AGI Can Be Safe
Data Skeptic
00:00
The Importance of Safety Interlocks in AI AGI Safety Research
The concept of AGI is it's can no longer be the bureaucracy it needs to be a mind and that's where I also I defer with many philosophers. For all sports like to think about machine smart learners as minds in the box not bureaucracies in the box. Still I should say here that what I just described it has its own residual failure modes so you still need to look at those and then also manage those if you have managed thoseyou will find that the management mechanisms to manage those failure modes have the run failure mode so you need to manage those too.
Transcript
Play full episode
