4 ms·
I just had a surprisingly coherent conversation with it mainly about Apple and the DMA. For over an hour. The cracks started to show eventually when it rehashed
by skeledrew 3y ago
I just had a surprisingly coherent conversation with it mainly about Apple and the DMA. For over an hour. The cracks started to show eventually when it rehashed previous points/questions, but a simple "we discussed this" fixed it each time. Actually learned a few things in the session, and it isn't over (using via Telegram).
- gremlinsinc 3y agobut is it better than Claude 3, Gemini, or gpt 4? Because that's kind of the bar, or well Mistral on grok is actually pretty good for a 75b parameter model.
- skeledrew 3y agoI haven't tried Claude, Gemini was pretty bad the first time I tried it, and GPT4 is pretty good... for coding. I never thought to use any of them as a "conversational agent" before as they weren't advertised as such. Note I consider this distinct from models just having a conversational interface.