> I guess that means we should absolve ourselves of any responsibility for it? I don't think so, I just think it's a complicated topic without many easy answer

> I guess that means we should absolve ourselves of any responsibility for it? I don't think so, I just think it's a complicated topic without many easy answers. > How do we apply that idea to our involvement in building future tools? If we pick a less sensitive topic like power tools, they have some basic safety features. I know some folks get annoyed by those, but most folks find them helpful and can still accomplish their goals with the tools. But the safety features still don't 100% prevent misuse. > Should we go ahead and build "sentient" autonomous weapons because "any tool can be turned into a weapon" anyway? Most production AI applications have a "human in the loop", especially when accuracy is essential. The AI will alert a human and provide info & context, but a human still makes the decision. This is because even our best AI isn't smarter, safer or more ethical than a skilled human (at this point). At the point where a "human in the loop" starts generating more errors than autonomous AI, then it might make sense to have autonomous AI, with a kill-switch of course. I don't know what the advantage would be in making AI "sentient" though.