Negative users dismiss the feasibility of multi-government AI datacenter monitoring as unrealistic because it assumes international cooperation between historical adversaries is viable.
Based on 1 visible X reactions from 3 accounts; directional sample.
Ask a question below.
Published answers will appear here.
The entire scenario smuggles in the assumption that international cooperation between historical adversaries is not only viable but likely, which is the least plausible trajectory unless we experience a globally galvanising Chernobyl-like event (but perhaps this is the implicit assumption)
@DKokotajlo This dodges the question with a hypothetical. I'm asking, for this to be true, someone would have to force all labs and all prospective labs and all training-ish runs to enforce total transparency. Who's doing that? Is your answer is "all the govts just work together" to do it?
@krishnanrohit You can read the scenario if you like to get a sense of what it takes and why the various governments might go for it. We aren't claiming that it's likely to happen; we think it is unlikely but possible and at any rate we think it would be good if it happened.
@DKokotajlo I do not understand how you do not think this level of centralisation of power is a) not centralisation, or b) desirable. Do you really not see the negatives with this? (I've read the scenario)
Negative users dismiss the feasibility of multi-government AI datacenter monitoring as unrealistic because it assumes international cooperation between historical adversaries is viable.
Based on 1 visible X reactions from 3 accounts; directional sample.
Ask a question below.
Published answers will appear here.