Hacker News Reader

Comment by atemerev | Hacker News Reader

> I can't wait for the revel that GPT-4 is just a chat application connecting you to a the worlds biggest center in India.

Tempting, but you can launch Mistral-7B on your laptop and see the same coherent responses and reasoning from a 8GB model file. The magic is really there.

View on HN

Theranos level fraud combined with the impossibility barrier?

OpenAI fakes their AI to be first to market and then everyone else, thinking that it's possible, finds ways to do it for real.

(this is 100% a joke. I know OpenAI's work is the real deal)

Also, the technology of finding the right set of human experts on basically any topic in milliseconds in India would be a massive feat in of itself

lol yea. The idea that someone would be convinced after running an LLM on their laptop and not after seeing ChatGPT expertly respond at 400 WPM in 0.2 seconds, is funny.

mistral 7b foundation model is the first AI related thing that excites me since the launch of ChatGPT. shows us that this tech is poised to be commodified, self hosted at a low cost eventually and won't necessarily be tied to and controlled by tech billionaires.

In fact, Mistral is so exciting that I think it has some direct relationship to the topic of this discussion (they still don’t talk about how they trained it, at all).

Perhaps there was some, how to say it, unexpected technology transfer, with Sam somehow involved.

interesting take, we'll see how things evolve.

First good take I've seen. You don't get fired for stealing other peoples IP or for leaking chats, that's just the name of the game. You do get fired if you sold off OpenAI's IP (or gave it away because you're an idealist)

Did you try with your Internet off?

Reply on news.ycombinator.com