2 ms·
I can't wait until LLMs are fast enough that a single response can actually be a whole tree of thought/review process before giving you an answer, yet is still
by Mockapapella 3y ago
I can't wait until LLMs are fast enough that a single response can actually be a whole tree of thought/review process before giving you an answer, yet is still fast enough to not even notice
- joshspankit 3y agoI would bet a chunk of $$ that right before that point there will be a shift to bigger structures. Maybe MOE with individual tree of thought, or “town square consensus” or something.
- bsima 3y agoWhy wait? This is pretty much what Groq has in hardware, just need the software layer to do the review process.