Tim O’Reilly forwarded a wonderful article concerning the OpenAI cleaning soap opera to me: Matt Levine’s “Cash Stuff: Who Controls Open AI.” I’ll skip most of it, however one thing caught my eye. Towards the top, Levine writes about Elon Musk’s model of Nick Bostrom’s AI that decides to show the world to paperclips:
[Elon] Musk gave an instance of a synthetic intelligence that’s given the duty of selecting strawberries. It appears innocent sufficient, however because the AI redesigns itself to be simpler, it would determine that the easiest way to maximise its output could be to destroy civilization and convert your entire floor of the Earth into strawberry fields.
That will get me, however not in the way in which you suppose. It’s personally poignant, for causes fully totally different from the AI-doomerism cults that Musk, Bostrom, and others are propagating.
After I was a graduate scholar at Stanford, I used to be driving round with a good friend by the infinite maze of parking heaps and strip malls in that nondescript a part of Silicon Valley the place Sunnyvale, Santa Clara, and Cupertino come collectively. My good friend identified the window and mentioned, “That’s the place my father’s farm was.” I requested what his father grew; it was very tough to think about a farm at that location. He grew strawberries. And what occurred to the farm? His father misplaced it when he was put right into a World Battle II internment camp for Japanese. An actual property investor ended up with it. My good friend’s father ultimately dedicated suicide. The farm grew to become a car parking zone.
This will get me again to an argument that I’ve made in older Radar articles: Our fears of AI are actually fears of ourselves, fears that AI will act as badly as people have repeatedly acted. We don’t want AI to show the world into strawberries any greater than we’d like it to show the world into parking heaps. We’re already turning the world into parking heaps, and doing so with out regard to the human value. We’re already spewing CO2 at a charge that may quickly make the world uninhabitable for all however the few who can insulate themselves from the results. If we’re going to resolve these issues, it gained’t be by know-how. It’s by discovering higher people than Elon and, I worry, Sam Altman. We don’t have an opportunity to resolve the AI downside if we will’t remedy the human downside. And if we don’t remedy the human downside, the AI downside is irrelevant.