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