YT YouTube2Text

YouTube2Text → Transcripts → L'histoire se répète… On en est au point de non-retour.

L'histoire se répète… On en est au point de non-retour. — Transcript

by Grand Angle · 3,707 words · 587 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00On July 22, France refused to authorize
0:02Tesla's FSD full-sell driving on its
0:04territory, even though five European
0:06countries had already given it the
0:08green light. And why? Well, because of
0:10the safety guarantees deemed
0:12insufficient. In the aftermath, Musk
0:14publicly responded that this delay
0:15would cost lives. For my part, during a
0:18trip to Florida, I drove nearly 2,000
0:20km in a Tesla, on the city, on the
0:22highway, and in traffic jams. I did it
0:24almost without touching the wheel.
0:26Let's be clear, that didn't make it a
0:28self-driving car. The system was still
0:30supervised. I was responsible at all
0:32times, and a camera checked that I was
0:34looking at the road. But here's what's
0:36interesting about the Netherlands. The
0:37authority that studied this system for
0:40over a year and a half believes that it
0:42can improve road safety. France has
0:44come to the opposite conclusion. So yes
0:46, the debate on safety is legitimate,
0:48but a regulatory decision also has
0:50economic effects. It can preserve, at
0:52least temporarily, the positions of the
0:54industry in place. But what exactly are
0:56we protecting? And above all, are we
0:58really protecting an industry when the
1:00technology that now concentrates its
1:01value starts to slip away from it? To
1:03answer this question, we have to go
1:05back to the last time an entire
1:07industry saw its core value slip away.
1:09That industry was the horse industry.
1:12But what disintegrated its economy was
1:14not just the automobile, but
1:16motorization in the broad sense. Cars,
1:18trucks, and in the countryside,
1:20especially tractors. The American peak
1:23was reached around 1918 with nearly 27
1:25million horses and mules on farms. A
1:28few years earlier, in 1915, some 93
1:30million acres of cultivated land were
1:32used to feed them. And yes, before
1:35gasoline, the fuel that powered the
1:37economy was the fodder needed to feed
1:39the horses. So what we are talking
1:42about here was not just a means of
1:43transportation, it was an energy sector
1:45. The bulk of this horse-drawn carriage
1:48was agricultural traction. So, if we
1:50want to understand what happened to the
1:52Calèch and the people who made them,
1:53we have to look at the cities
1:55separately. And there, three things
1:57happen. First, the profession had never
1:59been as flourishing as on the eve of
2:00its decline. Between 1890 and 1900, the
2:03number of horse-drawn carriage makers
2:06jumped by 300%in New York, 350%in
2:09Philadelphia, and 675%in Chicago.
2:12Between 1870 and 1900, in 10 major
2:14American cities, the profession grew by
2:17330%while the population only increased
2:19by 100%. If you were a carriage maker
2:23in 1900, you were in a booming,
2:25hypergrowth sector. Second, and this is
2:28what I want you to remember: between
2:301910 and 1920, the American
2:31agricultural livestock did not collapse
2:33. It even went from 24 million to 25
2:36million horses and mules. But over
2:39exactly the same period, the number of
2:41horses in New York fell by more than
2:43half, from 128,000 to 56,000 in 10
2:46years. In other words, anyone looking
2:48at national statistics at that time
2:50would see absolutely nothing. The
2:52aggregate was calm, but underneath, an
2:54entire ecosystem was emptying. Thirdly,
2:57it didn't happen all at once.
2:59Historians Clé McChain and Joel Tar
3:01have reconstructed the sequence that
3:03unfolded. The horse first lost its
3:06fixed uses in the 1830s, mills, pumps,
3:09then trams between 1885 and 1895, then
3:12light passenger transport from 1895 to
3:151915, then heavy freight from 1915 to
3:171930. Function by function, never the
3:20profession all at once. And in the end,
3:23the dividing line is not what you might
3:25think. Let's talk about carriage makers
3:27. The Fisher brothers made Fisher Body
3:29and supplied their bodies to the great
3:31automobile manufacturer D3. Studaker
3:33switched from carriages to automobiles.
3:36William Durant, king of the carriage in
3:38Flint, founded General Motors and
3:40Hermes Sélipuis 1837 left the horse
3:42without leaving leather. Farriers and
3:46stud farms did not disappear. They
3:48still exist today but their mass
3:50economic function has collapsed and
3:52what remains of them has focused on
3:54leisure, sport and luxury. So why one
3:57and not the other? Because a seat is
3:59still a seat, a carriage is still a
4:01carriage. The know-how was detachable
4:03from its former use. An internal
4:05combustion engine has nothing to do
4:07with a draft horse. Absolutely nothing.
4:10That's what we can call the
4:12technological gap. So, let's restart
4:14the question. What exactly are we
4:16protecting? Before going further and
4:18using the autonomous car example again,
4:20we must distinguish two things. On the
4:22one hand, WMO is real autonomy, with no
4:24human driver behind the wheel. But in
4:27controlled fleets and within delimited
4:29service areas mapped to the millimeter.
4:32Around 500,000 paid trips per week in
4:34more than 10 American cities. $ 16
4:36billion raised for a valuation of $ 126
4:39billion. On the other hand, Tesla. No
4:41autonomy without a driver. FSD
4:43Supervise remains a driving assistance
4:46system. The driver remains responsible.
4:48But this supervised driving is deployed
4:50on the scale of a particular fleet of
4:52vehicles. Wimo completely replaces the
4:54driver, but only in certain cities and
4:56certain areas. Tesla is deploying its
4:59system on a much larger number of
5:01vehicles, but the driver must always
5:03remain attentive. One starts with real
5:06autonomy and gradually expands its
5:07scope. The other starts with a massive
5:09deployment to progress towards autonomy
5:11. In both cases, the element that
5:14differentiates vehicles and could
5:15capture a growing share of their value
5:17is not just the car, it's the system
5:19that drives it. But if all this is
5:21indeed coming, then why haven't the
5:23traditional manufacturers taken this
5:24turn? I've often been told, but Richard
5:28, you'll see, Mercedes, Volkswagen and
5:30all the rest, Toyota, Fiat are also
5:31going to make autonomous cars. Well,
5:34it's not for lack of trying. General
5:36Motors took control of Cruise in 2016
5:39for nearly $ 600 million. 8 years later
5:41, more than $ 10 billion in operating
5:43losses for less than $ 500 million in
5:45revenue and therefore the end of
5:47development. And General Motors 'Cruise
5:49system, I tried it on rental cars in
5:51the United States and frankly it's all
5:53over the place. Let's take Ford now. It
5:56invested in Argo Eai with Volkswagen.
5:592.7 billion depreciation in 2022 and
6:01another project halt. Same for Ford;
6:04I've driven and tried them. Well,
6:07autonomous driving, assisted driving,
6:09is absolutely nothing like a Tesla or a
6:11Van. It must be said, these are systems
6:14that can barely follow white lines
6:15drawn on a road. Two of the world's
6:18largest industrialists, so not idiots,
6:20remarkable engineers, almost unlimited
6:23resources and in the end, well, nothing
6:25, nothing at all. So, we could conclude
6:29that they simply executed poorly, but
6:30that seems a bit thin as an explanation
6:32, especially when you consider who
6:33we're talking about and the guys they
6:35recruited, they weren't idiots either.
6:37Personally, I prefer to rely on the
6:39famous quote from Charlie Manger. Show
6:41me the incentive and I'll show you the
6:43result. So, let's look at the
6:45incentives of Mary Bara, the CEO of
6:47General Motors. On the one hand, an
6:50asset that brings in factories, a
6:52dealership network, engines, cash every
6:53quarter. Conversely, a software bet
6:56burning 2 billion yearly after having
6:59already racked up over 10 billion in
7:01losses. And once again, an entire
7:04corporate structure that is
7:05acculturated, designed and led to
7:08manufacture gasoline-powered cars that
7:10are driven with both feet and hands. A
7:13manager judged on a quarterly basis
7:15with a balance sheet to defend and
7:17shareholders to reassure can hold this
7:18bet for a certain time. And the further
7:21the bet moves away from the core
7:23business and corporate culture, the
7:24more each new billion must be justified
7:26in the face of an activity that is
7:28currently profitable. At some point,
7:31the incentive structure pushes for cuts
7:33, and at Alphabet, Wimo is backed by a
7:35group whose core business is already
7:37software, data and artificial
7:38intelligence. At General Motors, Mary
7:41Bara ended up saying it herself:
7:43Operating a fleet of robot taxis is not
7:46the core business. Add the technical
7:49distance. A manufacturer designs a new
7:51platform over several years. It tests,
7:53it approves, it freezes an architecture
7:55, then it produces it and amortizes it
7:57for years. And of course, all of this
8:01is accompanied by processes, a way of
8:03doing things, and a corporate culture
8:06that supports but also rigidifies a way
8:08of thinking and designing. A driving
8:12system must improve through continuous
8:14remote updates, fueled by the
8:15kilometers that the fleet covers. These
8:19are not simply two versions of the same
8:21job, they are two industrial cultures,
8:23two different rhythms, and two
8:24different incentive systems. The
8:27autonomous car system at Tesla is a
8:29system of software development and a
8:31start-up. It has nothing to do with the
8:34automotive world. And honestly, you've
8:36all seen that, right? If you've been in
8:39a Tesla or a Wayimo and you compare the
8:41on-board computer system with that of a
8:43Toyota, a Fiat, a Ford, or whoever,
8:45well, you realize that there's a world
8:47of difference between the two. It's
8:50just two people. On one side is my
8:52grandmother who does computing, and on
8:54the other, a young École Polytechnique
8:56graduate who designed it all. This
8:58difference in approach is really
9:00reflected. So what are they doing today
9:01? Many have stopped trying to cross the
9:03border alone. They buy the brain or
9:05provide the body. Toyota approached
9:07Wimo to bring the technology to
9:09passenger cars. The most striking
9:11example is Hyundai. The agreement
9:13provides for Hyundai to supply Wimo
9:15with a large volume of Ionic 5s over
9:17several years. The vehicles are
9:18assembled in the Georgia plant. Then
9:20Wimo integrated its own autonomous
9:22driving system. Take a close look at
9:24the division of labor. Hyundai supplies
9:26the body, the brakes, the battery, and
9:28a vehicle designed to accommodate the
9:30system. Wimo adds the sensors, the
9:32computing, perception, and decision. In
9:34other words: the driver, the vehicle on
9:36one side, and the technology driving it
9:39on the other. And if tomorrow the
9:41customer chooses a car primarily for
9:43its technology, then the manufacturer
9:45risks becoming a subcontractor for the
9:46carcass. It manufactures the body,
9:49someone else sells the mind. So, have
9:51the markets understood this? Tesla is
9:53valued at around $ 1.43 trillion. The
9:57world's 10th largest company. But Tesla
9:59is too specific a case. It has
10:01narrative, cult, promise, in short,
10:04muscle. Let's take Wimo instead. $ 126
10:07billion based on its latest private
10:09fundraising for a business that is
10:11still a mere speck compared to global
10:13transportation. And on the other side,
10:17Mercedes-Benz, the heiress of Carl Benz
10:19who filed the patent for the first
10:22automobile in 1886. Annual turnover of
10:24€ 132 billion, market capitalization
10:27of around € 42 billion. Mercedes
10:29generates about three times its own
10:31market value in turnover each year, and
10:34Wimo, whose business is not even to
10:35manufacture cars, is valued about two
10:38and a half times more than Mercedes.
10:41Obviously, market capitalization does
10:43not measure yesterday's turnover. It
10:46measures what investors think they can
10:47earn tomorrow. And in this regard, we
10:50have a rather interesting case with
10:52Armes. With a market capitalization of
10:55around 165 billion euros, the saddler
10:57from 1837 is worth more than the three
10:59German automobile giants. Now, be
11:01careful, Hermes is not worth 165
11:03billion because the market would
11:05remember 1915. Hermes is worth that
11:07much for strictly contemporary reasons.
11:10An operating margin of 41%, exceptional
11:12brand power, rare products and a
11:14capital intensity that has nothing to
11:16do with that of the automobile industry
11:18. By way of comparison, Mercedes'
11:21automotive division posted an adjusted
11:22operating margin of 4%in the second
11:24quarter of 2026. And that's not nearly
11:26the same for all major manufacturers.
11:28But precisely, and this is the heart of
11:30this video because if Hermes is still
11:32around, it's because Emile Hermes
11:33understood something while looking at
11:35the hood of a military vehicle during a
11:37trip to Canada. He discovered a closing
11:39system. In 1922, he secured the
11:42exclusive rights, and the following
11:44year, the house adapted this innovation
11:46for an auto bag. He didn't try to
11:49preserve the world of horses. He
11:51detached his know-how from its former
11:53use and adapted it to the next world.
11:56Leather survived the horse-drawn
11:57carriage. When France opposed the
11:59deployment of the FSD, the economic
12:01effect could be independent of the
12:03stated intention. If the core value is
12:06indeed shifting towards software, then
12:08while we deliberate, we may not be
12:09protecting the future of the European
12:11automobile industry, we are protecting
12:13its carcass. Except that in the days of
12:16the horse, this dividing line mainly
12:17crossed one sector. Today, it crosses
12:20them all at the same time. Let's take
12:22the example of accounting. And that's
12:25rather convenient because from
12:27September 1, all French companies must
12:28be able to receive electronic invoices.
12:31This is not a reform of artificial
12:33intelligence, it is a tax reform. But
12:36its real effect is to connect the pipes
12:39. Because once all invoices are
12:42circulating in a machine-readable
12:44format, it is easy to imagine how AI
12:46could automatically reconcile invoices
12:49and bank entries. And in the end, the
12:52accounting firm recovers accounting
12:54that has already been largely processed
12:57, except that the accounting firm would
12:59disappear. What is disappearing step by
13:03step is data entry, lettering, MOC
13:05reconciliation, and
13:06function-by-function consistency checks
13:09until extinction. Just like horses.
13:12There will still be chartered
13:13accountants. Of course, some will
13:15supervise, arbitrate, advise, and sign.
13:19But if the same volume of files can be
13:21processed with far fewer people, it is
13:22not because the profession survives
13:24that employment will survive in the
13:26same proportions. Just as there are
13:29still farriers today. And we risk
13:32seeing the same thing in customer
13:33support, in legal research, in the
13:35first pass of coding, in scientific
13:37literature reviews, and so on, even in
13:39literature and competition prizes. And
13:43that is my favorite moment because if
13:45you have looked into the subject, you
13:47could very well tell me that all this
13:48is a bit shaky and that we have a whole
13:50bunch of studies proving that jobs are
13:52not more threatened than that. Stanford
13:54, for example, concluded that there is
13:56no generalized shift in employment
13:58across the economy. And you could argue
14:01that it’s actually the opposite,
14:02since the hiring intentions announced
14:04in July alone are the highest for a
14:06July since 2022. And while all this is
14:08true on paper, we still need to dig a
14:09little deeper. Employment among 22-to
14:1325-year-olds in the jobs most exposed
14:15to AI is about 19%below the trajectory
14:17that would have followed if it had
14:19evolved in the same way as that of
14:21young people working in less exposed
14:23jobs. A year ago, the gap was 15%.
14:27Experienced workers, on the other hand,
14:28are not showing any comparable drop-off
14:30. The aggregate is calm. Below that,
14:33some early-career jobs are emptying.
14:35It’s the same statistical trap as
14:37with horses. The agricultural sector
14:38remains almost stable while in New York
14:40, horses are disappearing from the
14:42streets. And from what we can see, this
14:44decline seems to come mainly from a
14:46drop in hiring, not an increase in
14:48layoffs. So now, let's take Munger's
14:50famous quote again. A manager doesn't
14:53necessarily need to fire a lord. It's
14:55costly, sometimes legally risky, and
14:57very visible. He can simply stop
15:00recruiting beginners. That's much more
15:02discreet and can remain buried in the
15:04general statistics for a long time. The
15:06door doesn't close behind those who are
15:09already in, it closes in front of those
15:11who wanted to get in. And then, at that
15:13point, you could tell me that all we
15:14need to do is train people. They'll do
15:16something else. There's no shortage of
15:18retraining programs. Except that today,
15:20we're starting to get a measure of what
15:22these programs actually do. A
15:24meta-analysis published on August 12
15:26aggregated 56 American trials on
15:28professional retraining. Okay, I'll
15:30spare you the details, but it's also
15:32important to know that this study was
15:33co-authored by a researcher from
15:34Anthropics, who is therefore a
15:35stakeholder, but that doesn't make
15:37things any less interesting. Remember
15:39that while the most effective programs
15:41exist, and some of them manage to
15:43drastically increase the income of the
15:45people concerned, the problem is that
15:47they are very, very selective. In
15:49several cases studied, they only retain
15:51about one candidate out of every three.
15:54In other words, yes, some people can
15:56cross the Rubicon, but it's not the
15:58same as getting everyone to cross the
15:59Rubicon. The best candidates retrain.
16:03The masses have no guarantee of
16:04following suit. And that's likely to
16:07get worse. Look across the table. When
16:11Google was approaching $ 30 billion in
16:13annual revenue, the company employed
16:16about 24,000 people. At Sales Force,
16:18there were nearly 80,000 when they
16:20crossed that same threshold. In Tropic,
16:22it has just exceeded a revenue per user
16:25rate of 65 billion dollars per year
16:27with an estimated direct workforce of a
16:30few thousand employees. So yes, around
16:32Entropic, there are obviously data
16:34centers, chip manufacturers, cloud
16:36providers and so on that should be
16:37taken into account. But ultimately, the
16:40trend is relatively clear. The core
16:43value can be captured by companies that
16:45mobilize relatively few direct jobs.
16:47And then we must also talk about the
16:49speed of adaptation. At Figer, the
16:51robot manufacturer, the boss explains
16:53that the first industrial use case at
16:55BMW took 12 months to deploy. The
16:58second at another client specializing
17:00in logistics took 30 days. This is
17:03obviously not comparable to the
17:05diffusion of a technology throughout an
17:06entire economy, but look at the
17:08compression of the adaptation cycles.
17:1012 months for a first industrial use
17:12case, 30 days for the next. The urban
17:14horse takes several decades. Which also
17:17means that the frontier is moving at an
17:19increasingly crazy speed and therefore
17:21that it could move again before what it
17:23has just crossed has finished being
17:25converted. So, let's recap the three
17:28mechanisms. 1, core value is shifting
17:30toward a technology outside the
17:31business. This technology potentially
17:35makes it possible to operate a growing
17:37part of the economy with much less
17:39human labor. And 3, the transition
17:42could progress faster than our ability
17:45to convert what it leaves behind. So
17:47where does all this lead us? What
17:49follows is my vision, my interpretation
17:51of everything I have just told you, and
17:53I am happy to accept that I am wrong as
17:55usual. My intuition today is as follows
17:58. Yes, artificial intelligence will put
18:01a large number of people out of work,
18:04who will be expelled from a whole host
18:06of economic sectors, whether in
18:08tertiary, office, or intellectual work,
18:11but also in secondary and primary work
18:13once robotics has fully adopted AI.
18:17However, once we have said that, if you
18:19expel, let's say, 90%of the population
18:22who are no longer part of the leading
18:24edge of the economy, what happens? All
18:27these people are not going to raise
18:28their heads to the sky and wait to die.
18:30They will probably do what all human
18:32beings have always done, come together
18:34and create a social fabric. So in this
18:38social fabric, we should see a
18:40technologically deleveraged economy
18:42where people will generally have a
18:44technological ceiling because beyond
18:46this technological ceiling, they do not
18:48have sufficient productivity and
18:50complementarity to allow them to
18:52integrate the upper classes. So I think
18:57we are heading towards an economy that
18:59is becoming stratified, with radically
19:01different productivity levels that
19:02coexist but that rub shoulders less and
19:04less because their material realities
19:06no longer have much in common. And yes,
19:09that means an explosion of inequalities
19:11in living standards. On the one hand,
19:15there is a very small, extremely
19:16productive layer that has the models,
19:18the infrastructure, the robots or the
19:20capital needed to exploit them and,
19:22ultimately, why not the entire
19:23transhumanist part that allows us to
19:25cognitively support the explosion of
19:27computing power and the emerging
19:29artificial intelligence. Because yes,
19:32in a world where AI continues to
19:34progress, well, it is no longer
19:36artificial intelligence but
19:37superintelligence that is emerging. And
19:40if you want to remain competitive with
19:42this type of tool, well, once it has, I
19:44don't know, to caricature 5,000 IQ,
19:47it's not with your 160 IQ from Arward
19:49that you will make the difference. From
19:52its point of view, you will be an ant
19:54twice as intelligent as the ant next to
19:56you, but it is still an ant. So
19:59inevitably, if we continue on this path
20:01, I imagine, I'm a, I expect that
20:03the humans who will remain at the
20:05cutting edge of productivity will be
20:07augmented humans. And on the other side
20:10, we will have human societies that
20:12will be increasingly alien to this
20:13cutting edge of technology and who will
20:15potentially live very well and perhaps
20:17even in abundance. It depends on the
20:20social compromises that will be made
20:22because in reality, the difference
20:24between the starting point and the end
20:26point that I describe, the path itself
20:28could be very disrupted because these
20:30profound social transformations will
20:32inevitably cause a huge stir. So if you
20:36think that the world is unequal today,
20:38as things are going, tomorrow it will
20:41really, really be much more so. It's
20:44the explosion of gaps, it's social
20:46fragmentation and above all it's this
20:48intermediate period during which our
20:50institutions continue to function as if
20:52human labor remains essential to
20:53production when a growing part of this
20:55production can already do without it.
20:59Income, social protection, pensions and
21:01more broadly our place in society, all
21:03of this is still based in one way or
21:05another on the idea that most people
21:07contribute economically through their
21:08work. At some point, we must deal with
21:11social classes that are drifting apart
21:13and no longer speaking to one another.
21:15So a new social compromise. If you want
21:18to know more about the transhumanist
21:20dimension, which is the most
21:21interesting aspect of this end of the
21:23video, check out this video on Grand
21:25Angova where I share with you a way of
21:26approaching this subject.

About this transcript

This page contains the full transcript of L'histoire se répète… On en est au point de non-retour. by Grand Angle, generated from the public captions YouTube serves with the video. The transcript has 3,707 words across 587 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 translate the transcript into 100+ languages.

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, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.