Share this comment
The assumption that supporting the US military is necessarily evil always bugged me. What is your view of the Ukraine conflict? How about the looming invasion of Taiwan by the PRC? Or the PRC attacks in the Philippines and Vietnam? What about cyber defense against the PRC and the Russians? Could an AI system have detected the mapping err…
The assumption that supporting the US military is necessarily evil always bugged me. What is your view of the Ukraine conflict? How about the looming invasion of Taiwan by the PRC? Or the PRC attacks in the Philippines and Vietnam? What about cyber defense against the PRC and the Russians? Could an AI system have detected the mapping error in the Kosovo conflict where the US bombed the Chinese embassy? There are many use cases for AI in DoD (and in any bureaucracy). The good or the evil must be considered case by case.
AI could also have caused problems in military applications. The guaranteed hallucinations fraction is problematic.
Aside from that, war is, unfortunately how nations maintain and establish their rights to exist. But I don't argue with academics about military applications of technology.
That Waymo cars do well so far gives me hope that with further improvements current AI methods will be made smarter and more reliable.
Although I don't believe AI is relevant outside math, coding, and the physical sciences, it would be neat to see people prompt some AIs with, "How can we make the world a better place?"
Nice thought but more precision is called for. "Better" implies valuation, and values are subjective. I wouldn't expect something trained on reams of Internet text and then fine-tuned by a rarefied set of young college grads to behave in a way that aligns with my values.
I am genuinely puzzled by talk of “AI alignment” for precisely that reason.
Aligned with whose values? Sam Altman’s?
For a field that calls itself “science” there is a great deal of sloppy terminology..
It's one big circlejerk in search of VC funding. Science is hardly anywhere to be found.
If it were up to me, I’d call it “AI malignment”