“They” is the copyright industry. The same people, who are suing AI companies for money, want the Internet Archive gone for more money.

I share the fear that the copyrightists reach a happy compromise with the bigger AI companies and monopolize knowledge. But for now, AI companies are fighting for Fair Use. The Internet Archive is already benefitting from those precedents.

What kind of person owned a computer as a hobby in 1976?

That’s true in the same way that Trump’s tariffs are paid by other countries. Which is to say: Not at all.

Bill Gates was no billionaire at the time. His background was probably shared by almost all computer hobbyists at the time.

“Property is theft.”

the caveats that commercializing someone else’s work or taking credit for someone else’s work should be illegal.

So, not actually abolishing IP, then.

I really don’t get how opinions on intellectual property and its “theft” turn 180 whenever AI is mentioned.

I find it very unexpected. It used to be understood that IP laws favor monopolies. EG I don’t remember the OS community being on the side of Oracle in https://en.wikipedia.org/wiki/Google_LLC_v._Oracle_America,_Inc.

Maybe it just passed me by.

For fastest inference, you want to fit the entire model in VRAM. Plus, you need a few GB extra for context.

Context means the text (+images, etc) it works on. That’s the chat log, in the case of a chatbot, plus any texts you might want summarized/translated/ask questions about.

Models can be quantized, which is a kind of lossy compression. They get smaller but also dumber. As with JPGs, the quality loss is insignificant at first and absolutely worth it.

Inference can be split between GPU and CPU, substituting VRAM with normal RAM. Makes it slower, but you’ll probably will still feel that it’s smooth.

Basically, it’s all trade-offs between quality, context size, and speed.