Any one used this? Or is this sponsored content in disguise? I've become quite a cynic whenever an AI company touts their supposed leaps in progress. But would be good that the US hegemony be broken in terms of LLM.
pull down to refresh
21 sats \ 2 replies \ @Arceris 26 Jan
There’s some stuff on X about it. Saying it’s much cheaper than ChatGPT, and also much less censored/lobotamized.
reply
122 sats \ 1 reply \ @Arceris 26 Jan
Ok, I just started playing with it… it’s pretty sweet. So far, the free version seems at least as good as the $20/mo ChatGPT version…
reply
10 sats \ 0 replies \ @south_korea_ln OP 14h
I've played a bit with it too. The DeepThink looks cool. It was pretty interesting to see it write: "but wait..." and each time go off on a tangent. It didn't improve much the quality of its response without DeepThink, but all in all, was a better experience than the completely free ChatGPT model. And the question I asked would have required really advanced reasoning to give a good answer.
Will try to use it as my default this week and see how it compares in terms of coding productivity.
reply
99 sats \ 3 replies \ @xz 26 Jan
It's nice that, like the original Llama model and it's myriad iterations, it's free to download, but I feel this is being made into more of a big deal than it is. As I hear, this is optimized to perform as well as open ai's subscription service, the industry being touted as the thing we need to throw billions at, as well as nuclear powered data centres for the computational requirements of Sam Altman.
First, optimizing something is a lot easier than discovering and implimenting new things, secdond, I don't really think there was much hegemony as the llama model was initially made public, which started this whole optimization game. What is Nature saying?
reply
21 sats \ 1 reply \ @SimpleStacker 22h
Agreed. If they stay open weight that would be good, but as a technical accomplishment I don't really see anything to get impressed about yet.
For all we know, its energy and other costs are highly subsidized by the Chinese government, which we know is something they do. For another, they could have stolen OpenAI secrets, which is another thing the Chinese are well known for
reply
0 sats \ 0 replies \ @xz 17h
Yeah, I'd imagine so.
reply
0 sats \ 0 replies \ @south_korea_ln OP 14h
Yeah, that wasn't really justified. I forgot that Facebook's model is open-source (up to a point).
reply