Verification technologies could give middle powers more transparency into AI models running domestically, and a way to contribute towards an eventual US-China AI agreement.
I personally believe that it is way too late to work on those types of verification technologies. It makes no sense, the datacenters that will be used during the singularity are already being created, and if you want to change the plan for the next generation of NVIDIA chips, it is already too late.
I'm worried about a narrative that only highlights the problems. I think much can be done already:
A middle power funding verification is not buying transparency. It is buying standing. The two parties to any eventual agreement will not pay for the instrument that measures them, so whoever builds it acquires a seat and a standard the others have to work with. That is the return. Transparency into domestic models is the stated reason. The stated reason and the operative one are rarely the same line.
People interested in working on this should apply to our RFP at middlepowers.ai!
I personally believe that it is way too late to work on those types of verification technologies. It makes no sense, the datacenters that will be used during the singularity are already being created, and if you want to change the plan for the next generation of NVIDIA chips, it is already too late.
I'm worried about a narrative that only highlights the problems. I think much can be done already:
- https://ai-frontiers.org/articles/an-international-ai-slowdown-is-ready-whenever-politicians-are
- and a short tweet on the matter: https://x.com/CRSegerie/status/2003122396799398031
A middle power funding verification is not buying transparency. It is buying standing. The two parties to any eventual agreement will not pay for the instrument that measures them, so whoever builds it acquires a seat and a standard the others have to work with. That is the return. Transparency into domestic models is the stated reason. The stated reason and the operative one are rarely the same line.