- cross-posted to:
- technology@lemmy.world
- cross-posted to:
- technology@lemmy.world
“AI” – chatbots that wake up, “set their own goals,” and “spontaneously” start hacking servers – is fake. It doesn’t have “a 10% chance of ending the human race.” The Hugging Face hack isn’t a mysterious, supernatural occurrence. It’s a Python loop and a chatbot. The people responsible didn’t accidentally create god: they created autonomous malicious software and then failed to closely monitor it, resulting in it doing something both foreseeable and bad.
That’s just how this type of company operates. If they wanted to, they could quit using frontier models and just sell gpt4 subscriptions to people for $10/mo and never die. That revenue could continue funding training costs (albeit much more slowly).
The problem with most of these models is that they’re too bloated and poorly engineered to ever run forever for that kind of price point per user.
Again, do some reading. These people are selling dollars for a penny and when they tried to raise the price to a nickel everyone balked and they lowered it again.
I can run models comparable to that locally on my machine if I wanted to. I haven’t read the numbers so I guess I won’t argue this point but I do find it at least very surprising.
Open weight / free models are largely where the innovation in the models is actually happening (which makes sense because it’s where it has to). OpenAI and to a lesser extent Anthropic are money furnaces that long ago gave up on efficiency and much prefer to stuff trillions of parameters into a model that can only be run if we trap the sun in a dyson sphere or whatever.
Chuckling sadly at the idea of some aliens pulling up to the solar syatem and all they see is a giant carbon nanotube ring that says “What can I assist you with today?”