Sharon, yes, I'll absolutely ask David about this. I hope he knows about the infrastructure aspect of A.I. like he does the technological and sociological aspects of it because I'd like to know more about this myself. Thanks.
I feel like we're thinking in a similar direction if along different lines. Two questions come to mind:
1. How can we talk about AI in a way that doesn't flatten the subject to a yet another cultural contest between opposing camps (ie for or against)?
2. I'm working on developing smaller, smarter, networked models, playing with the idea that groups of these, through their interactions with each other, might have the potential to demonstrate better response quality in certain domains than much larger models. Currently I'm exploring a.) combining vertical recurrence with ternary quantization (BitNet b1.58 based) and b.) personality-weighted reasoning specialization (Phi4 based). Which approach do you think will yield better results and why?
Thanks again for the questions, Mark. They actually were a little involved even for David, but we discussed your "opposing camps" topic. I got to your questions at the end of the interview: https://youtu.be/Amj2o0JeJ_4
Thanks, Mark. I'm going to think about #1. That question applies broadly. I guess it's easy, tempting even, to analyze issues this way. It's baked into discussion formats from debate teams in school onward. It seems natural for us to organize this way, like all such things are binary. But I agree this is simplistic. So thanks for this reminder on this issue.
Would David address the issues and possible solutions that AI data centers can place on water systems and the energy grid.
Hi Sharon, thanks again. We got to your questions toward the end of the interview: https://youtu.be/Amj2o0JeJ_4
Sharon, yes, I'll absolutely ask David about this. I hope he knows about the infrastructure aspect of A.I. like he does the technological and sociological aspects of it because I'd like to know more about this myself. Thanks.
I feel like we're thinking in a similar direction if along different lines. Two questions come to mind:
1. How can we talk about AI in a way that doesn't flatten the subject to a yet another cultural contest between opposing camps (ie for or against)?
2. I'm working on developing smaller, smarter, networked models, playing with the idea that groups of these, through their interactions with each other, might have the potential to demonstrate better response quality in certain domains than much larger models. Currently I'm exploring a.) combining vertical recurrence with ternary quantization (BitNet b1.58 based) and b.) personality-weighted reasoning specialization (Phi4 based). Which approach do you think will yield better results and why?
Thanks again for the questions, Mark. They actually were a little involved even for David, but we discussed your "opposing camps" topic. I got to your questions at the end of the interview: https://youtu.be/Amj2o0JeJ_4
I will add David is working on "smaller" models as well, if that's the correct term. He's built a tool for more personalized A.I.
Brainyus — We build AI that compounds human knowledge. https://share.google/s7DO1gBOfkJelTVnm
Thanks, Mark. I'm going to think about #1. That question applies broadly. I guess it's easy, tempting even, to analyze issues this way. It's baked into discussion formats from debate teams in school onward. It seems natural for us to organize this way, like all such things are binary. But I agree this is simplistic. So thanks for this reminder on this issue.
For #2, I will take it straight to David. 👍
And here's the interview: https://youtu.be/Amj2o0JeJ_4