Based on the current rate of progress in AI/robotics, when do you think humanoid androids will be comparable in quality to Data from Star Trek: TNG, or David from Prometheus?
Some additional questions to consider for this discussion:
What are the major hurdles still to overcome?
How soon after the first public demonstration of such a model would they be common in society?
What effects would these androids have on culture, business, politics, and society in general?
Please vote in the poll I've created. I may use these results for a timeline prediction, depending on how many responses I get. Thanks!
I think AGI is likely sometime in the next 50-150 years. As I've talked about in the status updates I think that we need to understand the brain more first and get better at building mathematically similar simulations. They don't have to be the same as the human brain exactly, but more topological similarity to animal brains than our current models provide will be required I think. This tech is still quite a ways off even if we're firing on all cylinders as a civilization. Which gets into why this is a range. I won't go into detail, we all talk about it more than enough. But if our civilization is in a period a global decline in the latter half of this century it could stifle technological progress and I don't see us recovering until the 22nd century if that happens.
In terms of physical appearance I think the physical robotics are a secondary and lesser challenge. Robotics have been progressing rapidly and in 50 years organic movements will be very convincing. By the time we've created AGI I expect that it will just be a matter of giving them robot bodies we've already had for decades.
This is kind of cheating! If we're talking about an android like Data that is clearly alive, possesses their own consciousness, and can creatively reason; I think that's 50-150 years away. If we're talking about something that looks like a convincing android, moves efficiently but still robotically, and has a really powerful narrow AI powering it that can trick laymen into thinking its actually alive--that's much much closer, more like 30-100 years. I lean towards the earlier part of this range for a faux-android.
I believe every country in the world or at least USA & Europe have these. Same as China with the digital society they created I'd look to the news from there as an idea of when they will move this along. I think in the span of these 2 months there is Chinese androids that look pretty real to me. The demand for change is being created here in the USA as usual so for now the androids look like androids as annoying as that is.
wjfox wrote: ↑Fri Sep 15, 2023 8:30 pm
Based on the current rate of progress in AI/robotics, when do you think humanoid androids will be comparable in quality to Data from Star Trek: TNG, or David from Prometheus?
Some additional questions to consider for this discussion:
What are the major hurdles still to overcome?
How soon after the first public demonstration of such a model would they be common in society?
What effects would these androids have on culture, business, politics, and society in general?
I can't really comment on the latter two, because those are their own variables, and depending on how advanced generative AI becomes, the question will inevitably be raised about just how necessary perfectly lifelike humanoid robots even are. On some level, I think they'll definitely be around because god knows a lot of people aren't going to be satisfied with VR and computer screens, but perhaps they'll only ever be a novelty, something created by the coming AI-ran manufacturers to show that we really can create Data, Deckard, and sexbots.
The primary hurdles are entirely in motion. I believe you asked this very question in the previous forum, and unfortunately, not much progress has occurred since then in granting robots truly convincing motion, though to be fair, the two main limitations are power and articulation.
And the reason wy articulation is such a big deal is because, to mimic a human, you need to mimic muscle movements, many of which are imperceptible to the conscious mind but your unconscious mind WILL pick up. That level of muscular articulation is theoretically possible, but requires some hefty computing power and very detailed mechanical work.
After all, remember that the human brain isn't just powering higher cognition: it's powering our entire bodies. In fact, that's what most of its power and matter is used for. All our high intellectual abilities are more or less what we in Louisiana call "lagniappe": extra goodies left over from how overpowered our brains are. All those unconscious functions are what allows most people to pass the Uncanny Valley.
Don't ask me the details of how I know or anything, but some sexrobot manufacturers have also started utilizing hyperrealism. Considering how desperate humans can get to stimulate those regions, the first widely manufactured lifelike androids are undoubtedly going to be gynoids, and they're going to be used for exactly that purpose.
The million dollar question is "When?" because as anyone even passingly following AI progress knows, it all really hinges on that. If AI was a nonfactor— if AI Doomers and Safetyists won and forced all progress to slow down to a crawl until the 22nd century, well first of all, I think we'd still have decades' worth of economic impacts out of AI from even 5 years ago and GPT-4-level AIs could probably power a full century's worth of social, political, and economic issues without any further progress. But let's say that's actually the case, and GPT-4 is the best AI gets for the rest of our lives, meaning that future progress has to be done driven 95% by human ingenuity like it has been ever since the 80s. I assume this is what most futurologist timelines are based upon, because these are relatively reliable and don't rely on "AGI is the God Mode cheat, all Tech Trees Unlocked by 2029" to solve everything at once.
Under those circumstances of pure human artistry and mechanical/computer engineering progressing further and further into the future, I can imagine that some big lab would definitely attempt to get computer-powered hyper-articulated artificial muscles into a hyperrealistic-style android and be able to pull it off convincingly by the late 2030s at the earliest. Maybe not necessarily 'perfectly," but just looking at how well Ameca convinces us of its intelligence through decent-enough facial muscles, another 15-20 years of progress and funding could definitely get us there.
Of course, if Q* is an AGI and we lick the basics of AGI by 2024, it's probably going to be child's play within 2-3 years once all the problems are laid out for us to solve (and then solved for us for that matter).
But again, how useful are perfectly humanlike androids? It's kind of like asking how useful genetic engineering is. If you only run with that avenue of technological progress, "we can create genetically-perfect superhumans, what are the implications" have been explored in sci-fi for decades. Easily you can assume genetic social stratification and discrimination that becomes full casteism even in the West as Brave New World-type speciation occurs. But IRL, the first genetic Aryan that is created is already deeply obsolete by some vastly superior AI superintelligence (and maybe even transhuman superintelligence as well) so whatever "elite" status he might take in the economy is moot barring pure novelty.
Similar situation here. A perfectly lifelike humanoid robot would be dangerously disruptive if they were they peak of technology and engineering, because such an android would essentially be a true "artificial human" in full, with all our strengths and none of our weaknesses and able to blend in with our population perfectly. There's a lot you can do with that in science fiction.
But in real life, the question inevitably becomes, "Oh, the Overmind already does all high-cognitive labor and the only jobs that exist are ones humans elect to do to keep ourselves busy or keep old traditions alive." It becomes novelty at that point. "My co-worker is an android!" Or conspiracy theories that the human governor is actually an artificial human, their birth certificate generated by an AI and all evidence of their early life generated video and text! A paranoic's worst nightmare, really (then again, those types already believe this is the case; there's even a recognized verified mental illness for that).
Assuming a lot of people don't want to get involved in VR for whatever reason, or at least not fully-immersive VR (who knows what the adoption rate'll be, but I strongly doubt it'll be anywhere near 100%), androids will probably be most widely used as servants for those who either want humans but it's illegal to own people, or those who just love the novelty of such lifelike entities.... and again, THAT purpose.
Though of course, again, if GPT-4 or even GPT-5 is about as good as AI progress ever gets, then there is an argument that lifelike androids will serve a much greater, much more disruptive role in society. Probably the most disruptive possibility, actually. Because imagine GPT-5 is Turing Test-passable, but that's as good as we do, and yet we get artificial humans anyway.
And remember my friend, future events such as these will affect you in the future
