5 Comments

It seems to me that the biggest problem with AI Alignment - and this is practically the elephant in the room - is that it only takes 1 non-aligned AI with access to the external world to create so many problems that even a hundred thousand aligned AIs wouldn't be able to fix them.

Expand full comment
author

I don't necessarily think this has to be the case – I could imagine other powerful aligned AIs being able to monitor and spot actions from unaligned AIs, correcting them, flagging them, etc.

Expand full comment

The link leads nowhere?

Expand full comment
author

Should work – if you open it many times successively, it will lead to a 404 (security measure by squarespace).

Expand full comment