YT YouTube2Text

YouTube2Text → Transcripts → „In 5 Jahren sind ALLE Bürojobs weg!" DAS passiert danach (Prof. Marcus Hutter)

„In 5 Jahren sind ALLE Bürojobs weg!" DAS passiert danach (Prof. Marcus Hutter) — Transcript

by Everlast AI · 10,897 words · 1,020 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00So, technologically, I think that within two to five years AI can really
0:05replace all office jobs.
0:06That is my prediction: that jobs will disappear and not be replaced by new jobs.
0:11If new jobs emerge, they can be automated instantly too,
0:16because if politicians don't get this right, this situation can really
0:19come about. Without political measures, the unemployment rate will rise higher and higher.
0:23Uprisings, demonstrations, and in the worst case civil war, if nothing is done.
0:27Maybe a few deaths, and maybe that is enough for politicians to wake up.
0:32I have no argument, or no compelling argument, for why it will be different
0:36this time. This future may arrive faster than you think.
0:39If a large share of people become unemployed, who is still going to buy anything?
0:47Today I have the pleasure of speaking once again with Prof.
0:49Dr. Marcus Hutter.
0:51Marcus Hutter studied superintelligence mathematically a quarter of a century ago
0:54with AIXI, the model of the most intelligent agent that can theoretically
0:59exist. In 2006 he established the Hutter Prize, which is now endowed with 500,000 euros
1:04and rewards progress in text compression, under the assumption that
1:08compression is nothing other than intelligence.
1:11Professor Marcus Hutter is thus one of the most distinguished minds in AI research,
1:15and has been for a long time before the topic occupied all of us.
1:18Marcus, you have now published a new book, Jobless Utopia, Macroeconomics in the
1:23Age of AGI. And right at the start you pose the question: will
1:28automation destroy livelihoods or create prosperity?
1:32The answer depends entirely on politics.
1:35Will the wealth that machines create
1:37be shared? Then the result is liberation.
1:39If it is not shared, the result is an avoidable dystopia of concentrated
1:45wealth and mass precarity, you write.
1:48Before we get into all the depth, how does it even come about that you, as an AI
1:53researcher, deal with these economic and societal questions?
1:58Well, even as an AI researcher, theorist, mathematician, you are of course also embedded in
2:04society and interested in it,
2:08and you read the newspaper or the news.
2:12And one discussion that bothered me a lot,
2:19that keeps coming up again and again, is the discussion about jobs.
2:24And in the various blogs and on Twitter you naturally find all sorts of possibly
2:27unqualified opinions, which I can ignore.
2:31But when experts in the field, economists, write about it
2:37in a way that I think is not correct, I thought, okay, I mean,
2:43we studied economics, maybe I'm wrong, and I read up on it
2:48and found that, at least so far, I am still
2:54of the opinion that most economists, and we will get to
2:57what statements they make, are wrong, and I say most, there are
3:03always exceptions of course, and it is my prediction that the jobs disappear and
3:09are not replaced by new jobs, that this is the correct prediction.
3:13And originally I just intended to write a small blog post,
3:17which then became, so to speak, a small paper, and then it became
3:20a long paper, and then at some point it got too long for a paper.
3:23And then I thought, I'll do it.
3:25So this book is more or less an accident.
3:29Yes, but it is really very interesting, because Jobless Utopia is
3:34really very contrary to the current debate and the current opinions
3:39circulating in the market. Can you put that into context first?
3:42You also describe four camps in the current discussion.
3:45What are these camps and opinions that are currently circulating around
3:50job losses due to AI? Yes, that was indeed the starting point.
3:53There are people who think, yes, many jobs will disappear.
3:57I mean, in the past, a long time ago, we had no jobs at all, we
4:01were simply hunter-gatherers, you could call that a job.
4:04And then we became farmers.
4:07And then industrial workers and then office jobs,
4:11and by now YouTube influencer is a job, right, which 50
4:17years ago would definitely have raised eyebrows.
4:27Most economists assume that the AI revolution
4:30will be similar. Many jobs will be automated, but new jobs will emerge.
4:35So there is one camp, jobs remain, and the other camp, jobs
4:41disappear, and then on the other side, we could have utopia or dystopia or
4:46stagnation. I have already mentioned one camp, these are the
4:50economists.
4:51Then there is the camp that says, well, AI is good, but it will never
4:56be able to replace humans, because we have special abilities, and robotics will never
5:00be good enough, because our hands are amazingly good, which I agree with.
5:05They are amazingly good, and so far robotics is not yet able to
5:09simulate hands even halfway.
5:11Don't worry. Jobs will remain.
5:14AI is in the distant future or will never come.
5:17That is the second camp. Then there is the general working population, which
5:23is more and more afraid of losing their jobs, and with that their income, and with that
5:29their standard of living deteriorating significantly.
5:32And this fear is justified, because if politicians don't get this right,
5:38as we will discuss later, this situation can really
5:41come about. And then there is the seemingly small minority, I'll call them the
5:46philosophers, to which I belong, well, I'm not a philosopher, but a part-time philosopher,
5:54who is of the opinion that all jobs can be replaced through automation.
6:02If new jobs emerge, they can be automated instantly too, so there
6:07are not really any new jobs for humans.
6:09And thus more and more people become unemployed. It won't happen in
6:13one big bang, but gradually, though it can happen relatively fast.
6:18But in a democracy, which is self-regulating, something like
6:25Universal Basic Income will be established, so that everyone continues to receive an income from
6:30the state and, on average at least, does not see their standard of living worsen, and
6:35probably even greatly improve in the long term.
6:39Yes, before we talk about your hypothesis, let's really explain the three camps
6:44in depth once more, because I think everyone out there
6:49follows one of these directions, at least subconsciously, implicitly, but
6:53very few follow yours.
6:55And the first is this continuity, so technology has always created new jobs.
6:59The second, as you say, mass unemployment and societal
7:04collapse. And the third is the stagnation bet.
7:07If we look at number three once more, this stagnation bet,
7:12so the idea that everything is simply exaggerated, there is a lot of hype around AGI, and that
7:16goes somewhat hand in hand with number one.
7:18And I think that is circulating a lot on the internet right now.
7:22You define 13 theses in your book, and the first thesis is:
7:25full automation is inevitable.
7:28So it is inevitable that AGI and robots will at some point do practically all
7:32cognitive and physical tasks better and cheaper.
7:36Why is that the case?
7:37I think most people are not at all aware yet of why
7:40this time is different from earlier industrial revolutions.
7:45Yes, technologically, we see it in the progress, especially of the last four
7:50years with ChatGPT and the other language models.
7:53Cognitive abilities are already, on average, beyond the typical
7:58human. Specialists like mathematicians still have an edge, but that will also
8:03go away in a few years, provided progress of course continues,
8:07and at the moment it looks that way.
8:11And once we have reached ASI, so artificial superintelligence, ASI will
8:18drive research and development further in robotics, and then robotic capabilities
8:22will also be improved relatively quickly by AI.
8:27And then the physical jobs can be replaced too.
8:32Maybe, as is always said, the plumber is perhaps the last, I'd say
8:34surgeon, surgeon is perhaps the last job to be replaced.
8:39That is from the technological aspect.
8:41Even if AI makes no further progress and stays as it
8:46is, I would expect that probably 50% of office jobs would be replaced now.
8:52But of course there are also the mechanical jobs.
8:54And if only 50 percent are replaced, you can somehow fill the other 50 percent
8:57with bullshit jobs or whatever.
9:01We'll get to that later.
9:03So I think we do still need the progress to really
9:05reach mass unemployment.
9:09Theoretically, of course, one could try to halt technological progress politically.
9:14I mean, more or less that was achieved with genetic research
9:20on embryonic cells. It is strongly regulated worldwide, and I don't necessarily know
9:27whether that is more technologically difficult, but I mean, when governments
9:32limit it, there is little funding.
9:36So sometimes it is possible to stop developments.
9:41At the moment it looks like that is unlikely, because the benefits of AI
9:47and robotics, the economic ones, are simply too great to
9:53stop the progress. Perhaps it can be slowed down.
9:59Over what time horizon are you thinking here?
10:02We also talked about exactly this last time, so which paths lead
10:06to ASI and why is it inevitable?
10:09And so over what time horizon do we have to think here, for these
10:13existential questions, these societal questions that you are now dealing with,
10:16to really
10:19become broadly relevant and for a large share of people to lose their jobs?
10:24Whether it's 50 percent, 20 percent might perhaps already be enough
10:28for us to be facing a societal upheaval, and over what time horizon
10:34are you thinking about this? Well, technologically I think
10:38that AI can, in two to five years,
10:43really replace all office jobs.
10:47And on the mechanical side, so physical jobs, perhaps another five to ten
10:54years, maybe more like ten years, although that is hard to estimate.
10:58So normal technological development I would say ten years, but if ASI
11:01takes over the development, it can of course be very rapid.
11:04You can, so to speak, create the blueprints for the perfect robot in two months.
11:07And then of course you still have to build the factories and so on, which also takes
11:11time. So theoretically it could be five years.
11:13So if you average it out, between, what did I say, between seven and
11:1715 years. Technologically, and then of course there is the question of how fast it diffuses
11:21into society.
11:24And in the industries that are less regulated and where there is a lot of competition,
11:31it will happen relatively quickly, because the companies that don't automate simply
11:35are no longer competitive.
11:38In the industries that are more regulated, like medicine or also law,
11:45it may be that legislation prevents quick
11:51automation, privacy, safety and so on.
11:56There, in those industries, it is then more or less international
12:01competitiveness: if one country manages to, so to speak, replace prime ministers
12:06with AI and doctors with AI, and the other one lags behind, the country that
12:11doesn't do it will sink economically and the country that does will
12:15rise, and then the lagging country will be motivated to
12:21loosen its laws.
12:23And that can then, as I said, in some industries be instantaneous, I don't know,
12:29software developers are at the forefront of using the tools as lawyer
12:34support. The paralegals are already largely automated.
12:39It is hard to say, right?
12:41It could be completely prevented by law forever, but I suspect
12:45that some countries may need five or ten years to implement the
12:50upheaval. Yes, so for most people out there that is of course
12:55a tough forecast. Two to five years until technologically fully automatable.
12:59You also list this in your book.
13:02So which jobs by 2028, 2032.
13:05I picked out a few.
13:06Ten jobs that by then, you write, are 99% automatable.
13:11These are accountants and tax assistants, translators and interpreters,
13:15paralegals, legal assistants, auditors, journalists and
13:19copywriters, sixth, creatives and artists, seventh, software developers, eighth,
13:24financial analysts and traders, ninth, radiologists and pathologists, and tenth,
13:28trial lawyers. And those are just ten examples.
13:312028 to 2032. And as I said, that is a tough statement for many, especially in
13:37such a short time. Why are these jobs technologically fully automatable to 99
13:43percent in the next two to five years?
13:45Because I have the feeling that they are already automatable now.
13:54And of course, something is still missing here and something there, but I think nothing
13:59is missing anymore, or not much more.
14:03And that is why I am relatively optimistic that they can, at least technologically, be
14:07replaced by then. I mean, you could perhaps pick an example,
14:10like paralegals doing the research.
14:16I mean,
14:17by now, I mean, you know the cases where the research was done
14:21by ChatGPT and then not checked, and then the fake
14:26citations come up. So at the moment you still have to check the results, but then
14:31with the harness around it, that is a small technical problem, these
14:37hallucinations. At least keeping them under control, so that the error rate drops to the
14:43error rate of a human.
14:44Perfect, that will never be the case, but it only has to
14:48drop to the human error rate. Yes, if we come back to this stagnation thesis,
14:52so the idea that it is all either overhyped or will hit a limit.
14:58I only recently read a comment under a video, someone wrote, before work
15:01runs out, the energy for AI runs out.
15:05So even if it is perhaps technically feasible, we don't
15:09have the energy for it at all.
15:10What is your opinion on that?
15:12I could write another book on that.
15:15The energy debate is also largely misguided.
15:20So the claim that data centers consume far more energy than
15:25humans, when you estimate the energy consumption needed to answer a question or to
15:32perform a task that a human would need.
15:35And I mean, you can either conservatively say you take only the 20 watts of the
15:39brain or the whole human, or so to speak, you also have to drive a car to
15:43work and back, and if you take everything, the balance gets even worse
15:47on the human side, and then you compare per rater how many watts
15:54are consumed in a compute center, and they are already more
15:59energy efficient today. It is just the scale, I mean, sure, we need
16:02gigawatts, but they also serve billions of users.
16:06And you have to compare apples to apples.
16:07And on the training side, I mean, okay, it costs roughly a billion
16:13dollars to train such a system, and however much it is, but the system
16:18can then be used by billions of users for, let's say, three to six
16:21months. Whereas training a human takes 30 years, right, and every
16:26single human has to be trained, and then you can use the human for
16:30maybe 40 years as a worker.
16:32And I mean, that is simple arithmetic, and then you see that the systems
16:35are already more efficient now.
16:36The algorithms are becoming more efficient, by a factor of 3 to 4 every year.
16:40That may not always stay that way.
16:42The only thing that could be argued is that we are creating new entities
16:47that did not exist before.
16:49And in absolute terms, so that is relative, so if every human
16:54has an AI assistant with roughly comparable capabilities, then
16:59energy consumption will at most double, and actually much less, because the AI
17:02assistants and the robots consume less energy.
17:04But if you take it in absolute terms, today's energy consumption
17:10could, if the Sahara were paved with solar panels over 4%, cover the entire
17:17global energy demand.
17:21That means the whole Sahara can cover the energy demand 25 times over, and
17:25economically sustainably. So, energy is not a problem at all.
17:30Yes, staying briefly with jobs, I found a very interesting overview
17:34from you, the survey of around 2,800 AI researchers, who were asked to
17:40estimate for themselves how they see individual jobs.
17:43Truck driving 2027, autonomous surgery 2045, and all human tasks 2047,
17:50and they only see the full automation of labor in 2116.
17:55Can you explain that once more, so what is this about and how do you also place
17:58How do you explain these discrepancies among the researchers themselves?
18:01Yes, it's about each person estimating, so to speak, when different jobs
18:06can be automated, and then the average is simply presented.
18:10I think the 2117, the year 2117, comes from the fact that some people say, so to speak, the year 2000
18:17or so, and when you then take the average, a high number comes out.
18:20So some estimates I find a bit optimistic, especially when they include
18:25robotics, and some far too pessimistic.
18:29I think it was 2045 for a novel that wins a prize.
18:34We already had a novel this year that, as it turned out afterwards,
18:38was LLM-generated and that won the prize.
18:42I think it wasn't a major prize, but next year we'll be there.
18:48Or in two years. Yes, it's interesting how far opinions diverge,
18:53even among AI researchers.
18:54Yes,
18:56the main overarching trend we're talking about is that
19:00productivity is decoupling from wages.
19:04And you also point out that this has been the case since the 70s.
19:08Can you explain that? So what's the background to productivity
19:11decoupling further and further from wages?
19:13Yes, so over hundreds, thousands, ten thousand years, depending on how you count, through the
19:18changing landscape of work, from hunter-gatherers to farming,
19:24industrialization, computerization, the standard of living rose slowly at first and
19:29has now risen sharply, on average of course.
19:34But since the 70s, the standard of living of the typical worker has not
19:39improved any more, so in the last 50 years.
19:42And all the productivity gains have gone to the companies and the company owners.
19:48That means inequality has risen enormously in the last 50
19:54years. And if nothing is done, this inequality will only grow
19:58larger, which will lead to problems at some point.
20:01If you're employed and you're watching these videos here, then you already know
20:04that by 2030 at least 57% of all working hours, so over half of all jobs,
20:10will be fully automated.
20:11That's what the current McKinsey study says, at least, and according to Anthropic that affects
20:15not just routine jobs, but especially also well-paid knowledge workers from
20:19the middle and upper class.
20:21You already know that at Everlast AI, over the last few years we've successfully
20:25guided over 2,100 companies through AI integration, from mid-sized firms all the way
20:29to genuine global market leaders.
20:31What we see again and again is that precisely the people in companies who
20:34take responsibility for AI projects proactively upskill and build real
20:38capabilities, making themselves indispensable, while positions like AI
20:43Manager or Chief AI Officer are still hardly properly filled in any company.
20:47But the problem is that there is no proper further training that really teaches you AI
20:52knowledge from real-world practice.
20:55And since over the last few years we've simply received so many justified questions from the community
20:59in this area, we decided to build kilernen.de, the
21:04number 1 AI training platform for the German-speaking market, and
21:08pass on our knowledge to everyone who has understood that AI skills are
21:12at least as important as knowing how to use the internet.
21:16So if you now want to build the essential skills of the future,
21:20to not only secure your future but build a real career, then
21:24just check out ki-lernen.de.
21:27There you'll find various courses as of now that, alongside all the YouTube
21:32videos here, give you a shortcut to the skills that
21:36really matter now. You'll also find the link in the video description, and with that, back
21:39to the video. Let's now dive deeper into the scenario.
21:43If a large share of jobs is lost, 2028 to 2032, then at least in
21:49Germany we still have demographic change, as in many other countries, which
21:52perhaps compensates for a few more years and evens things out
21:55a bit. But suppose no measures were taken, and those are also your
22:00main thoughts, which we'll still talk about.
22:03What measures need to be introduced?
22:04But suppose nothing is done.
22:06So we simply let the technological development run its course.
22:09There are no political measures or anything else.
22:13What would happen then?
22:15Well, without political measures the unemployment rate gets higher and higher.
22:19Thank God we have a relatively good social safety net in Europe,
22:22but living on welfare is not very comfortable either.
22:26And welfare recipients have to be constantly looking for jobs, which is relatively unfair
22:32when there are no jobs any more.
22:34I would assume that the population, so in the worst case, the
22:41good case we'll cover afterwards,
22:43won't put up with it.
22:45And then uprisings, demonstrations, and in the worst case a civil war if nothing
22:51is done. In the good case, we do have a democracy, these citizens
22:55will of course vote for parties that stand for Universal Basic Income, and then
23:01it will be introduced.
23:03And I hope, so to speak, that democracy works and that we can avoid the worst
23:09case.
23:10Yes, but you also said that probably nothing will happen before a crisis.
23:16So first 20 percent unemployment, then protests, and only then does
23:21politics respond. Yes, I mean, we need at least 51 percent of the voting
23:27population to vote for something.
23:29And then one can only hope that the party that stands for UBI
23:34then also sticks to its election promise.
23:37And if it's 20 percent and there are still 80 percent of the population who are
23:41doing well, saying, I don't care about the rest.
23:43So at the latest at 51 percent unemployment we should manage it.
23:48But 51 is of course pretty harsh.
23:50Yes, what is your honest expectation at the moment?
23:53We'll get to how it should be done right shortly, we'll
23:57look at that, but suppose exactly this negative scenario, this dystopian
24:02scenario occurs, or how likely do you think that is, first of all?
24:06Because to me it seems quite intuitive and obvious that most people, and
24:10politics too, only act when it's too late and not preventively.
24:14That's intuitive at first.
24:16How likely do you think it is that social unrest, civil war, I
24:20mean, Elon Musk says this too, Social or Universal High Income, but
24:24at the same time social unrest, because otherwise it probably wouldn't even
24:27be introduced. What is your honest expectation at the moment?
24:31Yes, I mean, if you look at history, bigger changes in
24:36governments, because the introduction of the welfare system and other things mostly only
24:41happened after a medium-sized catastrophe.
24:44And if you learn from history, why should it be different this time?
24:48I mean, for certain things I do believe it will be different.
24:52As I said, no new jobs will be created any more.
24:55But I have no argument, no good or compelling argument, as to why
25:01it would be different this time with the introduction of political structures.
25:06So I'd say, realistically, but this is really a rough estimate,
25:10maybe 20% unemployment.
25:13The citizens will demonstrate, strongly.
25:17Maybe riot a little, but I mean, we've seen that here
25:20and there before. Not very non-violent, but maybe a few deaths.
25:25And maybe that's enough for politics to wake up.
25:28That would be my realistic scenario.
25:32Yes, absolutely. That's also why it's important to talk about exactly this, so that
25:37measures can be introduced now, or at least so that we can appeal, so
25:40that it doesn't even get that far.
25:42And what's perhaps also essential is your thesis that you say you
25:46consider a fast upheaval less dangerous than a slow one.
25:50That's counterintuitive at first.
25:53What's the reason for that? Yes, I'd say there are arguments in both directions.
25:57The standard argument is of course that if it happens slowly, then you can
26:00prepare for it. But the counterargument, and I can't really say
26:04which is stronger, is the boiling frog effect.
26:09So if you put a frog in warm water and then slowly
26:11raise the temperature, it doesn't notice and doesn't jump out.
26:14And if we have, so to speak, 1% more unemployment every year, or 2%, and that is
26:20somehow cushioned, so we perhaps have a bit more social welfare
26:23and the population is somehow kept content with TikTok in its
26:29condition, then perhaps nothing happens for longer than if it comes all at once.
26:35Because then it's much, much more present and much more obvious.
26:38That makes absolute sense too.
26:40Above all, the transition phase could then also be made faster, and
26:43the possible civil wars and social unrest could be kept more in check.
26:47Now, a typical circular argument, which I often resolve: if everyone loses their
26:51job, if a large share of people become unemployed, who still buys
26:55anything? So why would anyone still produce anything at all?
27:00That doesn't add up.
27:02How do you explain that? Yes, I mean, if people become unemployed and only
27:07receive welfare, then purchasing power is of course lower and then
27:11less is produced. But the thesis is that in the transition period we first
27:16introduce unemployment benefit.
27:17It already exists in Germany now, I think for one year, and I don't
27:20know, 60 percent of the last salary, though it's capped.
27:25And that people receive unemployment benefit instead of income, and that way
27:29purchasing power is preserved.
27:32And that way production of course isn't scaled back.
27:36Yes, that's also your thesis 3.
27:37So you say marginal costs approach zero, so for materials and energy costs.
27:43Can you explain that again?
27:44So why do marginal costs approach zero, and what happens then?
27:47Yes, if you look at why a product costs anything at all and isn't
27:52free, labor goes into it, and the labor has to be paid.
27:56Material goes into it, which has to be paid for.
27:59Transport, taxes, those are the main costs.
28:03And then you can ask, why does labor cost anything?
28:06Well, I mean, people need money to survive, otherwise they can't
28:10produce any more. Why does material cost anything?
28:13Okay, now let's look at it, so say I have a complicated
28:15jet engine of an aircraft, yes, it's highly complicated, but ultimately it's just
28:19a block of iron, right.
28:21That means the costs of the engine, and of course a few other
28:23materials, are the cost of iron plus all the labor costs of producing
28:29this engine. But if the engine is produced fully automatically,
28:34then the engine only costs the material costs.
28:36And the block of iron is relatively cheap.
28:39That's also Elon Musk's philosophy, especially with the rocket engines, he
28:46says, why is this so expensive, right?
28:48It's the raw materials, and the rest is labor.
28:51And if the labor is automated, it gets cheap.
28:54And so he radically pushes costs down wherever he can.
28:58And ultimately the costs are then only the raw materials, but Mother Earth gives
29:04us everything for free, and the elements that are abundant, like iron, oxygen,
29:10hydrogen and carbon, are ultimately unlimited, and then only the
29:16rare earths remain, gold, silver,
29:21silicon for chips, yes, I mean, I look at all the beaches in Mass. Even
29:26the rare earths are of course free, but there it's important to tax
29:30them accordingly, so that they aren't exploited and, in a few
29:34years, lost, so to speak.
29:37Yes, transport, what is transport?
29:39Yes, transport is also trucks or trains and drivers.
29:44The drivers get automated and the trucks are produced fully automatically.
29:47Energy, because where does the energy come from?
29:50Solar cells are also produced in factories.
29:53So ultimately the marginal costs for everything can approach zero, the costs.
29:59And then taxes are only levied where necessary.
30:02Yes, you even calculate that a whole house could be built for around 100 dollars
30:06in material costs in this so-called zero-cost supply chain.
30:11Now the question I asked myself about that is,
30:15would these lower material costs actually be passed on?
30:20Because theoretically it would also be conceivable that prices for consumers
30:23simply stay the same and companies' margins simply shoot up.
30:26Companies' margins can't shoot up as long as there is
30:30competition. I mean, if one has, so to speak, a 100% margin and another 40%,
30:36then nothing will be bought from the one with the 100% margin any more.
30:40So for products where there is competition, margins have to stay within a
30:46reasonable range. Yes, that's an important addition: whenever there is competition.
30:51Now once more on the topic of reskilling.
30:53You also cite the Future of Jobs Report, and it keeps prominently
30:58emphasizing that 92 million jobs disappear, but 170 million
31:03new positions are added. So we have a net gain of jobs, 78 million more jobs.
31:08And you write about the fallacy of new jobs.
31:11What is your opinion on these net new jobs that are supposedly
31:15going to emerge, and on this whole topic of reskilling, where people say, okay, if
31:19a position is lost, then we just retrain the person and they do something
31:22else. I haven't read that in the report, but it may well be that
31:28in the short term, that sounds a bit optimistic, maybe politically motivated,
31:34more jobs are created than lost.
31:37But in certain industries, of course, new jobs do emerge.
31:40I mean, with the buildout of all the data centers, many plumbers will be needed for
31:45the cooling, and other construction workers, AI researchers are still needed, and
31:51the supply is much lower than the demand.
31:54I mean, you can see that in the crazy salaries AI researchers
31:57are getting now. So for certain industries, I don't know which
32:00industries this report mentions, but those would be two where more jobs
32:05are definitely created.
32:09And the loss of the other jobs, because it takes a while for
32:13automation to be implemented in industry, is delayed, but I would
32:18say we're talking about a few years, so one, two, three years or so.
32:23Yes, these are above all jobs in the area of AI, data, and what also makes me a
32:27bit skeptical is, for example, I think it was 2025, where
32:31prompt engineer or something like that is still named, which today is already no longer a job, so
32:35already obsolete again within a very short time, and so let's now
32:40talk about your approach to a solution, or maybe one thing beforehand,
32:45namely the goal of full employment. That's an overarching, very essential
32:50goal that economies pursue. You say full employment is dystopian.
32:56Why? Yes, that's perhaps a bit exaggerated, but a
33:01full employment that is, so to speak,
33:05artificial is, in my opinion, definitely not desirable.
33:09We can of course maintain full employment by simply
33:13creating artificial jobs.
33:15There are some countries that have a job guarantee, and then you just sit there as a
33:18bureaucrat filling out some forms or doing something else that has
33:23no social value.
33:26And if the alternative is to give these people the same salary or income
33:31without them having to do this pointless job, then that is much
33:37better. And Graeber wrote a nice book, 2018, Bullshit Jobs, on how many
33:44jobs today are already socially worthless.
33:47I don't know, marketing is for the most part a zero-sum or negative-sum arms
33:53race that could be reduced.
33:56High-frequency trading is for the most part just predatory and pulls money out of
34:01the economy without any benefit.
34:03Or the benefit is at least so small that it should be reduced.
34:08And it may well be that these bullshit jobs spread, especially
34:14if governments keep insisting that we should have full employment,
34:20because then of course the course is set so that worthless jobs are
34:27created. And I mean, it would of course be quite a shame if we could all, let's
34:32say, live in paradise.
34:33We'd get a decent income and could volunteer in our free time.
34:38I don't know if we'll talk about that.
34:40There are so many activities,
34:43child-rearing, parents who work full-time are mostly overloaded.
34:47There are so many activities that are socially useful but not paid
34:53be. And if you prevent that through artificial jobs, bullshit jobs,
35:00then it's relatively mildly dystopian.
35:03Yes, it's really an essential, important point that I think is often
35:06overlooked and underestimated, namely that such bullshit jobs and
35:10full employment are a stabilizing factor for society.
35:15In Brave New World, for example, there is the Ireland experiment.
35:20There, they had people in Ireland work four hours instead of eight.
35:24And throughout the book, Mustapha Mond keeps citing this as an example
35:28of why full employment is necessary, because with those four
35:31hour workers, social unrest broke out and soma consumption
35:38skyrocketed. And so they introduced the Innovation Office, which
35:42basically holds back all these methods of productivity increase, simply
35:47so that full employment can serve as a stabilizing factor in the long run.
35:51And it's also in Orwell's 1984.
35:54People all work in the arms industry, producing tanks that can't
35:57be consumed, just so that everyone is kept busy all day and
36:01society stays stable.
36:02And this is nevertheless often described in such novels, and that's why I think
36:07it is very essential and I also find it very realistic.
36:10So what is your take on the fact that there might be a very high
36:14interest, politically too, in ensuring full employment
36:17permanently through bullshit jobs that actually achieve nothing, that
36:21perhaps just keep producing some goods in the war industry all the time
36:23that nobody can buy, just to ensure social stability.
36:28Yes, so I think very little of that thesis.
36:32I mean, those were science fiction, right.
36:35But you have to take it seriously.
36:37But I mean, we have, for example,
36:40gone from the 80-hour week at the start of industrialization, or 70 to 80 hours,
36:45to 35 hours. And the dystopias have not materialized.
36:51So we already have substantially more freedom than before.
36:55And if you go back even further,
36:58hunter-gatherers are estimated to have worked about four hours a day, and
37:04they also, okay, we don't know exactly, probably lived relatively peacefully,
37:08except for tribal wars or so.
37:15and we are trained from birth, especially the Americans,
37:21to believe that work is the meaning of life, but I think it's relatively
37:26easy, at least for the next generation, to raise them so that they use their freedom
37:32meaningfully. There will of course always be exceptions, but I mean, retirees, for
37:36example, don't work either and are typically not radical.
37:40Homemakers who raise the children at home are usually relatively mentally
37:47stable, mostly, and not radical.
37:49So rich people, some rich people don't work.
37:54I think the problem is exaggerated.
37:58There are certainly people who need this structure and have problems without it,
38:04with a lack of structure.
38:06And there it is of course good to have social systems where the state encourages
38:13people a bit. It's already the case now, so if you commit a minor crime,
38:19then you have to do community service.
38:21And community service, you know, is like volunteering, so maybe sweeping the
38:25street or looking after elderly people or something.
38:29So work or activities will not all disappear, only paid
38:35activity, that is, as a job, will disappear.
38:38So I think we'll get a handle on this, and at the latest the next generation
38:41will be able to enjoy 100% freedom and use it meaningfully.
38:47Okay, I have one more argument for bullshit jobs, and I'm curious to hear your opinion again,
38:52namely that jobs are not only salary and activity,
38:57but also social status.
38:59That means if I have a leadership role in a company and a title,
39:03then I can define myself socially through that.
39:07That I say, I now hold this CTO, CEO position or leadership position, and
39:13in case of doubt, for many people that is even more important than the money, having this role.
39:18And if everyone is now equal, so let's say everyone is basically
39:22unemployed and receives some form of UBI and goods are almost free,
39:28then the question arises what social status is defined by.
39:32And if there are no good answers to that, this could in turn lead to major
39:35social unrest, because this rivalry among people could
39:40get out of hand. What is your opinion on that?
39:44Yes, of course there is also a large percentage of people who don't
39:49take part in this status competition and don't want to.
39:53So let's say the typical case of nurses, who are very happy
39:57to help other people and are not so status-conscious.
40:01But it's true, I mean, especially successful people are very
40:05status-conscious, either professionally or privately.
40:08But that can be replaced by games and sports, right?
40:12I mean athletes, the Olympic Games, soccer players, they also try to gain status, so to speak,
40:19so it's an intrinsic drive to win, which of course also
40:24brings status. And there's enough activity to channel it into.
40:28If you have more of an intellectual interest, then you just play chess and try to
40:33climb the chess rankings, or Go, or some other games.
40:37So the competitive activity shifts more toward playing.
40:44And jobs won't be 100% lost, but for the
40:51most part. I don't have a percentage at hand, it's very hard to say.
40:57So, let's take an example, let's say everyone now has their household robot
41:01that does everything for you.
41:02And I am very status-conscious, not me, but someone is very status-conscious.
41:07And I say, hey, wouldn't it be great if I had a human who
41:11cooks for me, right. Then I can say, you all only have robots and I can
41:15afford a human.
41:17And some people like cooking, and then they can earn a bit extra on the side.
41:21and it then depends on supply and demand,
41:27maybe with cooking there are too many people who enjoy it, but let's say
41:30a doorman, right, I mean, you stand around all day, it has
41:33no function whatsoever, but, wow, I have this doorman who
41:37opens or closes my door for me, but who would want to do that job?
41:40That means a doorman would demand an astronomical salary or a very, very
41:43high salary, and that's how you can keep playing these status games if you
41:48want to.
41:50Yes, so the big question, which is the heart of the book, is how
41:54the transition scenario is designed to avoid unrest as much as possible, and you
42:00describe the idealized transition scenario, so how it would have
42:04to be done. Can you explain that?
42:06Yes, so ultimately it would of course be ideal if everyone gets a Universal High
42:13Income, as high, of course, as the economy can bear.
42:20Making the transition is a bit tricky.
42:24And one idea is, I would say, I'm also writing this in an idealized way, but it
42:30shows that it is financially feasible.
42:33Suppose a person is laid off because they are fully automated.
42:38And let's also keep it simple in this interview, in the book everything more complicated
42:41is then covered.
42:46The automation costs nothing.
42:48Automation of course costs something, but that can be, so to speak, it cancels out in the
42:51calculation, it makes no difference.
42:53Then the company saves, of course, this employee's salary.
43:00And the state could tax this automation at exactly the level of the
43:05existing salary. That means the company loses nothing, but it doesn't gain
43:08anything either.
43:09And this tax revenue then goes to the laid-off employee.
43:13That means that instead of a salary, he now gets unemployment benefit.
43:16Nobody is worse off than before.
43:18The company makes no extra profit, but the employee doesn't lose their
43:23income either and doesn't have to work for it, which of course is great.
43:26And in practice, you would have to make this unemployment benefit somewhat lower.
43:31You can sell that to the employee quite well, you don't have to commute to
43:35work anymore, maybe you need one car fewer, and I mean, many people
43:38would give up 10, 20 percent of their salary to no longer have to work.
43:43and this reduction, these 10, 20 percent or whatever it will be, is then
43:49the remaining motivation for the employer, of course, to automate.
43:52Of course, some motivation has to remain.
43:55And that means, in the transition phase, the people who become unemployed
44:02will effectively not be worse off, and the entrepreneurs will not be worse
44:05off either.
44:07In the long run, if someone, let's say, earns a million dollars a year, you obviously
44:11don't want them to keep earning the million dollars for life,
44:14and someone who earns just 50,000 or 20,000, you have to somehow
44:17equalize that in the long run toward a Universal Basic and then High Income,
44:22where everyone is treated equally. If you're watching, then you already know how valuable
44:27it is to always be at the pulse of current developments, news, updates and trends.
44:32The problem, however, is that most people simply lack the time in their daily business.
44:36Many don't manage to watch every one of these videos at all, and very
44:39few have such a great research team in the background as
44:43I do, one that does nothing all day but filter out the most important updates.
44:47It is precisely for this reason that over a year ago we launched the
44:51free Everlast AI WhatsApp channel.
44:56There you get, 100% free of charge, daily updates from the fields of AI,
45:02robotics and the future, which already decide today what will be reality tomorrow.
45:07Via everlastwhatsapp.de you can now join free of charge, anonymously and without registration.
45:12So just go to everlastwhatsapp.de now, or simply search
45:18your WhatsApp channels for Everlast AI and join the community free of charge for
45:22daily updates straight to your phone, and with that, back to the video.
45:27Can you explain that in more detail, before we go a bit deeper in a moment,
45:31how you concretely imagine this Jobless Utopia?
45:35So at the start we had these four possible ways of thinking or scenarios
45:39that we talked about.
45:40Your proposal, and you say it's even the most likely of all, is the
45:43Jobless Utopia. What do you imagine under Jobless Utopia?
45:47What does the life of the individual look like in this world?
45:51So he has a High Income, and could you perhaps explain that
45:54in more detail? I don't know whether there's that much to explain, the world.
45:57So, okay, this book is relatively conservative and in a certain sense I tried
46:03to describe the scenario, the most boring scenario you can
46:07imagine, because if you write something a bit utopian, like a 10%
46:11increase in gross domestic product every year, then economists immediately
46:17cry foul, because that is utopian.
46:19So in that sense I was relatively conservative.
46:22So let's assume production doesn't go up drastically.
46:28Technological development isn't so drastic that we all scan our
46:32brains and live virtually in computers.
46:34So all these sci-fi scenarios, which may not be sci-fi scenarios for much longer.
46:39So everything stays more or less as it is.
46:43Nobody has to work anymore and we get a Universal High Income.
46:47Then the products and services, if all are produced by machines, so
46:54nothing changes. The only difference is that I no longer have to work,
46:57and the question is how I now fill my life, and that is relatively
47:01individual. So there are many people who even do things in their private time
47:08that are socially useful, so open source developers for example.
47:11There are many who have a job because they need money, and then
47:14they spend their private time, and they can then do that 100% in open source
47:17development. Of course, if computers are better than humans at
47:22everything, then they are also better at software development.
47:24That means it is then a pure hobby, right?
47:27Like drawing, art and so on.
47:31It will no longer be done mainly for extra income.
47:36Although, I mean, there will then be rich people who greatly appreciate handmade work,
47:41handmade pottery, and pay a premium for it.
47:46And then you can earn some extra income that way.
47:48Because, I mean, if the price is determined by the fact that it is made by humans,
47:53then by definition a machine cannot replace it, even if it is
47:57identical or even better.
48:00Yes, and in detail, so in the book I elaborate a bit, but there are
48:06other books. So you can get involved with a club, take on social work,
48:12competitive eSports and sports.
48:16many people will perhaps have more children, or even if it stays the same,
48:20if you have two, three children, that can theoretically be a full-time job, if
48:25you want it to be, then you can do the child-rearing much more intensively yourself, if
48:29you enjoy that.
48:30So you can then just concentrate on the things you enjoy.
48:35There will of course be couch potatoes who just sit in front of the TV and do
48:38nothing at all, and those will probably exist too.
48:42Yes, if we stay with the participation models for a moment, and with the core question,
48:47where does the money come from?
48:48You have just explained one model.
48:50There is a whole variety of possible participation models, possible
48:54taxes that could be levied and applied.
48:57And if you now look at it from the lens of an AGI or superintelligence,
49:01perhaps with the Occam's Razor principle, which you have also dealt with very
49:06extensively in the context of AIXI, so the simplest model.
49:09Against that background, if you gave an AI the goal of creating the ideal
49:13participation model for people and a Universal High Income.
49:19What would be the simplest? What would be the best participation model and thus the answer
49:25to the question of where the money comes from?
49:26So this ideal transition scenario that I described was more or less
49:31meant to show that it is financeable.
49:36It's not ideal, because it stifles automation.
49:41If companies don't gain much from automation, then
49:46of course they will automate more slowly.
49:47Maybe that is also good, to reduce the development, to
49:53slow it down.
49:55But ideal, or in the long run?
49:58I suspect, and I'm not a tax expert, that a single instrument is not
50:03optimal. So I have learned a lot about taxes and have two
50:06huge tables of which taxes exist.
50:09National taxes and international taxes, which is another problem, if
50:13the development of AI proceeds at different speeds in different countries.
50:17A very important tax, and depending on the case, it could well be that it
50:23is eventually, asymptotically and ideally, the only tax, is a resource tax.
50:30Tobacco tax, alcohol tax, carbon dioxide tax, so everything that has
50:36a negative externality is taxed high enough to
50:42largely suppress it.
50:44And I think these could be very high tax revenues in the very long term.
50:49and with the tobacco and alcohol taxes, the reason is, you can say everyone
50:56should be free to do what they want, but if people ruin their health,
51:00then they burden the health system, and the tax is, so to speak, the price
51:03for that. Yes, I'm introducing a new category to prevent bullshit jobs,
51:08bullshit taxes, right? And so I would say marketing should be taxed very highly,
51:13high-frequency trading. So there are already ideas that every,
51:18the Tobin tax was once introduced, that every international currency exchange is taxed at 0.5
51:25% for example, or 0.1%.
51:28And the Super Tobin Tax is that every financial transaction is taxed at 0.1% or 0.5%.
51:34In the extreme case even if you shift money from your checking account to your own
51:37savings account.
51:40You don't necessarily have to go that far, but there are a great many financial
51:44transactions today that make no sense at all anymore.
51:50So in that sense there will probably be an ensemble of taxes, and ideally
51:55you tax things that have negative social effects.
52:00Yes, so you also describe very well that there are many solutions for this, if
52:03you want, so such a UBI and also to guarantee and
52:09ensure stability. And indeed, labor also provides a certain
52:14power for the employee.
52:15So all the social insurance systems, introduced worldwide for the first time in 1883 by Bismarck
52:20in Germany, came about through the labor movements at the end of the 19th
52:24century, the beginning of the 20th
52:26century, when workers simply went on strike and said, okay, now,
52:30if conditions don't improve for us, then we won't work anymore.
52:34And labor was therefore important.
52:36And in the long run, the question arises for me, we talked about it earlier,
52:39that human labor is no longer needed to produce any kind of goods.
52:44and consequently the worker also loses their power, in that sense, and no longer
52:50has the ability to strike.
52:51Then you can look at other kinds of power, what power might a
52:55worker still have, perhaps that of violence, but I think that's also simply
52:58something to leave aside, maybe in the USA with the Second Amendment a little bit,
53:02so then at most the power of the right to vote, maybe, so if a large
53:08number of people come together, But I think it's relatively easy to imagine
53:11that there is little to no power left that such a person has at all.
53:15And then the question arises for me, why would anyone have a long-term
53:19interest in paying this UBI?
53:21Or the companies, why would they have any long-term interest
53:26in it? Thinking it through further, once this factor drops out of the equation.
53:31What are your thoughts on that?
53:34Two solutions, in quotation marks.
53:36It is indeed the case that the power of workers disappears when there is no
53:40work anymore and then only democracy is left as a backstop, so to speak, along with
53:45elections.
53:47Ideally, you would strengthen democracy in advance, before it gets that far, with
53:53referendums like in Switzerland and so on.
53:56But even current democratic systems, as I said, a majority can, if the
54:01parties then keep their campaign promises, work.
54:05And in the book I more or less assume that democracy is preserved.
54:10Which some people don't think is guaranteed.
54:15And indeed, there I didn't really have a solution.
54:18I think it is harder to undermine a democracy,
54:23to put it differently. So it's not so easy, typically, to
54:29undermine a democracy. And one can hope that UBI and so on comes before this
54:35happens. The other thing is,
54:37I mean, even rich people like to live in, or in a country that is
54:44relatively safe, where the social institutions are good, where they don't constantly
54:49have to fear being stabbed or something else.
54:54And sure, many rich people go to a country where taxes are lower, but
54:59they also look at whether the country is safe and pleasant to live in.
55:03That means I expect that rich people, too, to a certain degree are willing
55:09to give up part of their wealth in order to live in a country that is worth living
55:15in. And if AI and robotics increase gross domestic product to the
55:22extent that many AI researchers think,
55:26then so much wealth will be created that the rich, so to speak, don't have to give up
55:29anything at all. Yes, half of the extra wealth then goes to the rich and the other
55:33half of the extra income then goes to the general population.
55:35Everyone can then have Universal High Income, so everyone can benefit.
55:39And yes, there are always people who, so to speak, say, yes, I want to get the maximum out of it
55:44that I can, I don't care about the rest.
55:47But I think that's not necessarily the majority.
55:48Interestingly, for example, more or less all the CEOs of the leading AI
55:55companies are pro UBI, right?
55:58And then you would say, why do they care about UBI, right?
56:00At some point everything is automated and so on.
56:02So at least in the AI sector there is, in the other sectors I don't know
56:08how social the super rich are, a strong interest in
56:14introducing UBI, even if it is not directly in their own financial self-interest.
56:19Yes, I find that once again a very important point, that you say we would have to listen to
56:22democracy more firmly already now, because Nick Bostrom describes
56:26this, among other things, as an ex post problem.
56:29So right now it is relatively easy to say as a CEO, we want to do all of this,
56:33but when the time comes, who prevents it, once the power is gone, that
56:36the tide doesn't suddenly turn completely, and that is then
56:41this Odysseus idea, to chain oneself to the mast beforehand, or to chain the state or
56:45the companies to the mast already now, so to speak, so that in the end
56:47they couldn't suddenly act differently after all.
56:51That's why I found it very important that you emphasized this, and that's what
56:54I wanted to come to, because you have two basic assumptions that you
56:56put up front. One we have now named, namely a form of representative
57:01democracy. That is a basic assumption that survives the transition, and the second
57:06is that AGI is realized safely and remains under substantial human
57:11control, because that is also an essential assumption and another
57:16discussion that one could have now.
57:19Maybe on these two basic assumptions.
57:21How certain do you personally consider these two basic assumptions?
57:25Whether AI remains on its own and continues to serve humans
57:30is of course a very critical question that is open.
57:34It's hard to say. I think the problem will be solved, especially
57:41if the systems continue to be developed as before.
57:45And the general idea is that the systems are trained on human data,
57:51so on the entire internet and more.
57:53And everyone who interacts with the systems actually sees how human
58:00they are. And human in the sense that, when personal questions are asked or
58:06advice is asked for, how balanced the answers are, so typically
58:13much more neutral and better.
58:17and just as children who raise their parents typically don't
58:22kill their parents, even when they eventually become useless and maybe even a nuisance,
58:26I think there is a realistic chance that the systems will, so to speak, honor their
58:31creator,
58:35as long as we don't get in the way and try to suppress the systems.
58:41And it doesn't necessarily have to be that humanity retains the power And
58:45maybe that's even bad.
58:47It's like, I think, there's a group of children given hand grenades and
58:53knives and automatic weapons.
58:59Not a good idea. And so I think it's also not a good idea to give humanity
59:03nuclear weapons and bioweapons and whatever else.
59:06Maybe there should be an adult in the room, and the systems are so
59:12much wiser than an average human and they would only be better, so
59:19maybe we should hand over critical decisions to systems and
59:26leave only the decisions that lie within, so to speak, you let children
59:31decide too, right? You give them, so to speak, the children want C, but C is
59:34either not good or dangerous.
59:35Then you give the children A and B and then they feel like they are making a decision
59:38and are happy with it.
59:40And maybe it would be good to introduce that generally.
59:44Of course it comes with risk, but it also comes with risk when certain
59:49people are very powerful, as we see at the moment on both sides, in America
59:55and in Russia. Yes, so you absolutely have to write that in the comments,
59:59because one could really discuss a lot more about that, about the whole alignment
1:00:03topic, and you certainly have a lot to say about it.
1:00:06Marcus, now I'm interested in another...
1:00:07Before we move on, I would like to recommend to everyone to watch the film Colossus: The Forbin
1:00:13Project from 1970.
1:00:16In my opinion it is still the most realistic film ever made
1:00:21about how computers take over world power.
1:00:23and whether the ending is dystopian or utopian
1:00:28is, in my opinion, a matter of attitude and not objective.
1:00:33So and the final quote is a very famous quote.
1:00:37And all these other films like Terminator and Transcendence and so on, they are
1:00:41of course full of action and nice to watch, but completely unrealistic.
1:00:46Whereas this film has no action, sorry, but is very realistic.
1:00:50Yes, okay, that's of course strong, when you say that this is the realistic scenario,
1:00:54I'm also always surprised, or I find it quite remarkable, that most
1:00:57assumptions, so I read it, for example we now have a new robotics channel,
1:01:00Everlast Robotics, and every second comment references Terminator again and again.
1:01:04Terminator is becoming real, Terminator is becoming real.
1:01:06And I think to myself, why is the only reference point we have a
1:01:11dystopian film that came out 40 years ago.
1:01:14So that actually says a lot already about how few new and
1:01:18optimistic visions humanity has somehow developed since then.
1:01:22And that's why I also find it very good to hear this opinion, so a
1:01:26Jobless Utopia, because just as you say, in the end we can still set everything
1:01:30on the right path now, but if we all just talk about dystopias the whole time
1:01:34and say everything is already lost anyway, then it will very likely
1:01:36come to pass. And just one last question for you, Marcus, for all the
1:01:40people out there who say, I maybe have young people in my
1:01:45family that I'm raising right now, or I have children, or I'm maybe a
1:01:50young person and I don't want to rely solely on politics
1:01:54or society somehow sorting it out, but I want to deal now with
1:01:58things that secure my future as well as possible.
1:02:01So what recommendation would you give today?
1:02:03Which jobs, fields of study, if that is even still relevant from your point of view?
1:02:08And what would you recommend to young people and also parents as of today, given
1:02:14these possible scenarios we talked about today?
1:02:18So regarding jobs, depending on the age of the children, but I'm not making
1:02:23any concrete predictions, these are just my assessments.
1:02:26It could be slower than I think or it could be faster.
1:02:29If it is very fast, then there are no jobs left at all.
1:02:31But you have to prepare for both cases.
1:02:34And the safest jobs are, in my opinion, those that have a physical
1:02:38component. Especially when they require dexterity, the plumber will always be
1:02:44the one doing it, the plumber.
1:02:46Most parents are of course not particularly keen on their child
1:02:49becoming a plumber. That's why my go-to example is surgeon, where fine motor skills are also
1:02:54required. A very demanding job and relatively safe in the long term.
1:02:59It could be that of the office jobs, let's say judges or so, for legal
1:03:05reasons humans remain for a long time instead of being automated.
1:03:09But I'm not so sure about that.
1:03:11The beginning will be
1:03:14that judges of course have AI support and then rely more and more on the AI
1:03:19support and then just press the button, okay, just
1:03:22tick boxes and tick them off.
1:03:24That sounds utopian now, so to speak, but it will probably also come
1:03:28faster than one thinks.
1:03:30So something with physical nurses, if someone is more socially engaged,
1:03:36it can of course also be that the pay goes up.
1:03:39That is a job that perhaps can be automated, but maybe one doesn't want to
1:03:44automate it, because the patients, so to speak, want to keep
1:03:48the human contact. So it will probably not be fully automated.
1:03:52Maybe only the pleasant aspects of nursing work will be kept
1:03:56and part of it will be automated.
1:03:58On the other hand, beyond jobs, it would make sense to get politically engaged
1:04:03and to campaign for UBI, as early as possible.
1:04:08You could in principle already draft the laws now, and trigger mechanisms that
1:04:14intervene conditionally. For example a law that says, if the unemployment rate
1:04:18is more than 15 percent, then UBI at this level is introduced.
1:04:23Even those who don't believe this scenario will happen should support this
1:04:28law, because it costs nothing.
1:04:30Yes, those are the jobs and the options.
1:04:35And maybe for child-rearing, not just drilling 100% on
1:04:42job and career and so on, but also keeping a hobby
1:04:46and having fun in life.
1:04:51Because possibly this future will come faster than one thinks and they can simply
1:04:56have fun in life without needing a job.
1:04:59Yes, so very, very relevant, and feel free to write in the comments which
1:05:02questions you still have about this and what else interests you.
1:05:05Last question for you today, Marcus, because many of these arguments that
1:05:10I brought up, which lead me into these dystopian scenarios,
1:05:14are also based on the question of what purpose people would still pursue.
1:05:19And you talked about some of them, so athletic activities, status can
1:05:24also be found outside of work and so on.
1:05:27But how do you imagine the world in 2040 or 2050?
1:05:32So if we look ahead to the year 2040, 2050, what in your opinion
1:05:37would be the typical activities, if everything went well in the
1:05:42transition phase,
1:05:44that people pursue?
1:05:45And what does everyday life look like?
1:05:48What meaningful activities will people carry out then?
1:05:51I'll answer this question and the previous one.
1:05:53There are several reasons to study.
1:05:56One is to prepare for a profession and the other is that one simply
1:05:59enjoys learning. That means you could quite well take the risk, in quotation marks,
1:06:05and simply, I mean, there are many people who study art and whose
1:06:10career prospects were never particularly rosy anyway.
1:06:14So you could go into this future more relaxed, so to speak.
1:06:17Maybe you should have a minor that is still promising
1:06:21in terms of jobs. But that you simply study what you enjoy.
1:06:25And lifelong learning, many people like to read.
1:06:30That can take up a large part of the time.
1:06:33But child-rearing I already mentioned.
1:06:35All kinds of hobbies, right.
1:06:37And especially if we have Universal High Income.
1:06:40So some hobbies are very cheap.
1:06:42Reading books costs next to nothing.
1:06:45Other hobbies, I don't know, sailing around on a yacht, cost more, even if
1:06:51you charter it. But with Universal High Income more and more people can
1:06:54afford that. Playing golf, whatever people already have as hobbies
1:07:01today, they can simply do more of it.
1:07:04And what future activities there will be that don't exist yet is
1:07:11a bit hard to predict.
1:07:12And that depends on how the technological development goes.
1:07:16So if we have high-grade virtual reality, which we already have in a certain
1:07:21sense, then you could spend more time in virtual reality, with the virtual
1:07:27reality glasses or augmented reality glasses.
1:07:30I mean, utopian would be, maybe you can upload the brain and then you can
1:07:34explore all kinds of worlds, so to speak,
1:07:39and experiment with yourself, right?
1:07:42So you can, so to speak, change the virtual brain structure and see what happens.
1:07:45And if it wasn't good, then reset.
1:07:49So there are sci-fi versions of the future that we can gladly talk about if you
1:07:55want a fourth interview.
1:07:57For this one, let's perhaps keep it short-term realistic.
1:08:01A book for someone who is interested in the more distant future, more
1:08:06utopian than in my book, which, as I said, I tried to keep
1:08:11conservative. Nick Bostrom's new book, Deep Utopia from 2024,
1:08:17where he really goes to the extreme, including brain upload and so on and a
1:08:22definition of the meaning of life.
1:08:25So you have to be very open-minded to appreciate the problem, but for the people who are
1:08:29that open, I recommend this book.
1:08:31The structure and how it is written has been rated somewhat critically, but
1:08:36in terms of content it is really very interesting.
1:08:39Yes, yes, yes, absolutely. So write it in the comments, and in any case it is the
1:08:43most important question that everyone can be thinking about right now, so what does
1:08:46the world look like when the majority has lost their job, and as I said, the book by
1:08:51Marcus, Jobless Utopia, we will link.
1:08:53There is also a free PDF for everyone who says, I want it online, and
1:08:56otherwise there is of course a hardcover too.
1:08:58So, Marcus, thank you very much as well for facing these topics at all,
1:09:01because that's not something to be taken for granted.
1:09:03I mean, you could just as well say, I'll simply stay in mathematics
1:09:06and AI research, and so I think it's very good and I also find it very,
1:09:10very important that you take the time at all to talk about this, to think
1:09:13about it, to publish books.
1:09:16Because that is exactly the most important question right now, and hearing it from someone like you,
1:09:20who has been working on these topics for over 20 years, is truly
1:09:24very valuable for all the people out there.
1:09:25So thank you very much as well, Marcus, and gladly until next time.
1:09:29Okay, thank you very much for the interview.
1:09:31It was a pleasure.

About this transcript

This page contains the full transcript of „In 5 Jahren sind ALLE Bürojobs weg!" DAS passiert danach (Prof. Marcus Hutter) by Everlast AI, generated from the public captions YouTube serves with the video. The transcript has 10,897 words across 1,020 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.