File under Really?
It's amazing to me the amount of magical thinking supposedly rational people do. AI will be an unflinching reflection of humans with none of the forgiveness. That is why diversity matters.
https://www.itbusiness.ca/news/ai-can-be-as-as-fallible-as-the-humans-that-make-it-microsoft-panel-concludes/100522
no shit.
ReplyDeleteI kinda disagree, at several levels. There used to be an AI strategy called Minsky Frames. Don't see it much anymore, but when AI was more philosophy than technology, it was (and remains) a serious approach to knowledge acquisition.
ReplyDeletehttps://web.media.mit.edu/~minsky/papers/Frames/frames.html
You do see a fair bit of the actual substance of Minsky Frames in speech recognition. If the AI is ever to reason in any recognisable way, comparable to human reasoning, it will be in something resembling Minsky Frames.
AI is not going there. It will never go there, unless it's some sort of Philosophy Bot. AI will learn to reason, but it will be wildly different than human reasoning. Though it might be fallible to some extent, it will not be fallible as humans are fallible.
The first intelligences to emerge from AI ( we're nowhere near there yet, but we're on our way ) will more closely resemble insect intelligence, the Hive Mind. Lots of "instinct" which can collectively interoperate, precious little "reasoning"
I have argued with the pro-robot driver crowd that empathy and moral judgments matter. It’s not just a matter of mechanical driving ability of robots being, supposedly, better.
ReplyDeleteBrian Arbenz How can a robot experience empathy?
ReplyDeleteRemove ‘artificial’ for a moment. We have trouble teaching diversity and positive values to organic intelligence. How will we teach our machines?
ReplyDeleteSo we’re left with two possibilities. AI’s will be just as biased, even racist, as the humans that program them, or they determine that humans are wrong and pose an inherent risk to the planet and other humans.
My point exactly, Dan Weese.
ReplyDelete