.320 --> 00:09:51.200
Let's use that word.
00:09:51.279 --> 00:09:54.639
And this is sort of question number one, very, very small question, right?
00:09:54.799 --> 00:10:16.720
And then question number two, I'll skip the number three, is if if indeed then where the breakthrough happens is not fundamental, then does that leave some hope also in the AI development for Europe, where no frontier company has been born, no frontier model, and and so on and so forth.
00:10:16.879 --> 00:10:18.080
So these are my two questions.
00:10:18.159 --> 00:10:21.039
And you we can go with the first one, which is US-China.
00:10:21.200 --> 00:10:23.120
How how do you see that laying out?
00:10:23.440 --> 00:10:33.360
So I do actually think that where the breakthrough happens matters, in large part because industries tend to emerge around those breakthroughs.
00:10:33.600 --> 00:10:44.159
The job hopping that then emerges from that then transfers a lot of tacit knowledge, gives rise to new businesses, and gives rise to dynamism more broadly.
00:10:44.320 --> 00:11:01.279
And so I don't think it's quite as simple as you know you have a breakthrough, and then you know that knowledge proliferates everywhere and every place adopts the technology, and if that was true, obviously every country, every society would be at the technological frontier and rich and prosperous, right?
00:11:01.360 --> 00:11:09.679
And so I I do think that being at the frontier matters because that is also where global talent tends to go.
00:11:09.840 --> 00:11:13.440
But you can obviously undermine that advantage as we're seeing right now as well.
00:11:13.679 --> 00:11:15.200
Secondly, uh on adoptions.
00:11:15.440 --> 00:11:22.480
I think it's absolutely true that it's when technology is becoming adopted that it's become beginning to have an impact on the economy.
00:11:22.639 --> 00:11:24.960
But I think adoption is also not one thing, right?
00:11:25.200 --> 00:11:27.840
Both you you and I we've adopted the computer, right?
00:11:27.919 --> 00:11:32.080
We we probably use it for a lot of similar things, we probably use it for different things as well, right?
00:11:32.240 --> 00:11:38.720
And so just by looking at computer adoption, well, it you know drives some differences in productivity across space.
00:11:38.879 --> 00:11:42.080
But in the end of the day, it's a question of what you use the technology for.
00:11:42.159 --> 00:11:44.879
Do you use it to develop new ideas, new businesses, etc.?
00:11:45.200 --> 00:11:47.679
Do you just use it to automate a few processes?
00:11:47.840 --> 00:11:51.840
And so I think the more relevant factor is dynamism.
00:11:51.919 --> 00:12:01.919
And I think dynamism is related to where these innovations happen, because that's where you get the inflow talent, that's why you get new firms emerging, you get clustering.
00:12:02.080 --> 00:12:05.360
And so I think that part is actually underrated.
00:12:05.440 --> 00:12:07.759
Um and I forgot the second part of your question, sorry about that.
00:12:08.000 --> 00:12:08.639
It's my fault.
00:12:08.720 --> 00:12:10.320
I always anticipate too many questions.
00:12:10.480 --> 00:12:12.399
The second part is about Europe, right?
00:12:12.480 --> 00:12:15.360
And I think that you've already in part already answered that question.
00:12:15.519 --> 00:12:21.200
If there is no breakthrough in Europe, does that mean that Europe is going to lag behind for a very long time?
00:12:21.440 --> 00:12:23.679
If yes, what we can do about it?
00:12:23.759 --> 00:12:38.000
And if no, what we can do about it in the sense of where Europe is now and how I'm not saying we should we should have a recipe, but what, according to your studies, what could be a way to catch up in in the current system?
00:12:38.639 --> 00:12:45.679
So clearly Europe has a tradition of being able to do catch-up growth pretty well.
00:12:45.759 --> 00:12:56.879
So it caught up to Britain in the 19th century, it closed the gap to the United States in the post-war period, but it hasn't really even closed the gap in digital.
00:12:56.960 --> 00:12:58.000
And so why is that?
00:12:58.159 --> 00:13:06.720
I think part of it has to do simply with the fact that the single market is less harmonized in digital and services than in goods.
00:13:06.960 --> 00:13:14.000
As a result of that, the market is more fragmented, and so the return to innovation in that space is simply lower.
00:13:14.159 --> 00:13:28.799
I think when it comes to trading services, language and culture matters more, and so that is inevitably going to be a bit of a drag um on Europe when it comes to scaling, although that may have other advantages.
00:13:28.960 --> 00:13:44.559
I think the costs of experimentation are higher in Europe, uh, in large part because of labor protection, which has some upsides, but it may mean that higher and fire is harder, and that means that it's more costly to experiment and pivot when a new technology arrives.
00:13:44.639 --> 00:13:55.519
So Meta can just you know pivot to the metaverse and then just you know reverse course and then pivot towards generative AI, whereas a firm and SP will struggle for years to restructure, and for example.
00:13:55.679 --> 00:14:07.039
I think barriers to entry are higher in part because of rules and regulations like the GDPR, which have introduced compliance costs which make it particularly difficult for younger firms entering the market.
00:14:07.200 --> 00:14:17.679
I think those things together probably can go some way in explaining why Europe hasn't uh caught up in digital the way it's caught up in mass production.
00:14:17.840 --> 00:14:19.200
So, what does Europe do?
00:14:19.279 --> 00:14:22.320
Well, I think there are sort of two broad school of thoughts out there.
00:14:22.480 --> 00:14:31.679
One is sort of more related to top-down industrial policy, and you will hear, you know, Airbus as the lead example, and that is a case in point.
00:14:32.000 --> 00:14:37.679
I think at the same time, it's also true that aircraft technology was fairly static at the time.
00:14:37.840 --> 00:14:41.919
The jet engine was the last sort of big breakthrough that had already been developed.
00:14:42.000 --> 00:14:46.960
And so Airbus was sort of catching up to a fairly static target, being Boeing.
00:14:47.120 --> 00:14:53.840
There were numerous policy initiatives to seed domestic semiconductor industries, which essentially failed.
00:14:54.000 --> 00:14:58.399
I think part of the reason being that semiconductors are more dynamic, it's a moving target.
00:14:58.559 --> 00:15:00.559
And I think that's even more true of AI.
00:15:00.639 --> 00:15:10.000
I mean, we don't know if the future of AI is large language models, small language models, world models, some fusion of you know symbol symbolic uh AI with something else.
00:15:10.159 --> 00:15:15.759
And so I think that's a process we will figure uh something we will have to figure out through trial and error.
00:15:15.840 --> 00:15:21.200
And that means that we need to, you know, take a fairly diversified set of bets.
00:15:21.360 --> 00:15:31.519
And for those bets to be taken, I think we need to sort of solve some of the underlying challenges, such as the lack of a harmonized single market in services, for example.
00:15:31.759 --> 00:15:46.159
I'm very glad that I didn't even suggest it, and uh a more unified or harmonized single market was your first answer, is something quite recurrent in several episodes, is my personal thought, but I never try I always try not to bias the question.
00:15:46.240 --> 00:15:48.799
So to the listeners, this was genuinely from Carl.
00:15:48.879 --> 00:15:51.200
It wasn't it wasn't me or any of the other speakers.
00:15:51.440 --> 00:15:52.559
I'm gonna let you go very soon.
00:15:52.639 --> 00:15:57.200
So I'll close with the last question, which is a little bit more broad and about AI.
00:15:57.360 --> 00:16:10.399
The book ends on a note of uncertainty on what the impact of AI on progress uh could be and whether it will depend on how institutions foster their culture, openness, control.
00:16:10.720 --> 00:16:20.879
But if you had to place a bet right now, does the world end in a new golden age of progress, or are we at the beginning of what could be great stagnation?
00:16:21.039 --> 00:16:27.039
Now I realize you don't have a crystal ball, but at the same time is is is a fun question to ask at the end of this episode.
00:16:27.360 --> 00:16:36.080
So I think my base case would be something like the computer revolution, but a bit more moderated in terms of its impact on productivity.
00:16:36.159 --> 00:16:44.159
So with the computer revolution, what you got was sort of basically productivity upsurge over around a decade, mostly confined uh to the United States.
00:16:44.320 --> 00:16:50.799
I would be surprised if we don't see AI in the productivity statistics at all a decade or two from from now.
00:16:50.960 --> 00:16:52.240
I think it will show.
00:16:52.399 --> 00:16:57.840
But I also think that certain challenges that we currently have will persist for some time.
00:16:58.080 --> 00:17:00.399
One of those challenges is verification, right?
00:17:00.480 --> 00:17:05.359
And so, you know, uh with computers, they were more predictable in terms of what they produced.
00:17:05.519 --> 00:17:13.759
If they they wouldn't fabricate sources and that sort of thing, they you know certainly had some challenges, but the error was usually human error.
00:17:13.920 --> 00:17:29.440
So with artificial intelligence, I think even as it becomes better, it's going to be harder and harder for us to verify the output because only fairly specialist individuals will be able to tell truth from falsehood, right?
00:17:29.519 --> 00:17:34.640
And so that means that we will have to learn with the algorithms in order to verify the outputs.
00:17:34.799 --> 00:17:44.880
And so in the end of the day, the productivity gains will be, you know, time saved doing the research or whatever it is you do, minus verification.
00:17:45.039 --> 00:17:56.160
And so, you know, the computer revolution essentially gave us instant access to a lot of knowledge, which save hours, you know, researching the library or you know, waiting for new material to arrive.
00:17:56.240 --> 00:17:58.319
And so it's essentially automated downtime.
00:17:58.559 --> 00:18:02.400
AI automates cognition and it still requires the human in the loop.
00:18:02.480 --> 00:18:11.039
And as long as that's the case, I think the productivity gains will be more muted than what we saw during the ICT revolution, but I still think that's the closest parallel.
00:18:11.359 --> 00:18:15.519
And this is a great ending, uh, especially because uh is interesting.
00:18:15.680 --> 00:18:25.839
Well, the impact on the knowledge sector, if so we want to call it, is quite interesting how universities think tanks research centers will change, both in terms of examination.
00:18:26.000 --> 00:18:35.920
I was reading something about, you know, a PhD thesis being submitted at a level that was uh completely different from the performance of the student throughout the four years that preceded.
00:18:36.079 --> 00:18:52.240
So is the output no longer going to be the way we judge the production of a student because it could change, but also how the organic, so how the staff is organized, are we going to be needing a lot more specialized people with a lot of expertise that can check the content?
00:18:52.400 --> 00:18:59.440
Of course, now it's relatively early, but I find it quite funny that you start seeing on LinkedIn and around all these positions for code verifiers.
00:18:59.680 --> 00:19:14.720
So because now coding has been automatized just so quickly and everyone was so enthusiastic about it, and then suddenly you have a whole new role, which is okay, yeah, the machine is producing this code, but you'll have you'll need a human being to verify whether that is uh is is right.
00:19:14.880 --> 00:19:19.599
So it's an open question whether all of these will create more productivity or just a different system.
00:19:19.680 --> 00:19:22.480
But I think it's a great question to leave the audience with.
00:19:22.720 --> 00:19:26.720
So, Carl, thank you so much for taking the time of to have this conversation with us.
00:19:26.880 --> 00:19:27.680
Thank you for joining me.
00:00:07.679 --> 00:00:09.439
Welcome to Geoconomic Competition.
00:00:09.519 --> 00:00:15.119
I'm Francesca Guerti, your host, and in every episode I dive into the intricacies of geoconomics with other experts.
00:00:15.279 --> 00:00:20.559
You might have noticed we haven't really done an episode on tech or on AI yet.
00:00:20.719 --> 00:00:33.359
So today I have Carl Benedict Frey, Dieter Schwartz Associate Professor of AI and work at the Oxford Internet Institute, but also the author of The Technology Trap and of How Progress Ends.
00:00:33.600 --> 00:00:37.200
Two beautiful, fantastic books that I recommend you go by and read.
00:00:37.359 --> 00:00:39.119
Thank you for being here with me today, Carl.
00:00:39.359 --> 00:00:40.079
Good to be with you.
00:00:40.320 --> 00:00:43.039
There are many very interesting things in your latest book.
00:00:43.200 --> 00:00:51.359
And one of the most interesting framings, at least for me, is that society needs two engines that often are treated as somewhat competing.
00:00:51.439 --> 00:00:56.880
And one is the decentralized exploration to generate new ideas, we could call it somewhat innovation.
00:00:57.039 --> 00:01:03.759
And the other one is centralized exploitation to scale them, adopt them, diffuse them, you you pick the term.
00:01:03.920 --> 00:01:11.840
Can you walk us through how that tension works and especially why getting the balance wrong has been fatal for so many great powers?
00:01:12.239 --> 00:01:23.200
So if you run an assembly line or if you run a factory and every single worker on that assembly line questions every single procedure, you're not going to produce too much, right?
00:01:23.439 --> 00:01:28.560
But if nobody questions anything, you're also not going to invent anything new.
00:01:28.719 --> 00:01:32.959
And sort of that's the tension I'm I'm trying to explore in the book.
00:01:33.120 --> 00:01:48.319
And the key point, as you mentioned, is that if you want to explore new things, a certain degree of autonomy, different people exploring different technological trajectories is usually important to come up with new ideas, new technologies.
00:01:48.480 --> 00:02:09.039
But when you then settled on what it is you want to do, and when you want to produce and make your goods or services available to as many people as possible, you want to cut costs, you want to exploit new technologies like factories, that then leads you towards managerial hierarchies and greater centralization.
00:02:09.120 --> 00:02:15.439
And so both are needed for progress, but different stages of the technology life cycle, if you like.
00:02:15.840 --> 00:02:20.960
And we get to what stage we're in and who's doing these best in in a moment.
00:02:21.199 --> 00:02:36.159
Part of your book is tracing the history and the experiences of some of these, let's call them actors' great power, whether it is some dynasty in China, the Dutch Republic, Victorian Britain, Soviet Union, and so on and so forth, but also modern Japan, and we'll get to that in a moment.
00:02:36.479 --> 00:02:41.439
And you notice how many of these surged ahead and then stalled, right?
00:02:41.599 --> 00:02:52.400
So is there a common pattern in how the transition from the rise to stagnation plays out and perhaps some lessons learned or elements we should keep our eyes open for?
00:02:52.719 --> 00:03:00.960
So the case I make in the book is that moving between decentralization and centralization means that you need some institutional flexibility.
00:03:01.039 --> 00:03:12.800
And that is hard to do over the long run because you need to change your institutions and you often have incumbent interests that have a stake in the exist existing order of things.
00:03:12.960 --> 00:03:17.520
They are often going to resist reform and new technologies as well.
00:03:17.759 --> 00:03:24.479
But you can grow quite quickly for a fairly long period of time without being at the forefront of technology.
00:03:24.560 --> 00:03:39.520
And so if you take the Soviet Union, for example, we tend to look back at it as a failure now, and obviously in many ways it was, but it was also growing at six percent per year over a period of around four decades or so, which is quite respectable.
00:03:39.680 --> 00:04:00.960
And it did so basically by taking technology invented elsewhere and scaling it domestically, and so it took ample advantage of the Ford Motor Company's open door policy, sent delegations to Detroit to look at uh the Ford factories, learn from that, then build a vehicle industry in the Soviet Union.
00:04:01.039 --> 00:04:10.960
And in the age of mass production, the authorities could could hold factory uh managers accountable by simply benchmarking a performance across plants.
00:04:11.120 --> 00:04:20.399
And so it could that way determine if some factories were slacking off and not producing as much as maybe should compared to its peers.
00:04:20.560 --> 00:04:24.319
But at the frontiers of technology, that becomes much harder, right?
00:04:24.560 --> 00:04:28.720
Because when something is new, what do you benchmark against?
00:04:28.959 --> 00:04:40.079
In addition to that, what what the Soviet Union needed for further growth as the mass production system petered out globally was something new.
00:04:40.240 --> 00:04:45.920
That something new was the computer revolution, to which Soviet contributions were essentially none.
00:04:46.000 --> 00:04:54.000
And a big part of that was that if you were an inventor in the Soviet Union, you could go to, let's say, the Red Army and ask for funding.
00:04:54.160 --> 00:04:57.279
If they declined, maybe two or three other options.
00:04:57.439 --> 00:05:01.680
If they declined, then basically your your idea would die with you.
00:05:01.839 --> 00:05:13.120
That's quite different from the American system of much more decentralized finance where Besma Venture famously declined to invest in Google back in 1999 and probably regret it today.
00:05:13.279 --> 00:05:18.800
But it also illustrates that in in the late 90s Google was not a safe bet, right?
00:05:18.959 --> 00:05:27.680
Alta Vista and Yahoo were dominating such, and so somebody needed to take the risk to invest in order to know if Google would catch on.
00:05:27.759 --> 00:05:32.319
And the fact that Bessmer didn't invest didn't mean the end of Google because others stepped in.
00:05:32.399 --> 00:05:40.079
And so in a more decentralized system, you have different players that can take different bets, you can explore more technological pathways that way.
00:05:40.240 --> 00:05:50.720
And so that's more important at the technological frontier when it comes to pushing innovation and where the opportunities for simply adopting technologies invented invented elsewhere diminishes.
00:05:51.040 --> 00:05:51.439
Alright.
00:05:51.600 --> 00:06:00.639
And the other side of the spectrum is sort of the adoption part, which you know you've already mentioned in relation to the Soviet Union, you've already mentioned the limits.
00:06:00.800 --> 00:06:12.720
Perhaps in authoritarian states, other limits could be sort of the political purge and the political cycles where then the same people who were praised 10 years ago that then put aside or even imprisoned.
00:06:12.800 --> 00:06:16.800
Or we've seen that in the Soviet Union, we've seen that in China as well most recently.
00:06:17.040 --> 00:06:20.720
But if I can ask you a question is so where are we now?
00:06:20.959 --> 00:06:27.680
What are we in the phase where we should be betting into innovation or should we bet towards diffusion?
00:06:27.839 --> 00:06:31.839
Or is never really a single phase and there's a bit of both every time?
00:06:32.319 --> 00:06:35.199
I think the answer is there's a bit of both every time.
00:06:35.439 --> 00:06:37.680
A second question is who are we, right?
00:06:37.759 --> 00:06:47.680
So uh if if the focus is on you know artificial intelligence, for example, and the European Union, you could argue that we have some catching up to do.
00:06:47.839 --> 00:06:57.600
At the same time, we could also make the case that AI, impressive as it is, may not be quite as transformative yet to have a material impact on the economy, right?
00:06:57.680 --> 00:07:05.360
And so we're not seeing much of an impact on artificial intelligence either in the United States or, although it's harder to tell, in China either.
00:07:05.439 --> 00:07:07.519
So both are having productivity problems.
00:07:07.680 --> 00:07:12.319
So my view is that we need more innovation in AI to push forward.
00:07:12.480 --> 00:07:34.079
I tend to think that artificial intelligence needs its separate condenser moment, by which I mean that you know, if you go back to the first industrial revolution and the proliferation of steam power, in the early days, you know, uh steam engines were tremendously energy inefficient, they were only used to drain coal mines, and even that they didn't do particularly well.
00:07:34.240 --> 00:07:47.120
And so for steam to be applied to transportation, to rail, to steamships and so on, you needed further innovation and you needed the separate condenser in particular for um energy efficiency.
00:07:47.279 --> 00:07:57.600
With AI today, I think we need greater data and compute efficiency, and so humans are very good at picking up patterns from just a few examples.
00:07:57.759 --> 00:07:59.680
We don't need millions of trials.
00:07:59.839 --> 00:08:09.519
And so if the world was just a you know static distribution of events, then you could you know probably brute force AI, just throw more data and compute at the problem and you will get there, right?
00:08:09.600 --> 00:08:12.000
But the world is changing every time uh every day.
00:08:12.240 --> 00:08:20.639
And so the question always emerges: even if AI was accurate yesterday, is it's going to be accurate accurate today?
00:08:20.720 --> 00:08:24.720
And I think the the key example here is um Stuart Russell in his lab.
00:08:24.879 --> 00:08:28.639
Everybody knows about AlphaGo beating the world champion back in 2016.
00:08:28.800 --> 00:08:40.960
What few people know is that actually human amateurs using standard computers exposing the best Go programs to positions that they would not have encountered in training actually beat those programs uh relatively easily.
00:08:41.120 --> 00:08:58.720
And so that answers puts forward some some fundamental questions in the sense that even if AI has achieved superhuman performance in a domain, when exposed to new situations, novel circumstances that might no longer hold, and I think that resilience challenge uh challenge still persists.
00:08:59.039 --> 00:09:00.639
I have three questions.
00:09:00.799 --> 00:09:06.480
I'm gonna I'm gonna anticipate all three of them, but we're gonna go through one by one very, very slowly.
00:09:06.720 --> 00:09:11.519
So the first question and is linked to what you've just said, and it is about the tech competition, right?
00:09:11.679 --> 00:09:20.879
So is innovation that important in the sense of having the breakthrough in your country, let's say the United States rather than China, is that fundamental?
00:09:21.039 --> 00:09:36.240
Or at the end of the day, if indeed knowledge flows, and this is some is also part of your book, part of the conversation in your book, if knowledge flows, then where the breakthrough happens is not as important as how you diffuse or you adopt the technology.
00:09:36.399 --> 00:09:50.240
And I'm asking you specifically because there is this sort of framework going around that the United States is innovating, innovating, innovating, but China is thinking a lot more about industrial implementation and how to turn AI specifically into productivity.
00:09:50.320 --> 00:09:51.200
Let's use that word.
00:09:51.279 --> 00:09:54.639
And this is sort of question number one, very, very small question, right?
00:09:54.799 --> 00:10:16.720
And then question number two, I'll skip the number three, is if if indeed then where the breakthrough happens is not fundamental, then does that leave some hope also in the AI development for Europe, where no frontier company has been born, no frontier model, and and so on and so forth.
00:10:16.879 --> 00:10:18.080
So these are my two questions.
00:10:18.159 --> 00:10:21.039
And you we can go with the first one, which is US-China.
00:10:21.200 --> 00:10:23.120
How how do you see that laying out?
00:10:23.440 --> 00:10:33.360
So I do actually think that where the breakthrough happens matters, in large part because industries tend to emerge around those breakthroughs.
00:10:33.600 --> 00:10:44.159
The job hopping that then emerges from that then transfers a lot of tacit knowledge, gives rise to new businesses, and gives rise to dynamism more broadly.
00:10:44.320 --> 00:11:01.279
And so I don't think it's quite as simple as you know you have a breakthrough, and then you know that knowledge proliferates everywhere and every place adopts the technology, and if that was true, obviously every country, every society would be at the technological frontier and rich and prosperous, right?
00:11:01.360 --> 00:11:09.679
And so I I do think that being at the frontier matters because that is also where global talent tends to go.
00:11:09.840 --> 00:11:13.440
But you can obviously undermine that advantage as we're seeing right now as well.
00:11:13.679 --> 00:11:15.200
Secondly, uh on adoptions.
00:11:15.440 --> 00:11:22.480
I think it's absolutely true that it's when technology is becoming adopted that it's become beginning to have an impact on the economy.
00:11:22.639 --> 00:11:24.960
But I think adoption is also not one thing, right?
00:11:25.200 --> 00:11:27.840
Both you you and I we've adopted the computer, right?
00:11:27.919 --> 00:11:32.080
We we probably use it for a lot of similar things, we probably use it for different things as well, right?
00:11:32.240 --> 00:11:38.720
And so just by looking at computer adoption, well, it you know drives some differences in productivity across space.
00:11:38.879 --> 00:11:42.080
But in the end of the day, it's a question of what you use the technology for.
00:11:42.159 --> 00:11:44.879
Do you use it to develop new ideas, new businesses, etc.?
00:11:45.200 --> 00:11:47.679
Do you just use it to automate a few processes?
00:11:47.840 --> 00:11:51.840
And so I think the more relevant factor is dynamism.
00:11:51.919 --> 00:12:01.919
And I think dynamism is related to where these innovations happen, because that's where you get the inflow talent, that's why you get new firms emerging, you get clustering.
00:12:02.080 --> 00:12:05.360
And so I think that part is actually underrated.
00:12:05.440 --> 00:12:07.759
Um and I forgot the second part of your question, sorry about that.
00:12:08.000 --> 00:12:08.639
It's my fault.
00:12:08.720 --> 00:12:10.320
I always anticipate too many questions.
00:12:10.480 --> 00:12:12.399
The second part is about Europe, right?
00:12:12.480 --> 00:12:15.360
And I think that you've already in part already answered that question.
00:12:15.519 --> 00:12:21.200
If there is no breakthrough in Europe, does that mean that Europe is going to lag behind for a very long time?
00:12:21.440 --> 00:12:23.679
If yes, what we can do about it?
00:12:23.759 --> 00:12:38.000
And if no, what we can do about it in the sense of where Europe is now and how I'm not saying we should we should have a recipe, but what, according to your studies, what could be a way to catch up in in the current system?
00:12:38.639 --> 00:12:45.679
So clearly Europe has a tradition of being able to do catch-up growth pretty well.
00:12:45.759 --> 00:12:56.879
So it caught up to Britain in the 19th century, it closed the gap to the United States in the post-war period, but it hasn't really even closed the gap in digital.
00:12:56.960 --> 00:12:58.000
And so why is that?
00:12:58.159 --> 00:13:06.720
I think part of it has to do simply with the fact that the single market is less harmonized in digital and services than in goods.
00:13:06.960 --> 00:13:14.000
As a result of that, the market is more fragmented, and so the return to innovation in that space is simply lower.
00:13:14.159 --> 00:13:28.799
I think when it comes to trading services, language and culture matters more, and so that is inevitably going to be a bit of a drag um on Europe when it comes to scaling, although that may have other advantages.
00:13:28.960 --> 00:13:44.559
I think the costs of experimentation are higher in Europe, uh, in large part because of labor protection, which has some upsides, but it may mean that higher and fire is harder, and that means that it's more costly to experiment and pivot when a new technology arrives.
00:13:44.639 --> 00:13:55.519
So Meta can just you know pivot to the metaverse and then just you know reverse course and then pivot towards generative AI, whereas a firm and SP will struggle for years to restructure, and for example.
00:13:55.679 --> 00:14:07.039
I think barriers to entry are higher in part because of rules and regulations like the GDPR, which have introduced compliance costs which make it particularly difficult for younger firms entering the market.
00:14:07.200 --> 00:14:17.679
I think those things together probably can go some way in explaining why Europe hasn't uh caught up in digital the way it's caught up in mass production.
00:14:17.840 --> 00:14:19.200
So, what does Europe do?
00:14:19.279 --> 00:14:22.320
Well, I think there are sort of two broad school of thoughts out there.
00:14:22.480 --> 00:14:31.679
One is sort of more related to top-down industrial policy, and you will hear, you know, Airbus as the lead example, and that is a case in point.
00:14:32.000 --> 00:14:37.679
I think at the same time, it's also true that aircraft technology was fairly static at the time.
00:14:37.840 --> 00:14:41.919
The jet engine was the last sort of big breakthrough that had already been developed.
00:14:42.000 --> 00:14:46.960
And so Airbus was sort of catching up to a fairly static target, being Boeing.
00:14:47.120 --> 00:14:53.840
There were numerous policy initiatives to seed domestic semiconductor industries, which essentially failed.
00:14:54.000 --> 00:14:58.399
I think part of the reason being that semiconductors are more dynamic, it's a moving target.
00:14:58.559 --> 00:15:00.559
And I think that's even more true of AI.
00:15:00.639 --> 00:15:10.000
I mean, we don't know if the future of AI is large language models, small language models, world models, some fusion of you know symbol symbolic uh AI with something else.
00:15:10.159 --> 00:15:15.759
And so I think that's a process we will figure uh something we will have to figure out through trial and error.
00:15:15.840 --> 00:15:21.200
And that means that we need to, you know, take a fairly diversified set of bets.
00:15:21.360 --> 00:15:31.519
And for those bets to be taken, I think we need to sort of solve some of the underlying challenges, such as the lack of a harmonized single market in services, for example.
00:15:31.759 --> 00:15:46.159
I'm very glad that I didn't even suggest it, and uh a more unified or harmonized single market was your first answer, is something quite recurrent in several episodes, is my personal thought, but I never try I always try not to bias the question.
00:15:46.240 --> 00:15:48.799
So to the listeners, this was genuinely from Carl.
00:15:48.879 --> 00:15:51.200
It wasn't it wasn't me or any of the other speakers.
00:15:51.440 --> 00:15:52.559
I'm gonna let you go very soon.
00:15:52.639 --> 00:15:57.200
So I'll close with the last question, which is a little bit more broad and about AI.
00:15:57.360 --> 00:16:10.399
The book ends on a note of uncertainty on what the impact of AI on progress uh could be and whether it will depend on how institutions foster their culture, openness, control.
00:16:10.720 --> 00:16:20.879
But if you had to place a bet right now, does the world end in a new golden age of progress, or are we at the beginning of what could be great stagnation?
00:16:21.039 --> 00:16:27.039
Now I realize you don't have a crystal ball, but at the same time is is is a fun question to ask at the end of this episode.
00:16:27.360 --> 00:16:36.080
So I think my base case would be something like the computer revolution, but a bit more moderated in terms of its impact on productivity.
00:16:36.159 --> 00:16:44.159
So with the computer revolution, what you got was sort of basically productivity upsurge over around a decade, mostly confined uh to the United States.
00:16:44.320 --> 00:16:50.799
I would be surprised if we don't see AI in the productivity statistics at all a decade or two from from now.
00:16:50.960 --> 00:16:52.240
I think it will show.
00:16:52.399 --> 00:16:57.840
But I also think that certain challenges that we currently have will persist for some time.
00:16:58.080 --> 00:17:00.399
One of those challenges is verification, right?
00:17:00.480 --> 00:17:05.359
And so, you know, uh with computers, they were more predictable in terms of what they produced.
00:17:05.519 --> 00:17:13.759
If they they wouldn't fabricate sources and that sort of thing, they you know certainly had some challenges, but the error was usually human error.
00:17:13.920 --> 00:17:29.440
So with artificial intelligence, I think even as it becomes better, it's going to be harder and harder for us to verify the output because only fairly specialist individuals will be able to tell truth from falsehood, right?
00:17:29.519 --> 00:17:34.640
And so that means that we will have to learn with the algorithms in order to verify the outputs.
00:17:34.799 --> 00:17:44.880
And so in the end of the day, the productivity gains will be, you know, time saved doing the research or whatever it is you do, minus verification.
00:17:45.039 --> 00:17:56.160
And so, you know, the computer revolution essentially gave us instant access to a lot of knowledge, which save hours, you know, researching the library or you know, waiting for new material to arrive.
00:17:56.240 --> 00:17:58.319
And so it's essentially automated downtime.
00:17:58.559 --> 00:18:02.400
AI automates cognition and it still requires the human in the loop.
00:18:02.480 --> 00:18:11.039
And as long as that's the case, I think the productivity gains will be more muted than what we saw during the ICT revolution, but I still think that's the closest parallel.
00:18:11.359 --> 00:18:15.519
And this is a great ending, uh, especially because uh is interesting.
00:18:15.680 --> 00:18:25.839
Well, the impact on the knowledge sector, if so we want to call it, is quite interesting how universities think tanks research centers will change, both in terms of examination.
00:18:26.000 --> 00:18:35.920
I was reading something about, you know, a PhD thesis being submitted at a level that was uh completely different from the performance of the student throughout the four years that preceded.
00:18:36.079 --> 00:18:52.240
So is the output no longer going to be the way we judge the production of a student because it could change, but also how the organic, so how the staff is organized, are we going to be needing a lot more specialized people with a lot of expertise that can check the content?
00:18:52.400 --> 00:18:59.440
Of course, now it's relatively early, but I find it quite funny that you start seeing on LinkedIn and around all these positions for code verifiers.
00:18:59.680 --> 00:19:14.720
So because now coding has been automatized just so quickly and everyone was so enthusiastic about it, and then suddenly you have a whole new role, which is okay, yeah, the machine is producing this code, but you'll have you'll need a human being to verify whether that is uh is is right.
00:19:14.880 --> 00:19:19.599
So it's an open question whether all of these will create more productivity or just a different system.
00:19:19.680 --> 00:19:22.480
But I think it's a great question to leave the audience with.
00:19:22.720 --> 00:19:26.720
So, Carl, thank you so much for taking the time of to have this conversation with us.
00:19:26.880 --> 00:19:27.680
Thank you for joining me.
