And right now China is trying to catch up with that technology. They might be lagging today, but they are a huge market, and they will eventually catch up, maybe not completely but enough to be competitive.

Plus their LLMs need a lot less hardware resources for inferences.

Don’t forget that all US frontier models are operating at a huge loss, because if they need to charge market rates their market will shrink significantly.

So this GPU crunch is forcing Chinese labs to innovate and find clever ways to achieve higher intelligence with constrained resources.

And cloud and computing cost will only go up now, due to exploding hardware prices, which will make the whole business model of the US frontier models even more unprofitable.

That’s why you should buy physical disks. In the worst case you can re-sell them to someone else.

Even if they don’t make it, the whole fact that the aviation industry goes less CO2 intensive would be a good win.

To be honest DRAM was dirt cheap for quite a while and the profit margins of those companies were razor thin. Only recently they saw a surge in demand and they are exploiting it to enrich themselves but also increasing the manufacturing capacity is a lengthy and costly process which I presume some of those companies are actively trying to delay, in order to preserve their fat profit margins now.

Competition is good though and hopefully will lower the price overall faster.

The beauty is when those companies run out of human training data and start training on AI slop, just to generate even more AI slop.

This is probably already happening though.

It depends on the country you are living in. There are plenty of people with restricted and surveilled internet.

Manjaro was quite messy last time I tried it a couple of years ago.

Yes, so pretty much they don’t want to be held responsible. And the government is bending over to comply with the tech bros.

But how would they know I have been truthful about my age? I mean I am never putting my real age when registering on websites.

Why do they need the age after all. What are authorities going to do with it?

That’s a great question. I like the concept of more stable OS and I played a bit with NixOS in the past but ultimately decided that it is too much of a hassle to learn Nix to use it and decided to try Bluefin. I am actually overall content as I was able to install the missing packages either in Toolbox/Distrobox or using homebrew.

By the way I have actually found a way, and rebased my OS to another immutable flavor. I know that’s more of a workaround.

AI should also be taxed proportionally then. And they should be liable and not exempt of copy rights infringements.

I believe most of the companies are doing it to inflate their share prices.

You also need to change the devices browsers, extensions and timezones to stay anonymous or buy a device and set the most common fingerprint settings, so it is harder for those companies to track you down. It is a slippery slope, and you can check your browser fingerprint and avoid adding unique settings, extensions or anything that can help them to track you.

Why is our company valued at 4 trillions instead of 10 trillions. Those peaky humans are not buying enough of our shares. /s

I don’t know but I am constantly hitting the RAM limit with 16Gb of RAM with around 20-30 open tabs and other apps, both on Linux and Windows

Welcome, glad to be of help.

This at the moment is a clear Ponzi scheme.

OpenAI or Nvidia announces some partnership or a deal for X amount of billions. As a result the valuation of the recipient company goes straight to the sky, increasing the market valuation of said company X-times more.

And then try to tell me this isn’t a bubble.

The future is very small models trained to work in a certain domain and able to run on devices.

Huge foundational models are nice and everything, but they are simply too heavy and expensive to run.

China right now is leading the way with releasing open weights models. The US lags behind, as they are all more concerned about releasing closed weights commercial models.