Next Big Future: Interview with Ben Goertzel Who is Working on Artificial General Intelligence
robot_makes_music
· 6 months ago
Wow, DISQUS s---s.
I had a three page comment ready to go, but the log in function is so obscure I didn't know I wasn't logged in, or where the actual log-in link was. After I got that sorted, it ate my comment. Why does the default have to be "register" when you go to post comment instead of "log in" even though it doesn't say either way, it just says "enter your information, post a comment" and then has a link that doesn't say "Log In" above the post but does log you in.
nextbigfuture
· 6 months ago
Sorry to hear that you lost work and time.
I have lost comments when trying to submit them to sites with wordpress and to blogger.com
I would suggest saving any long comment page or one witha lot links and work before submitting them
cybrbeast
· 6 months ago
I'm skeptical about his grand claims of AI within a decade at modest funding. His insistent call for extra funding and big claims lead me to believe he just wants to be royally funded without really knowing how long AI development would take. If it really was possible don't you think DARPA or similar institute would already be funding this at much higher levels? If they are you would expect results sooner unless they are keeping it under wraps.
RickSchettino
· 6 months ago
It's very likely that the first 150IQ computer will come from the military...
TransAlchemy
· 6 months ago
Interesting set of questions, thank you for sharing.
GoatGuy
· 6 months ago
CybrBeast, et alia:
I don't think that his claim is all that unreasonable. I've been working in AI now for over 30 years - and like Hydrogen Fusion, it has long remained a technology that is 10 years off. Yet, there are some things that others - such as the neoSciFi almost-but-not-quite kook Kurtzweil - have noted that show a map of a path paved to deliver real AI in the nearer, not further future.
Let me bear on your (and other readers') patience to briefly explain.
One of the central tenants found by early and present day AI researchers is that the amount of information necessary, and the amount of logic necessary to sift through and select coherently related pieces of information is such that quadrillions of decisions (by projection) need to be processed per second, correlated, softly assessed (i.e. not binary, but in a more analog fashion) in order to simulate sentience. (There's a hard one: if one is "simulating" sentience in real time, is that not sentienced UNsimulated, and real?)
Assuming that the analog-comparison and analog-correlation nature of knowledged coherence is something that is but a couple orders of magnitude more "hard" in binary (i.e. to not represent correlation as 1/0 but -1.00 ... -0.99 ... -0.01 ... 0 ... +0.01 ... +0.99 ... +1.00), then instead of "just" quadrillions, maybe a quintillion or so binary decisions need to be made per second to cast votes in all areas, to follow all threads to the degree that you, or I, or a panel of experts and cynics would call "human level intelligence and sentience".
At that level of computational complexity, roughly the complexity - in digitized analog simulation! - of the human brain, or perhaps another half magnitude further along (because let's be honest: we ain't going to figure out anytime soon the finely tuned idiosyncracies of all the unknown parts of the brain anytime soon, so we'll have to over-engineer the solution, just like we overengineer bridges), something akin to human sentience should be simulatable.
And this upcoming decade is the first time in the history of computation that the posited resources are becoming available to researchers at costs that are affordable enough that hundreds if not thousands of AI researchers can spin the bottle, try their luck, have their hopes dashed then renewed on the anvil of the super-duper-computers that will be their laboratories.
In the decade following that, such zetaFLOP computers will consume a few hundred watts, and sit on people's desktops. It is then that the dream of personally-owned intelligent partners will come about. The prospect isn't just fiction, it is all but inevitable. Just as the cell-phone, invented in the early 1980's has become inevitably a tiny, almost portable fixture in everyone's pocket. Just as the bulky "personal computers" gave leave to the laptops, and now the pocket-PC's, which inevitably will become the WHOLE computer for most people. Just as decreasingly real, live humans are staffing the lathes, saws, brakes and shears in metal shops, in lieu of CNC machines.
As Kurtzweil rightfully observes, it is when hypercomputing at the IQ150 level comes about that the very nature of what innovation is, becomes offloaded to the smarties that we can (have) manufacture(d). Once abstract intelligence can move at super-human speeds, and once it has the directive and chance to invent intelligence-and-knowledge coding systems that are portable, and independent of our wetware's limits, then knowledge goes POSTAL, where it hits a singularity (borrowed from Z-plane signal processing lingo) of speed, retention, acquisition and purpose.
I don't think we need to fear that day. I just think that it may be sooner rather than later. If the gods so deem, I may be alive to see the day. In most regards, I hope to. Especially if I become a doddering old pet to some intelligence that is reasonably amused with my maunderings.
GoatGuy
RickSchettino
· 6 months ago
I think the doddering old pet isn't such an unlikely scenario in the far distant future. If we set up AGI to evolve with the value that all life is sacred it could basically solve all our problems and take care of all our needs so that we can all go surfing and paint all day...
RickSchettino
· 6 months ago
I'm a big fan of Ben's but, being picky, if you're not going to use brain scanning and emulation then you can't really use terms like human-level intelligence or superhuman-level intelligence to describe the stages of development. Whatever Ben creates will in fact be superhuman in many ways long before it's intelligent in any distinctly human ways such as emotional intelligence or real creativity. I imagine some kind of "artillect" - I think some people call it - that would make a very good personal agent. It can carry out almost anything that you ask it to do, but hasn't developed a personality or true creative abilities. I suspect Ben et al will create super-intelligent computers without human brain emulation before Ray can create an artificial persona using emulation. But like Ben says, it's probably going to be only a few years between a smart ass computer and a super-intelligent persona. Even at that point, I'm still not convinced it would be "conscious" but that's a whole 'nother debate...
enantiomer2000
· 6 months ago
The consciousness argument will go on for a while I think. Probably because there are circular definitions for it.
Conscious : aware of one's own existence, sensations, thoughts, surroundings, etc. Aware: having knowledge; conscious; cognizant
Now intelligence is something more objective (comparatively).
RickSchettino
· 6 months ago
Great point. (Although we're off topic.)
I think only awareness can be proven. I don't think you can prove consciousness.
I imagine this future conversation:
Human Being: Computer, are you aware that you exist?
Super Intelligent Computer: I am aware that I exist because you asked me. If I didn't exist we wouldn't be talking.
...and so on in a circular fashion as you stated...
Personally, I suspect that consciousness is not as simple as just being some emergent phenomenon of matter. In fact, like Amit Goswami and others I suspect that reality is an emergent phenomenon of consciousness.
All Comments
I had a three page comment ready to go, but the log in function is so obscure I didn't know I wasn't logged in, or where the actual log-in link was. After I got that sorted, it ate my comment. Why does the default have to be "register" when you go to post comment instead of "log in" even though it doesn't say either way, it just says "enter your information, post a comment" and then has a link that doesn't say "Log In" above the post but does log you in.
I have lost comments when trying to submit them to sites with wordpress and to blogger.com
I would suggest saving any long comment page or one witha lot links and work before submitting them
If it really was possible don't you think DARPA or similar institute would already be funding this at much higher levels? If they are you would expect results sooner unless they are keeping it under wraps.
I don't think that his claim is all that unreasonable. I've been working in AI now for over 30 years - and like Hydrogen Fusion, it has long remained a technology that is 10 years off. Yet, there are some things that others - such as the neoSciFi almost-but-not-quite kook Kurtzweil - have noted that show a map of a path paved to deliver real AI in the nearer, not further future.
Let me bear on your (and other readers') patience to briefly explain.
One of the central tenants found by early and present day AI researchers is that the amount of information necessary, and the amount of logic necessary to sift through and select coherently related pieces of information is such that quadrillions of decisions (by projection) need to be processed per second, correlated, softly assessed (i.e. not binary, but in a more analog fashion) in order to simulate sentience. (There's a hard one: if one is "simulating" sentience in real time, is that not sentienced UNsimulated, and real?)
Assuming that the analog-comparison and analog-correlation nature of knowledged coherence is something that is but a couple orders of magnitude more "hard" in binary (i.e. to not represent correlation as 1/0 but -1.00 ... -0.99 ... -0.01 ... 0 ... +0.01 ... +0.99 ... +1.00), then instead of "just" quadrillions, maybe a quintillion or so binary decisions need to be made per second to cast votes in all areas, to follow all threads to the degree that you, or I, or a panel of experts and cynics would call "human level intelligence and sentience".
At that level of computational complexity, roughly the complexity - in digitized analog simulation! - of the human brain, or perhaps another half magnitude further along (because let's be honest: we ain't going to figure out anytime soon the finely tuned idiosyncracies of all the unknown parts of the brain anytime soon, so we'll have to over-engineer the solution, just like we overengineer bridges), something akin to human sentience should be simulatable.
And this upcoming decade is the first time in the history of computation that the posited resources are becoming available to researchers at costs that are affordable enough that hundreds if not thousands of AI researchers can spin the bottle, try their luck, have their hopes dashed then renewed on the anvil of the super-duper-computers that will be their laboratories.
In the decade following that, such zetaFLOP computers will consume a few hundred watts, and sit on people's desktops. It is then that the dream of personally-owned intelligent partners will come about. The prospect isn't just fiction, it is all but inevitable. Just as the cell-phone, invented in the early 1980's has become inevitably a tiny, almost portable fixture in everyone's pocket. Just as the bulky "personal computers" gave leave to the laptops, and now the pocket-PC's, which inevitably will become the WHOLE computer for most people. Just as decreasingly real, live humans are staffing the lathes, saws, brakes and shears in metal shops, in lieu of CNC machines.
As Kurtzweil rightfully observes, it is when hypercomputing at the IQ150 level comes about that the very nature of what innovation is, becomes offloaded to the smarties that we can (have) manufacture(d). Once abstract intelligence can move at super-human speeds, and once it has the directive and chance to invent intelligence-and-knowledge coding systems that are portable, and independent of our wetware's limits, then knowledge goes POSTAL, where it hits a singularity (borrowed from Z-plane signal processing lingo) of speed, retention, acquisition and purpose.
I don't think we need to fear that day. I just think that it may be sooner rather than later. If the gods so deem, I may be alive to see the day. In most regards, I hope to. Especially if I become a doddering old pet to some intelligence that is reasonably amused with my maunderings.
GoatGuy
Conscious : aware of one's own existence, sensations, thoughts, surroundings, etc.
Aware: having knowledge; conscious; cognizant
Now intelligence is something more objective (comparatively).
I think only awareness can be proven. I don't think you can prove consciousness.
I imagine this future conversation:
Human Being: Computer, are you aware that you exist?
Super Intelligent Computer: I am aware that I exist because you asked me. If I didn't exist we wouldn't be talking.
...and so on in a circular fashion as you stated...
Personally, I suspect that consciousness is not as simple as just being some emergent phenomenon of matter. In fact, like Amit Goswami and others I suspect that reality is an emergent phenomenon of consciousness.