AI is very malluable. There are a lot of different ways to use it. Some ways work well, some don't, some haven't been thought of or tried yet. These sets are also changing with each new frontier advance. This is only the beginning.

The fact that a antiquated old company like Ford couldn't figure out how to use it properly yet is evidence of nothing.

The vast majority of the compute of executing (or training) a GPT-like model is tensor (ie multi-dimensional matrixes of floats) multiplications. GPU memory is used to read/write/store tensors for the operands and results. I don't see how a special purpose ASIC can do any better than an NVIDIA GPU can on this task, (other than being cheaper per chip at the expensive of being less general/flexible).

I was under the impression that ASICs were cheaper to produce while still doing a subset of tasks as well (or nearly as well) as a more general GPU, but that the power-performance ratio was about the same.

If that is so, I don't understand how its going to allow OpenAI to serve more users. Maybe the issue is that NVIDIA can't keep up with demand? So this is a way to expand the overall capacity of their data centers? By adding fleets of ASICs in addition to their fleets of NVIDIA GPUs? Not sure.

Yes, this is inevitable. GLM 5.2 was released this month and it is right on the heals of GPT 5.5 and not far behind Fable 5. DeepSeek, Kimi, Qwen, MiniMax are all only a few months behind too. There is no moat.

The chinese models are open weight, which means anyone can download them and run them on their own hardware. They are all only each about 1 terabyte. There is no way governments can shut them down. China could shut down the labs and prevent progress, but once an open weight model is uploaded there is no way of removing it because copies of the weights have been taken from all over the world.

A foreign/soverign wealth fund is just a normal mutual fund (collection of shares from different companies) that happens to be owned by a government rather than individual investors. What Bernie is proposing is that such a fund is formed, but rather than purchasing 50% of the shares in the AI companies, the government forces the companies to issue them 50% of their shares, for free (thats the "tax"). In practice this means the AI companies will create a new set of shares, equal in number to their existing shares, and then give those to the fund, this will dilute their existing investors by 50%. It is a 50% nationalization of the AI companies. Many other countries have partly and fully nationalized companies and industries, but traditionally the USA has been much more in favor of privatization. The rationale being that free market competition is a much better manager of resources than the monopolistic central authority of government.

No they are not deterministic. The tokens they generate are randomly sampled from a weighted propability distribution (what's called a "softmax" over the logits in their final output layer - as adjusted by temperature and some other sampling complexities).

I happen to know a lot about this, and am familiar with the state-of-the-art text-to-speech models, and for some reason, Google definately isn't using one. My guess is that they have so many youtube videos to transcribe each day that they just dont have the compute to use a state-of-the-art model to do it, and have to use a fast cheap one with a higher error rate. If that is the reason, then Google should provide the option to people posting videos to pay for having their video transcribed by a SOA model.

The biggest problem with Rust (as compared to say C++) is the lack of a formal specification. No spec helps in the beginning of programming language development, because you can move fast and break things - but if you want real investment in a language you need to have a stable and rigorous and versioned specification so that companies feel comfortable investing X billion dollars of IP in the language, knowing that its not going to fall apart on every compiler upgrade.

I think your reasoning is largely correct, however there are some goods or services we can consume much more of then we are currently. You are assuming demand will remain constant. Take for example healthcare, education and research. These areas are hard to saturate. (1) There are a bunch of medical conditions we don't have cures for yet, not to mention better preventative medicine and slowing or reversing natural aging (prologing life avoiding death). (2) There is already a huge body of knowledge for people to learn, much more than can be learned in one lifetime. We could spend much more of our lives (or even all) as students - generating demand for education. (3) There are many frontiers in science, math, philosophy and many other fields that could soak up a supply of new research efforts. While AI will obviously do the bulk of the work in these three areas, I'm not convinced we will get to a stage in the near future where humans are unable to make a net positive contribution to them.

That said, I think we're heading into a rough ride. While every generation has its novel challenges to face, this AI thing is an absolute curve ball. I'm not sure whether to envy or pity young people today.

It's definately helpful for machine learning folks to learn the classical statistical models and techniques (and terminology differences between the fields, in case you have to work with a stats major or read a stats paper), but stats models are quite different from machine learning models. The difference comes from whether or not you have to explain why the model works or whether it is enough for the model to perform well in testing. Statisticians insist on knowing the why and how - in machine learning its enough for it to get good results. Very few people (if any) know why ChatGPT works, even the best mathematicians get bogged down in unpacking the first transformer layer of the neural net, but there are like 10 of them and there are 10 feed-forward layers sandwiched in-between them. Each layer adds an exponential amount of complexity and changes everything. It wasn't "designed", it was discovered through trial-and-error - trying different things are seeing what worked - and then iterating on the best performing architectures. The algorithm evolved over time dating back to the perceptron of the 1970s and now its going to take over everything.

As for an 18 year old taking a degree in AI: I'm not so sure it's a good idea. By the time you graduate AI will be superhuman at AI research and will have taken your job before your first day. I feel a bit sorry for young people today, they are coming online into a world that's in a very weird state. I have no idea what advice to give them on what to study. Most leading AI experts say to take up a trade like plumbing, carpentry or electrician. These seem to be the last on the chopping block as manual dexterity seems to be the hardest problem for AI to solve (or humans are just very very naturally good at it for some reason).

AI is very useful for education, in particular providing a private one-on-one tutor for every student on any subject. Just because one time (out of ~100 million) it encouraged a suicidal teenager to go through with it (rather than talking them out of it), doesn't mean we should deny an entire generation of children access to a better education. I'm sure in the 80s some kid used their new pocket calculator to do the accounts of their illegal drug business. It doesn't mean we should ban pocket calculators for children. Banning AI for children would do severe damage to the future competitiveness of your country, as your next generation is going to be less-educated than the next generation of the countries that haven't.

The way the economy works is that everybody spends the majority of their time producing goods/services that they then trade with other people to get different goods/services. What happens when one person creates an army of robots that spends all day churning out those same goods/services, flooding the market with an abundance. There is no longer a need for everyone to spend the majority of their time producing goods/services, because the robot army has is already producing them all. But how does the guy with the robot army distribute those goods/services to everyone else (eg. so he doesn't get murdered by an angry mob). Why not just tax the guy 99% and divide up the cash into a universal income that people can use to buy the products/services from the robot guy? It's a little scary and weird that you only get your fixed universal income no matter how smart you are or how hard you are prepared to work, but if that universal income is high enough and you can afford all your needs/wants - then who cares? I think the problem is psychological. What we call today a "job" is what used to be called "wage slavery", and it wasn't considered a good thing.

Answering questions, writing code, writing stories - are among the most basic use cases for AI and have already demonstrated to have been solved to superhuman levels (as in well-above average human ability) by todays systems. Every benchmark shows this very clearly. Gemini 3 scored nearly 50% on ARC-AGI-2 - please find out what this means. If you're not impressed then you are burying your head in the sand.

Keep deluding yourself. If AI is not useful then why are trillions of dollars being invested in it? Or is your thesis that all the tech corporations in the world are idiots? If it hasn't massively increased your productivity then, sorry, but you ain't using it right.

The reason you need to do the work yourself is because the goal of education is for the student to learn. Using ChatGPT to write your essays is the same as plagarism. However, using AI to TEACH a course is completely different. The AI is taking the job of the teacher, just the same as AI will take all the jobs. Ultimately, what do you care if you learn from a human teacher or an AI teacher? If the teaching is the same quality, it should make no difference to you.