4 ms·
We've started using meta.ai and the 200B llama model on a couple servers we run at my company. I really have to go with Zuckerberg's approach on this one. I'm n
by sfblah 2y ago
We've started using meta.ai and the 200B llama model on a couple servers we run at my company. I really have to go with Zuckerberg's approach on this one. I'm not sure OpenAI and Anthropic really have anything that can't be cloned much more cheaply.
- htrp 2y agoThere's a 200bn parameter llama model?
- sfblah 2y agoBleh you're right. I meant the 70B model. I think there's some other context where the number 200 makes sense. Thanks for the correction.
- jsheard 2y agoZuckerberg's approach may be good for you but it's costing Meta astronomical amounts of money for pretty nebulous benefit to themselves, so I wouldn't count on their current approach continuing forever. Enjoy it while it lasts.
- generalizations 2y agoMost of FAANG is effectively in competition with each other; OpenAI might be joining that cohort soon; as long as Facebook can kill their business model, Facebook is keeping a major player out of the running. Zuckerberg will keep doing this for as long as it takes to kneecap or bankrupt OpenAI.
- paxys 2y ago> for pretty nebulous benefit to themselves All of Meta's products are increasingly dependent on AI for engagement and ads. Their investments in the space are first and foremost for the benefit of the company itself. So no, they are not going away any time soon.