3 ms·
One way to achieve superhuman intelligence in AI is to make humans dumber.
by palmotea 2y ago
One way to achieve superhuman intelligence in AI is to make humans dumber.
- imoverclocked 2y agoThat’s only if our stated goal is to make superhuman AI and we use AI at every level to help drive that goal. Point received.
- yieldcrv 2y agoRight, superhuman would be relative to humans but intelligence as a whole is based on a human ego of being intellectually superior
- caseyy 2y agoThat’s an interesting point. If we created super-intelligence but it wasn’t anthropomorphic, we might just not consider it super-intelligent as a sort of ego defence mechanism. Much good (and bad) sci-fi was written about this. In it, usually this leads to some massive conflict that forces humans to admit machines as equals or superiors. If we do develop super-intelligence or consciousness in machines, I wonder how that will all go in reality.
- yieldcrv 2y agoSome things I think about are how different the goals could be For example, human and biological based goals are around self-preservation and propagation. And this in turn is about resource appropriation to facilitate that, and systems of doing that become wealth accumulation. Species that don't do this don't continue existing. A different branch of evolution of intelligence may take a different approach, that allows its affects to persist anyway.
- caseyy 2y agoThis reminds me of the "universal building blocks of life" or the "standard model of biochemistry" I learned at school in the 90s. It held that all life requires water, carbon-based molecules, sunlight, and CHNOPS (carbon, hydrogen, nitrogen, oxygen, phosphorus and sulfur). Since then, it's become clear that much life in the deep sea is anaerobic, doesn't use phosphorus, and may thrive without sunlight. Sometimes anthropocentrism blinds us. It's a phenomenon that's quite interesting.
- boringg 2y agoThe cultural revolution approach to AI.
- 6510 2y agoI thought: A group working together poorly isn't smarter than the smartest person in that group. But it's worse, A group working together poorly isn't smarter than the fastest participant in the group.
- jimmygrapes 2y agoanybody who's ever tried to play bar trivia with a team should recognize this
- rightbyte 2y agoWhat do you mean? You can protest against bad but fast answers and check another box with the pen.
- tengbretson 2y agoBeing timid in bar trivia is the same as being wrong.
- trentlott 2y agoThat's a fascinatingly obvious idea and I'd like to see data that supports it. I assume there must be some.
- 6510 2y agoI understand but the bug is closed now.
- xrd 2y agoIf you came up with that on your own then I'm very impressed. That's very good. If you copied it, I'm still impressed and grateful you passed it on.
- BrenBarn 2y agoWhat if ChatGPT came up with it?
- card_zero 2y agoRaises hand https://news.ycombinator.com/item?id=43303755 https://news.ycombinator.com/item?id=43303755 I'm proud to see it evolving in the wild, this version is better. Or you know it could just be in the zeitgeist.
- xrd 2y agoI'll never forget you, card_zero.
- ryao 2y agoThis reminds me of the guy who said he wanted computers to be as reliable as TVs. Then smart TVs were made and TV quality dropped to satisfy his goal.
- SoftTalker 2y agoThe TVs prior to the 1970s/solid state era were not very reliable. They needed repair often enough that "TV repairman" was a viable occupation. I remember having to turn on the TV a half hour before my dad got home from work so it would be "warmed up" so he could watch the evening news. We're still at that stage of AI.
- ryao 2y agoThe guy started saying it in the 80s or 90s when that issue had been fixed. Ge is the Minix guy if I recall correctly.