注册并分享邀请链接,可获得视频播放与邀请奖励。

bayes
@bayeslord
加入 January 2019
3.8K 正在关注    46.2K 粉丝
I prefer to contest the new species framing altogether, but let’s take it seriously for a moment. The concern that humanity is being totally dominated and supplanted in an evolutionary sense is a reasonable one. However, it’s not true that evolutionary fitness is in general determined solely (or mostly) by intelligence. This was true for humans when there were no other strong intelligences around, but we’re here now, and so it will not be such a slam dunk for any would be humanity destroying AI. Also, this world, ie a hypothetical new species’ would-be evolutionary environment, is run by humans. So it is completely up to humans to shape things in the way we want them. We can make Earth inhospitable to would-be competing replicators. Coordination is challenging of course. And things will only get harder as we scale, because AIs will mediate or facilitate an increasing fraction of humanity’s immediate control over matter, which is a potentially very vulnerable vector. Again, I want to be perfectly clear, this is the default state of things if we just build any random optimization demon and let it loose. But that is far from the limits of better things we could do in principle. It doesn’t help that many tech people seem intent on purposely enabling species succession, as in “Give the universe to AI once we solve alignment”, or “If AI beats us, that’s nature and therefore good”. One is a sad, tragic plan, not to mention overconfident and low resolution, while the other is a kind of perverted naturalism. We don’t have to do this. We can solve intelligence and use it wisely. AI could be another mere tool for us. Believing it must be otherwise is a failure of imagination. We went through something similar with physical strength. Yes sometimes the mechanization of force intersects with our lives in negative ways, but it is largely positive. Intelligence is categorically different, to be sure, but mastery over the production of intelligence need not mean ceding the cosmic endowment to what will almost surely be poorly understood minds, out of whatever weird and pathetic reflex people in the San Francisco Bay Area can’t help but exhibit. We need to be much more humble and not assume we can solve the future of the universe in the next fifteen years. We can try, and should, but it is so little time I believe we will need to pace the cosmic frontier. There are many mysteries left to solve before we can differentiate between heaven and hell. There are many mysteries left to solve in general. We are still children, and this is beautiful. There is no need to rush into adulthood.
显示更多