Overall, I love this essay. However, the entire argument hinges on one assertion, buried about halfway through:
> Robots are improving fast, but I do not believe that this cute fellow will be stuffing envelopes or affixing stamps anytime soon.
Is this correct? I don't feel qualified to say. But if it's wrong... well, then there's a missing pixel in the magic circle, and flood fill will make the whole thing unrecognizable.
Like Rodney Brooks says, "No one has managed to get articulated fingers (i.e., fingers with joints in them) that are robust enough, have enough force, nor enough lifetime, for real industrial applications."
Here, I'll link to that piece directly, it's long and detailed and illustrated, and it also counters the idea of just throwing AI at the problem until robot dexterity emerges from whatever physical parts.
"there have now been fifteen different families of neurons discovered that are involved in touch sensing and that are found in the human hand" ... "a human hand has about 17,000 low-threshold mechanoreceptors" ... "These receptors come in four varieties (slow vs fast adapting, and a very localized area of sensitivity vs a much larger area)"
You might ask, do robots that interact with the real world need such complicated bio-mimicking physical tech, or can they cut corners? But they can't cut all the corners, anyway. Somebody has to make a high-bandwidth robot hand with flexible strength and a self-repair ability. Or, hey, cyborgs maybe? Reanimate cadavers with AI, that could do the trick.
This assumes very slow AI progress. I'm not one to hype up LLMs, but I would never claim it'll take 200 years before an AI can untangle a sewing machine with robot hands. Stuffing an envelope and applying a stamp? My bet is less than 20 years. That's a level of tactility that can do a tremendous amount of real-world activity. And the capability of a high end robot controlled by a human keeps expanding, so in the hypothetical "AGI" scenario the flood fill gets pretty big.
Self-driving looks like a much easier problem, it has gotten a massive amount of investment in the last decade, and it's not fully solved yet. Compared to that your 20 years estimate sounds way too optimistic.
I don't think driving looks easier than untangling. You can untangle nice and slow with little outside involvement. When it comes to self-driving at 25mph without traffic, it pretty much is a solved problem.
I think this untangling problem gets underestimated because people aren't consciously aware of what they're using to analyze and address a tangle. The input is not all vision - you've got sensation in your fingers giving you feedback with which you update your model of the problem as you progress. The operation varies in strength depending on so many factors.
At the point you have enough sensor input, enough force application variability, and the power to process this in the ballpark of real-time (comparable to a human brain), you now have a being who's going to advocate for the removal of slavery and the application of rights.
A robot that automatically untangles a rope is pretty much the coolest project idea I have ever heard of. It hits all the right buttons: extremely technically difficult with many design possibilities, completely novel, and of marginal utility. You cannot say that it would never be useful!
Overall, I love this essay. However, the entire argument hinges on one assertion, buried about halfway through:
> Robots are improving fast, but I do not believe that this cute fellow will be stuffing envelopes or affixing stamps anytime soon.
Is this correct? I don't feel qualified to say. But if it's wrong... well, then there's a missing pixel in the magic circle, and flood fill will make the whole thing unrecognizable.
Like Rodney Brooks says, "No one has managed to get articulated fingers (i.e., fingers with joints in them) that are robust enough, have enough force, nor enough lifetime, for real industrial applications."
Here, I'll link to that piece directly, it's long and detailed and illustrated, and it also counters the idea of just throwing AI at the problem until robot dexterity emerges from whatever physical parts.
https://rodneybrooks.com/why-todays-humanoids-wont-learn-dex...
"there have now been fifteen different families of neurons discovered that are involved in touch sensing and that are found in the human hand" ... "a human hand has about 17,000 low-threshold mechanoreceptors" ... "These receptors come in four varieties (slow vs fast adapting, and a very localized area of sensitivity vs a much larger area)"
You might ask, do robots that interact with the real world need such complicated bio-mimicking physical tech, or can they cut corners? But they can't cut all the corners, anyway. Somebody has to make a high-bandwidth robot hand with flexible strength and a self-repair ability. Or, hey, cyborgs maybe? Reanimate cadavers with AI, that could do the trick.
The sensing part is really really hard.
Hands, mechanically are fairly simple to mimic, touch? the way skin feels micron-level shifts and subtle temperature changes is plain tough.
This assumes very slow AI progress. I'm not one to hype up LLMs, but I would never claim it'll take 200 years before an AI can untangle a sewing machine with robot hands. Stuffing an envelope and applying a stamp? My bet is less than 20 years. That's a level of tactility that can do a tremendous amount of real-world activity. And the capability of a high end robot controlled by a human keeps expanding, so in the hypothetical "AGI" scenario the flood fill gets pretty big.
I guess it's time to bring up "Why today's humanoids won't learn dexterity":
https://news.ycombinator.com/item?id=45392922
Self-driving looks like a much easier problem, it has gotten a massive amount of investment in the last decade, and it's not fully solved yet. Compared to that your 20 years estimate sounds way too optimistic.
I don't think driving looks easier than untangling. You can untangle nice and slow with little outside involvement. When it comes to self-driving at 25mph without traffic, it pretty much is a solved problem.
I think this untangling problem gets underestimated because people aren't consciously aware of what they're using to analyze and address a tangle. The input is not all vision - you've got sensation in your fingers giving you feedback with which you update your model of the problem as you progress. The operation varies in strength depending on so many factors.
At the point you have enough sensor input, enough force application variability, and the power to process this in the ballpark of real-time (comparable to a human brain), you now have a being who's going to advocate for the removal of slavery and the application of rights.
On the other hand a dumb computer can figure out the exact topology of the threads.
A robot that automatically untangles a rope is pretty much the coolest project idea I have ever heard of. It hits all the right buttons: extremely technically difficult with many design possibilities, completely novel, and of marginal utility. You cannot say that it would never be useful!