what are your thoughts on computer "consciousness" or "sentience" or "self-awareness" i am not sure you agree with goertzel , but from what ive read it seems like he is pretty confident along with quite a few others that we will achieve agi in the not to distant future (5-10 years or sooner).
I agree w/ Ben on a lot of issues, but the 5-10 time-frame seems too soon. Mostly because I kind-of back-of-the-envelop sketched a "road to machine sentience" some 5 years ago, and it has 10 or 15 or steps on the road, and "it doesn't seem that hard", and here I am 5 years later, and have taken a quarter step of the first step. Successfully, but golly, for something "not that hard" it sure is taking a long time.
There's a way to put it in perspective: ordinary "big science" takes decades. Want to built a telescope? Prepare to spend a few decades. Want to launch a science satellite? a few decades ... oh, and raise the many millions needed, and assemble the staff of dozens, hundreds needed to develop and test the instruments. Examples: look at timeline of LSST, or the Thirty-Meter-Telescope, or CERN, or the Event Horizon Telescope, SKA, the gravitational wave detectors, or any satellites or spacecraft missions to Jupiter or Mars or Pluto. Decades is the norm.
And this is, in a sense "known knowns" - "known science" ... AGI still has at least a few "unknowns" -- likely roadblocks that haven't even been encountered yet. Oh, and I'm working as a lone wolf. As is most everybody; there's no consensus, no collaboration, no organization, no funding. And then random crap happening behind closed corporate walls, which, from all indications, is funded but chaotic.
I think deep in his gut Ben might sense this is the case, but his optimism and natural promotional talents get the better of him. :-)
but what about Artificial Consciousness? is there narrow vs general, or what do you think?
My current favorite theory of consciousness is basically that we as humans are entities that observe the world and locate/identify patterns, our consciousness is essentially that mechanism observing itself and locating identifying patterns of itself. this coupled with theory that computation is really just a simpler form of thinking (at specific complexity levels the two become indistinguishable) and you could theorize that creating consciousness is doable and something that could be able to scale up..
Yes, that is an entirely valid and reasonable "theory" of (self-) consciousness. And that variant might actually be buildable in a decade(s?) time frame, maybe, if there was actual, sustained, coordinated effort (on the scale of other "big science" projects) The problems:
Combinatoric explosion remains a challenge. Neural nets perform a kind of "dimensional reduction" but they reduce too much, and there's this .. well, extracting the "right" patterns is still a problem of threading-the-eye-of-the-needle, but we can't really see quite where the eye is, so its confusion-prone. Easy to get mislead.
Early "sentient" systems are likely to be utterly psychotic, and nothing at all like humans. So like Star Trek Next Gen had episodes where the writers tried to imagine completely alien life-forms, like the sentient sand-pile, or the binarians, or Q or even Moriarty. So think of that, but even more alien and non-humanistic in values.
Speaking of humans, we have an invertebrate "gut brain" that controls, heart, breathing, digestion, its a billion years old. And a sympathetic nervous system, from the vertebrates, that controls/organizes the gut brain. And a mammalian neocortex that gives us rationality and language. But the vagus nerve still connects our gut to our facial expressions, which is why we wear our heart on our sleeves, which is why Star Trek's Spock or Data had trouble understanding humans... We are literally ruled by gut emotions, when we are not ruled by neuro/bio/hormonal cycles that control addiction (drug addiction, gambling addiction, sex addiction, video-game addiction, addiction/desire to be in the "flow state" -- intellectuals are addicted to erm.. the brute pleasure of thinking. The flow state.)
So this is what makes us human, and the first sentient machine is going to be so utterly different from that, that it will be psychotic by any and every human definition. And as AGI researchers, we're not prepared for that.
Currently existing examples of proto-AGI, "almost-AGI" are ... corporations. Self-aware, with super-human powers, working 24x7, doing .. things. Nominally controlled by humans, but not really. Once a corporation gets going, every human in it is replaceable. They exists as entities, not just legally, but in reality.
But following this same line of reasoning, WWI was a kind-of industrial machine that produced dead bodies, and only halted when it ran out of raw materials (young men to kill). I'm using WWI as an example of a near-AGI-level machine that was utterly alien, and completely hostile to humans, and it ran out of control: politicians could not stop it, generals could not stop it. Now, it ran partly as a brain-virus -- it was a memetic structure, a meme, that entrained entire economic systems, a memetic viral infection that could not be shaken off.
Compare the psychological state of European humans in 1915 to the psychological state of humans on facebook in 2019, and then realize that the modern proto-AGI's are not going to be like Star Trek's Data, created in some isolated laboratory, having a simple mechanical body, and a video camera for an eye, and a microphone for ears --- no -- its going to be getting sensory input from a firehose of petabytes per second of facebook feeds and google searches and youtube views. It will be a kind-of nervous-system, like the gut-brain, but instead of being attached to the heart and lungs and intestines, its attached to social media, i.e. to society-at-large. That's where, how, what the shape of the first AGI will be. The wiring of it is already being laid down. The higher-function intelligence is still missing, but its already gaining an IQ point or two per year. (and yes, this variant is not open-source; its corporate-controlled. Or in the case of China, controlled by the CCP.) We will all become a part of the Borg, and willingly, and joyfully, and probably not quite even aware that it is even happening.
Phew, I was going to post a short reply, but whatever.
#agi #opencog