dauertewigkeit

It is not the first war in which the "good guys" as posthumously determined by us, were disguising themselves as civilians, out of necessity. Multiple resistance groups in WWII to draw inspiration from. I don't know of anyone who would look at the SS/Wehrmacht retaliatory punishment of the nearby civilian population and think, "oh well, that's totally justified due to those pesky resistance groups".

I don't think most of them have any kind of secret sauce. I think the founders hope to get bought out simply for being able to train "near-SOTA" LLMs. I guess achieving that level of skill and infra could be valuable enough to build upon.

Millions of people including one side of my family have experienced the opposite. They were tenant farmers living in poverty, barely subsisting, before the hard left socialists both 1. developed the economy enough to give them jobs, 2. provided social safety nets such as tax payer paid healthcare so they wouldn't go bankrupt every time they needed to buy medicine or visit a doctor. If the country had never gone that direction it would have spent many more decades being a quasi-feudal land stuck the the middle ages. Not sure how "individualism" would have helped them at all. They didn't have any capital.

It would be interesting to know if this is explained by the status gain of tech jobs (w.r.t more traditionally prestigious jobs). This is one of the theories that tries to explain some of the gap in preferences between man and women of equally high aptitudes.

Being Christian in Europe and the US are two completely different experiences. The two chief differences that I notice:

1. In Europe it is sort of the default state. Everyone belongs to the same Church. It is not really part of one's identity. American Christians are so splintered that being part of a Church means being part of a significantly smaller community and that plays a much bigger role in forming one's identity. The community aspect of Church is similarly more pronounced in the US for this reason.

2. American Jesus is your personal trainer, your protector, your personal cheerleader, your forever companion. European Jesus is technically supposed to also be those things but nobody really buys into it. The fervent belief in the efficacy of prayer is completely lacking. That gives people less incentive to go to Church.

People who care about justice are not hypocrites. People who claim to care often are.

Its also possible not to see humanity as being composed of competing tribes.

Some men just care about their own tribe at the expense of all others, and other men care about truth and justice.

    "Unlike in the modern dating process, they cannot get to know a potential partner intimately before deciding whether or not to commit to a relationship."

This sentence caught my attention because it strikes me as false. People do commit in many ways, and rather quickly, and then they break up. A history of many break ups and similiar experiences is probably the chief difference between today and centuries past. IMHO all breakups and all experiences leave their mark.

Synthetic hydrocarbon fuels depend on both green hydrogen and DAC. So green hydrogen would be needed anyway. The only question is whether you build infrastructure with the goal of using hydrogen directly or not.

People want to have their cake and eat it too, it seems. When you enter into a marriage contract you are taking on risks in return for rewards. Trying to de-risk marriage for women will likely end up causing a whole bunch of other issues. Men should also be protected from marriage fraud and paternity fraud.

American culture is to be envied by all. Europe has a lot to learn.

No, don't generalize like that. What you have to understand is that even as a percentage of the total population, immigration is only a problem in certain countries/regions mainly both because of failed immigration policies and also because the people there are plainly more racist/xenophobic.

I would put the slowing economic growth and aging population as more substantial EU-wide problems.

This is a just-so story without any empirical evidence.

The model of hierarchy is too simplistic. If you zoom in on the nodes you will see a more complex structure. It's not just one person but a host of people, operating in a semi-formal somewhat also anarchic structure. At every level of the hierarchy there is ample backup that will keep the system from collapsing.

Then there is also the dimension of efficiency that is completely missing from the discussion. Hierarchical systems might be more fragile but also much more efficient.

There is no paradox as I see it. All involved believe that AI is going to be revolutionary. They only disagree on its potential impact. But either way, if you believe this is going to be something powerful, then you would want to be able to possess it, control it and learn all about it. Although I respect Mr. Musk, I am partial towards Mr. Page's prediction. We are not going to be able to keep AI in a bottle, just like we could never dream of keeping any other technological revolution of the past contained.

Not sure how they do it specifically for LLMs, but you can do what is called model or tensor parallelism where you can split a layer over multiple GPUs or even nodes. If you look under the hood it's the same distributed matrix multiplication stuff with MPI, as far as I know.

I think Deepspeed has bespoke transformer kernels which handle this stuff specifically.

Yes, because it is well know that Mohammad urged Muslims to create criminal gangs and start dealing drugs... Sorry, but what sort of stupid nonsense worldview do you have?

There are certainly immigrants with non-western values and also immigrants who are criminals but why are you mixing these two demographics together? If anything the ones dealing drugs are likely to be the most westernized.

People are forgetting that OpenAI started as a legit non-profit. It was not meant to be a big money making startup. So presumably getting a cushy board member position was not as hard because this was just some weird philanthropy thing that some SV people were funding...

It's part of the TensorLi definition where all the magic happens.

The multimesh is interesting. Still, I bet the Fourier Neural Operator approach will prove superior. Members of the same team (Sanchez-Gonzales, Battaglia) have already published multiple variations of this model, applied to other physical scenarios and lots of them proved to be dead ends. My money is on the FNO approach, anyway, which for some reason is only given a brief reference. To their credit DeepMind usually publishes extensive comparisons with previously published models. This time such a comparison is conspicuously missing.

Full disclosure: I think DeepMind often publish these bombastic headlines about their models which often don't live up to their hype, or at least that was my personal experience. They have a good PR team, anyway.

There are a whole lot of people whose "strategy" for navigating life is to always pick the side of the strongest group. For them being part of the mainstream as a form of protection is of higher priority than being correct/truthful/just/moral.

I don't think pg was referring to this demographic, but there are certainly a whole lot of people who even take pride in being mainstream on every topic.

Twitter was the first platform to get celebs posting on the daily so they had something new and exciting on their hands. TikTok then said, what if we turn normal people into celebrities?

TikTok would not exist without Twitter and I think Twitter could not have been TikTok. The two cater to different audiences. The elitism of Twitter is what kept the celebs posting there.

That is part of the hero narrative that I am also lamenting.

If there are 100 possible research directions and only 1 proves successful. The people who travel along the other 99 get no recognition, despite expending the same effort and despite their contribution being just as important.

I don't think a CUDA engineer, or a researcher working on optimization or distributed computing should get the same recognition as a Chief of anything. But dare I say that writing optimized code to run on clusters is just as hard as coming up with a new deep learning architecture? Actually having done both, it is harder, at least for me.

The authors of "Attention is all you need" got all the recognition. The people who worked on scaling deep learning models to huge clusters got no recognition. But both were just as instrumental to the ultimate success of LLMs.

Industry wide, not even in startups, the people who get screwed the most are the ones working under the hood. I am not even talking about the disparity between SWEs and C-suites. I'm talking within the engineering domain. The closer you are to the application side the more credit you get for the effort, even if it might have only been possible through the hard work of recent under the hood technological advances.

A prime example is deep learning. The people getting all the credit are the AI researchers. Now I am not trying to discredit them. But a huge part of what made the recent training of large models possible are all the other advances. All the dedicated PyTorch/C++/CUDA/MPI code that is written just to be able to scale on a HPC cluster, for example. Those people are handsomely rewarded, but they aren't getting the Chief AI scientist recognition, despite it probably being just as hard if not harder to write all the optimized code to enable and facilitate LLMs in the first place.

Before I got into deep learning I was into numerics. Numerics people get typical SWE salaries, but they aren't getting FANNG level salaries, even though their code is powering the simulations across whole industries that ultimately enable many of them to operate at all.

Now I am sure somebody will debate me on the "getting screwed" part, but it sure feels as if, the ones who use a technical library and build an application on top of it, get the recognition both for their own work and also get part of the recognition that should due-fully be bestowed on the author of the library.

But maybe this is how the entirety of society works, but one should still be conscious of this phenomenon. If we all stand on the shoulders of giants, then maybe we should also name the giants in the acknowledgements section.

I know a QC startup that sells software to compute simulations using classical methods on HPCs, but with the promise that once QC is available they will add QC acceleration.

People have been trying to link the two together, but I see no direct link between (good) economic growth and public social spending.

It is not because of welfare that there are less startups in Europe and it is not because of welfare that the big companies in Europe prefer to be more conservative with their spending on R&D. It's cultural. It's people being more financially conservative and risk averse.

Germans are extremely pessimistic and risk-averse for one.

First of all, I'm German.

Secondly, I think everyone has a right to have opinions, thoughts and values, that others deem hateful. I would not want to live under your dystopian rule, where there is only one correct way to think.

If you are going to impose impossibly high standards on these people BEFORE they are gainfully employed, you're the problem. They don't need to integrate nor learn the language to do a variety of jobs. Give them a job and a reason to exist and believe they are advancing in life. Then demand that they learn the language. Don't put them in bureaucratic limbo for years, because they aren't "integrated enough". No wonder you end up with problems then.