What are the relative speeds of AI capabilities and AI safety?

post by NunoSempere (Radamantis) · 2020-04-24T18:21:58.528Z · LW · GW · 2 comments

This is a question post.

If you want to solve AI safety before AI capabilities become too great, then it seems that AI safety must have some of the following:

Is this likely to be the case? Why? Another way to ask this question is: Under which scenarios doesn't aligning add time?


answer by technicalities · 2020-04-24T22:20:12.205Z · LW(p) · GW(p)

Some more ways:

If it turns out that capabilities and safety are not so dichotomous, and so robustness / interpretability / safe exploration / maybe even impact regularisation get solved by the capabilities lot.

If early success with a date-competitive performance-competitive safety programme (e.g. IDA) puts capabilities research onto a safe path.

answer by William Walker · 2020-04-28T22:38:44.714Z · LW(p) · GW(p)

Let's just save time by jumping to to the place where the AI in charge of AI Safety goes Foom and takes us back to the Stone Age "for safety" ;)


Comments sorted by top scores.