Sep 10Liked by Chris Best, Mills, ℭ𝔬𝔫𝔰𝔱𝔞𝔟𝔩𝔢 𝔬𝔣 𝔔𝔲𝔞𝔩𝔦𝔞
OK I rewatched this because it's deeply interesting to me. Because I think I'm in Mills' camp about superintelligence, I want to argue against myself for a second. So this is my devils advocacy for a superintelligent computer.
When I argued against superintelligence when Mills posted this on Notes, I argued that because the inputs are comprehensible to humans the outputs have to be comprehensible to humans.
so if I wanted to set out to create a superintelligence that was utterly incomprehensible to humanity, what would I need to do?
It would need to be able to observe the universe with a greater level of precision to our own. It would need to be able to make INDEPENDENT observations and NOT SHARE THEM. If humans are making the observations, we'll have as many brains chewing on the data as we'll have CPU's. So the superintelligence would need to make its own observations, at higher fidelity, and keep them secret.
It would also need to understand some things humans understand but which aren't objectively quantifiable. Transcendentals come to mind, beauty goodness truth etc. These are important because they speak to ways that humans value things and convey meaning and weigh options. A superintelligence would need to understand them ONLY AT A MINIMUM to be able to communicate with humans.
A superintelligence I think would need to be able to demonstrate proofs I think to humans too. Like a superintelligence that was capable of lying could say "the universe is an infinite torus and if you move far enough in one direction you'll arrive back where you started." And how would we know if it told us? There would need to be some kind of trust established. How do we know a superintelligence wouldn't lie to us, like Dr. Manhattan etc? And if we program a computer to NOT LIE can it BE a superintelligence in the first place?
All of these things feel like I'm just describing a person. But "what would a superintelligence look like" is an interesting question, I'd be interested to hear your thoughts on it. Maybe I'm hamstringing myself because I don't believe it's possible in the first place, but as a thought experiment it's been illuminating.
If I could describe a superintelligence it would be all about speed. Essentially it would just be able to think and learn faster than us.
Now I'm in the camp where I think superintelligence is impossible without a body. It would need some sort of physical feedback to understand the world. It would have to make mistakes and learn.
My guess though is that AI will peak like some sort of hyper intelligent animal. Probably incapable of questioning it's own existence or examining it's own thoughts from an outside perspective.
is speed really a superintelligence? Would a race of superintelligent computers have gone from burial-mound civilization to landing on the moon in 25% the time it took humans? And if so--is that really that much of an achievement since humans also did that regardless of how long it took?
That makes it sound like any outcome of superintelligence can be achieved by humans if we have patience. And then it's not that super at all--it's normal human intelligence just faster.
Now, a hyper intelligent animal is a cool idea. And that poses some alternate questions. Does a hyperintelligent cat NOT chase mice???? I almost think that it's like domestication. Animals are wild until adopted by humans and domesticated. Computers were "born" domesticated. Can they be wilded? Would a superintelligent computer need to be "wild" in order to develop these attributes we think a superintelligent computer would need?
Have a i spilled the banks of sanity and overflowed into science fiction???
Sep 10Liked by Chris Best, Mills, ℭ𝔬𝔫𝔰𝔱𝔞𝔟𝔩𝔢 𝔬𝔣 𝔔𝔲𝔞𝔩𝔦𝔞
I really enjoyed this. It was dynamic interesting and just long/short enough to be digested in my working day.
On a technical note, video quality was great and the vertical twin set up easy to follow. Although it might be that subjects that move around a lot use a Lav microphone. I only say that because Mills’ audio dropped out a few times which would hamper you scraping the audio to make a quality podcast. Also at times incorrect captioning added a little confusion to the context. But I’m not sure if that could be edited in post.
Mind expanding. Thanks to this I just watched a TED talk with David Deutsch. What a brilliant communicator. Now I need to rest my brain. Thanks again - amusing and informative, especially on the Apple stuff of which I know nothing. I'm Android (not literally).
OK I rewatched this because it's deeply interesting to me. Because I think I'm in Mills' camp about superintelligence, I want to argue against myself for a second. So this is my devils advocacy for a superintelligent computer.
When I argued against superintelligence when Mills posted this on Notes, I argued that because the inputs are comprehensible to humans the outputs have to be comprehensible to humans.
so if I wanted to set out to create a superintelligence that was utterly incomprehensible to humanity, what would I need to do?
It would need to be able to observe the universe with a greater level of precision to our own. It would need to be able to make INDEPENDENT observations and NOT SHARE THEM. If humans are making the observations, we'll have as many brains chewing on the data as we'll have CPU's. So the superintelligence would need to make its own observations, at higher fidelity, and keep them secret.
It would also need to understand some things humans understand but which aren't objectively quantifiable. Transcendentals come to mind, beauty goodness truth etc. These are important because they speak to ways that humans value things and convey meaning and weigh options. A superintelligence would need to understand them ONLY AT A MINIMUM to be able to communicate with humans.
A superintelligence I think would need to be able to demonstrate proofs I think to humans too. Like a superintelligence that was capable of lying could say "the universe is an infinite torus and if you move far enough in one direction you'll arrive back where you started." And how would we know if it told us? There would need to be some kind of trust established. How do we know a superintelligence wouldn't lie to us, like Dr. Manhattan etc? And if we program a computer to NOT LIE can it BE a superintelligence in the first place?
All of these things feel like I'm just describing a person. But "what would a superintelligence look like" is an interesting question, I'd be interested to hear your thoughts on it. Maybe I'm hamstringing myself because I don't believe it's possible in the first place, but as a thought experiment it's been illuminating.
Don’t worry. I’ll make a post about this and clean it all up. Then we will all understand.
We've all been focused on superintelligence and forgotten about the superextelligence walking among us
Superintelligent sounds like you’re trying to reimagine God the Father.
Superextelligence sounds like you’re trying to reimagine Jesus.
I don’t know how kosher that is but I agree when we’re talking about superintelligence were butting awfully close to concepts of divinity
Yes. Also, I would say that the word of God would be an example of extelligence.
And, by extension, superextelligence walking among us, would be God, The Word.
If I could describe a superintelligence it would be all about speed. Essentially it would just be able to think and learn faster than us.
Now I'm in the camp where I think superintelligence is impossible without a body. It would need some sort of physical feedback to understand the world. It would have to make mistakes and learn.
My guess though is that AI will peak like some sort of hyper intelligent animal. Probably incapable of questioning it's own existence or examining it's own thoughts from an outside perspective.
You never know though.
is speed really a superintelligence? Would a race of superintelligent computers have gone from burial-mound civilization to landing on the moon in 25% the time it took humans? And if so--is that really that much of an achievement since humans also did that regardless of how long it took?
That makes it sound like any outcome of superintelligence can be achieved by humans if we have patience. And then it's not that super at all--it's normal human intelligence just faster.
Now, a hyper intelligent animal is a cool idea. And that poses some alternate questions. Does a hyperintelligent cat NOT chase mice???? I almost think that it's like domestication. Animals are wild until adopted by humans and domesticated. Computers were "born" domesticated. Can they be wilded? Would a superintelligent computer need to be "wild" in order to develop these attributes we think a superintelligent computer would need?
Have a i spilled the banks of sanity and overflowed into science fiction???
https://www.youtube.com/watch?v=VUCNBAmse04
THIS DROPPED FIVE HOURS AGO HOW DID THEY KNOW
superintelligence
[rollsafe.gif]
I really enjoyed this. It was dynamic interesting and just long/short enough to be digested in my working day.
On a technical note, video quality was great and the vertical twin set up easy to follow. Although it might be that subjects that move around a lot use a Lav microphone. I only say that because Mills’ audio dropped out a few times which would hamper you scraping the audio to make a quality podcast. Also at times incorrect captioning added a little confusion to the context. But I’m not sure if that could be edited in post.
Looking forward to more of these :-)
Dammit, I should’ve also said you had me making notes. There was so much juicy content in your conversation. :-)
Jumped in for the Apple summary and got bonuses, including realities that humans can't access, for maybe as long as we're constrained by time. :-D
I laughed hard at the way Mills described the communication style from the humans of Apple. I think they caught the PBS bug. They're so modulated.
How do I join the beta for live? LinkedIn and other platform stats are sinking. Can give structured feedback or review it as I’m a product manager ;)
Love this, thank you.
Mind expanding. Thanks to this I just watched a TED talk with David Deutsch. What a brilliant communicator. Now I need to rest my brain. Thanks again - amusing and informative, especially on the Apple stuff of which I know nothing. I'm Android (not literally).