3 ms·
Response to Hawking: AI Helps Us Understand a World We've Endangered
Last week, a group of scientists led by Stephen Hawking warned that advances in artificial intelligence could produce uncontrollable and disastrous consequences for humanity.
The physicists and computer scientists who co-authored an article in the British newspaper The Independent raised the specter of superintelligent machines. Those machines, they wrote, may spin out of control one day, evolving without human intervention to become more powerful than, and uncaring toward, their creators.
That dystopia is compelling, and familiar. It’s been aired repeatedly every few decades as society’s fears attach themselves to each newly imagined technological threat. Of course we can't know all the repercussions of what we create, and by that token, the scientists’ article is irrefutable. That said, their fearful argument should not be the last word on AI.
- natch 12y agoThe writer is mostly talking about non-AGI AI, which results in his argument failing to address the point. I don't think Hawking was saying AI (meaning in Hawking's usage, AGI) was necessarily bad, but rather that we should just be careful about preparing for it.
- vonnik 12y agoThanks for the comment, natch. You may be right. That said, Hawking didn't use the term AGI. If he is only warning of strong AI, then there is nothing urgent or topical about his piece. It's a long way off. But Hawking seems to think his warning is urgent. That's one reason why I assume he's not just talking about AGI. In fact, many algorithms, including the deep-learning nets the LinkedIn author describes, are self-improving by combining stochastic gradient descent with a loss function. I think it is that aspect of self-improvement that frightens Hawking, because it allows machines to attain levels of performance and perfection beyond the reach of human beings. Neither I nor the author find that frightening.