@Okanogen Doesn’t this also imply that our current system is also racist and bigoted and can’t be trusted to make decisions for society?

I think you’re comparing these systems to a platonic ideal of a government which can’t exist in the real world.

Seems more useful these hypothetical systems to our current system. I realize that’s far more difficult, but it’s also far more relevant.

@Okanogen Note that I’m not advocating for AIs here, but I *am* advocating for logical rigor.

If AI government is simply a reflection of society, why exactly wouldn’t we trust it at least as much as we do human government? Conversely, why would one be preferable to the other?

Seems if anything it’d be easier to implement change across an ai-run government, assuming instant government-wide application of a single viewpoint is even a desirable goal. That’d suck with Trump.

@Okanogen Couple things:
- you’re saying an AI will be bigoted because it’s trained on data from our society, which itself is bigoted. Fair enough. But the alternative is a government made up of bigoted people, so how would AI be worse exactly?

- AIs can’t innovate (yet), but that’s not a positive in, say, the judiciary. Or in carrying out many executive or legislative functions. In these cases, uniformity in applying laws is the goal. Seems AI would be better that people there.

@Okanogen Are you even actually reading my responses?

No one is saying these systems are infallible - their “hallucinations” might be the most well-known feature of LLMs.

But it’s silly to think these systems need to be infallible to be useful. They just need to be *better than the system they are replacing*.

I suspect that depending on how you define “better”, LLMs could definitely have a role in government, because our existing system of government isn’t impressive.