[kylebrown]

> We recommend expanded research aimed at ensuring that increasingly capable AI systems are robust and beneficial: our AI systems must do what we want them to do.

The author points out that the meaning of this statement hinges on the word "we". Hasn't the world already arrived at a dystopian future where AI systems simply maximize revenue and ad clicks for their owners? Whereas "good" AI would be emloyed "for the people", i.e. to maximize social welfare while taking into account externalities (poverty and global warming).

The fear is of some future AI which acts only to further its own pathological interests... but how would that be different from present-day corporate bureacracies, with business processes increasingly driven by weak (but rapidly improving) AI? Seems to me that the owners of such systems are the only people who would fare differently under the two scenarios. But for the vast majority of the world, their experience would be the same in both cases: exploitation by autonomous AI, or exploitation by bureaucracy. Changing the parameters that control the way either one operates is equally difficult.

[eli_gottlieb]

>The fear is of some future AI which acts only to further its own pathological interests... but how would that be different from present-day corporate bureacracies, with business processes increasingly driven by weak (but rapidly improving) AI?

It's not that different at all, which is precisely why we need to avoid both options on this dichotomy. Also, I think you meant to reply to the thread about the other NYRB article above.