If you think of hiring as trying to get the best deal: best candidate for the most discounted price, what you want to do is find the proverbial gem in the rough. But AI does precisely the opposite of that. It will recommend exactly the same profile that appeals to everyone else. The gems will remain undiscovered and you will all be competing for the few individuals that according to their profile, are the safest bet.
He did a lecture. Not sure if you can still find it on Youtube, because IIRC, he published a paper and then redacted it. From what I can tell it was bits of old fashioned differential geometry and a whole lot of hand waving.
ML Research is ripe for such a subculture to emerge, because there are truly so many research directions that are nothing more than a tower of cards ready to be exposed.
You need an element of truth to capture your audience. Once you have an audience and you already deconstructed the tower of cards, you start looking for more content. And then you end up like Sabine.
My impression of NVIDIA is that internally the teams are quite independent and there's not much in terms of broad strategy. So it could be these two efforts are not related.
I mean, trivially true, if you consider that AI-enhanced programming requires 1000% more code written, deleted and rewritten, countless times, by AI itself in it's feeble attempt to "reason" about the problem.
I agree with you that tech stocks are overvalued. However, there is no free cake that I can think of. You can invest in bonds but in the short to medium term they under perform compared to equities by a lot.
And most industries have their own risk profiles that if anything are more tangible than the notion that someday, the tech bubble will pop.
Actually you must be thinking of the most reputable banks.
Because most of them will absolutely try to fleece their older customer demographic by trying to make them invest in high expense funds that perform terribly. For them to advice you to put them into a decent index tracking fund is already a good sign.
NVIDIA put a lot of effort into making their hardware and accompanying software useful and usable. CUDA by itself might have never got any attention if not for the effort that NVIDIA puts into helping their customers use their technology, effectively. And they are by no means perfect at it. Most of their products are a horrible mess and they often have 7 different ways to do the same thing. A lot of their libraries are closed source and you're forced to use an API that links to a black box. There's lots to complain about.
Maybe they also do that, but I work with a class of problems* that no other model has managed to crack, except for R1 and that is still the case today.
Remember that DeepSeek is the offshoot of a hedge fund that was already using machine learning extensively, so they probably have troves of high quality datasets and source code repos to throw at it. Plus, they might have higher quality data for the Chinese side of the internet.
* Of course I won't detail my class of problems else my benchmark would quickly stop being useful. I'll just say that it is a task at the undergraduate level of CS, that requires quite a bit of deductive reasoning.
I'm not at all familiar with this whole field, but why would you publish a trading strategy if it has potential? Why not sell it to a hedgefund, at least? Or is this research formally publishing what industry is already doing?
No expat means expat and immigrant means immigrant. Immigrant implies that the intention is to settle in the host country, whereas expat implies that the reason for staying is temporary and hence the intention is not to stay permanently.
For example, if I am doing a PhD in France, with no intention of staying in France afterwards, wouldn't it be rather inaccurate to call myself an immigrant to France?
The reason the EU cannot compete in tech, is because its market is way too tight with the US market. Any founder has a choice (if you can raise this insane level of seed capital you have the choice). They could pick US, Canada, UK, France, Germany ...etc. Given that choice, they will pick the US every single time. It's strategically the best choice, simple because of its size and wealth.
Politically, we need economic growth. It is what made us "civilized". Before we had ample economic growth, the human ambition to get more wealth and power used to mean, to go to war and take all that your neighbour has.
And people should not make the mistake of believing we are somehow not exactly the same people we were 100 or even 1000 years ago. People today, are still killing each other over land, for example.
So yes, the energy/climate crisis should be taken seriously, but proposing a utopia that ditches one of the major constraints on our way of life is no solution.
The job is too niche. This means there are very few employers worldwide that can make use of your experience. That actually puts downwards pressure on your salary (the employer is a quasi monopsony), not to mention that you are tied to a few locations.
The whole hype about AI safety is to some extent a shrewd marketing ploy. It's the whole, friends holding back their buddy who is amped up and ready to start throwing punches, act.
That is not to say that Hinton, Sutskever and others aren't genuinely concerned about AI safety. But I doubt that is the reason why the big names are paying lots of random nobodies to pretend to care about AI safety, because frankly, I do not see how they can output anything of use in a possible AGI future.
It's a matter of degree. The average Joe cares about privacy and tech sovereignty too, but not to the extent that he would sign off a platform where the rest of his friends are.
I don't think people care AT ALL about FOSS, however, they do care somewhat about privacy and sovereignty.
But at the end of the day, they need to have the network of people they care about already on the platform and they need a smooth experience.
There is nothing wrong with protectionist measures per se. These tariffs are just not set in reality. They demand the impossible, based on false premises. This won't happen. Therefore an economic recession will unfold.
You are free to come to Germany. Pretty sure you can get a VISA to do an Ausbildung at the end of which you will be working in some factory or another. Getting a German VISA is much easier than getting an American VISA and Germany is actively advertising abroad to attract workers. That should tell you something, but do come and experience it with your own eyes.
OK, let's put it this way. You would struggle to buy a house if you worked at VW assembling cars. You would be able to rent an apartment and have a relatively good if modest existence. Nothing close to what Americans are able to afford.
But if you wanted to create blue collar jobs and if the government was going to step in and contort the economy with heavy handed measures, anyway, then just setup a public works program and build a bunch of housing, build and maintain energy and transport infrastructure, build climate mitigation projects. That would actually address a real demand and make a whole lot of sense.
Americans should study Germany and Japan. Do their economies look healthy? Are those manufacturing jobs attractive? Are the Germans and the Japanese richer? Do Germany and Japan have good economic prospects?
I'll spare you the research. The answer is, "No". Turns out, having an economy based on manufacturing high-end door knobs in 2025 is not great. Economic growth and innovation is not there, because there is so much growth and innovation that you can eke out of high-end door knobs. These enterprises are great for the families that own them, but they employ a relatively low number of workers, those jobs do not pay that much, and they exist is a steady-state. Well, only until China and India figure out how to also manufacture high-end door knobs.
Another possible reason might be outreach. NVIDIA spends big money on getting people to use their products. I have worked at two HPC centers and at both we had NVIDIA employees stationed there, whose job it was to help us get the most out of the hardware. Besides that, they also organize Hackatrons and they have dedicated software developer programs for each common application, be it LLMs, Weather Prediction or Protein Folding, not to mention dedicated libraries for pretty much every domain.
I do not think we have enough historical evidence for such an extrapolation. Human societies have been getting more and more complex, necessitating more and more complex governance. We are at a point of crisis, where the electorate is supposed to decide through voting, and our elected politicians are supposed to decide through legislation, on topics that neither the electorate, nor the elected are able to fully understand. It is not clear to me that democracy will triumph.