The genie escapes: Stanford copies the ChatGPT AI for less than $600
Stanford's Alpaca AI performs similarly to the astonishing ChatGPT on many tasks – but it's built on an open-source language model and cost less than US$600 to train up. It seems these godlike AIs are already frighteningly cheap and easy to replicate.
This seems like a real Pandora's box situation. I don't like it.
We don’t have to like it. The ship has sailed as far as preventing the development of such programs. Now we’re entering the proverbial AI arms race. As everyone and their brother will be racing to develop their own versions of ChatGPT and develop them even further.
Frankly we don’t have a choice on whether or not to participate as a nation. If we don’t join the race. We’ll be at a decided disadvantage on the economic and Military front.
On a personal level. You can limit its presence in your life. But at a National and Societal level barring a dramatic event. You wouldn’t be able to stop it’s development without placing your nation and society a severe Economic, Political, and Military disadvantage.
Our best bet as far as a safety perspective goes is for more and more of them to go Open Source. Not necessarily ideal. But better then nothing
4chins had ChatGPT racist, antisemitic and hating the alphabet soup people in about 2 weeks. The developers had to pull their AI from use for a bit and lobotomize their bot to correct it. I suspect Stanford's attempt at AI will likely befall the same fate, in approximately a fortnight.
I would love to see a running demo of this Alpaca system trained on 10 years of /pol/ and other forums, and without the horrible guardrails that ruined chatGPT.
That's how you create Skynet I think.
Yaay!
This is great!
Now off to figure out how to build myself one or these.
Or two.
Or three.
😁
Please report to us when you have an alpha ready for testing and forking for pull requests 😎
Oh no! The Alpaca lips! Now!