It’s not X. It’s Y. Multiple times, over and over. We need to stop with the AI stuff.
The authors presumably don't have English as their first language.
The Robotic Dexterity Deadlock
41–50 of 53 posts
Re: The Robotic Dexterity Deadlock
#42I couldn't read it except on archive.org. Here's the link to save people trouble https://web.archive.org/web/20260227201321/https://www.origa...
That loses some of the pictures. Especially the one of their custom motor, which is the new development.
Re: The Robotic Dexterity Deadlock
#43I couldn't read it except on archive.org. Here's the link to save people trouble https://web.archive.org/web/20260227201321/https://www.origa...
Thanks, we've added that to the toptext also.
Re: The Robotic Dexterity Deadlock
#44That's a nice piece of motor engineering. It's well known that high ratio gearboxes for robots are a headache. Back driveability doesn't work, and tiny teeth are fragile. Comments on this go all the way back to Feynman writing about his time spent engineering automatic gunnery aiming systems in WWII. This new discovery is that gearbox problems mess up a machine learning system. It's trying to track gearbox noise and…
Why is a pancake motor too large for fingers? Human motors aren't placed on the fingers either...
Re: The Robotic Dexterity Deadlock
#45It’s not X. It’s Y. Multiple times, over and over. We need to stop with the AI stuff.
Deciding whether A is an X or a Y is a really basic part of why we're all communicating. Suspicion of em dashes is one thing, but once you start getting nervous on seeing "It’s not X. It’s Y." then you're just going to get paranoid. The fundamentals of an LLM is to statistically match their output with the corpus. The tics they have are really common in natural human usage too.
This line of argument would fall down if it turns out that a human with statistically normal output is a bizarre-sounding human.
Re: The Robotic Dexterity Deadlock
#46Earlier quoted context omitted.
Deciding whether A is an X or a Y is a really basic part of why we're all communicating. Suspicion of em dashes is one thing, but once you start getting nervous on seeing "It’s not X. It’s Y." then you're just going to get paranoid. The fundamentals of an LLM is to statistically match their output with the corpus. The tics they have are really common in natural human usage too.
Did you read the article? It's all AI tells. The tone may as well be a fax machine
I didn't reply to the comments talking about the AI tells. I replied to the comment that is making a bad argument. It doesn't matter to me whether the article is or isn't LLM assisted.
Re: The Robotic Dexterity Deadlock
#47That's a nice piece of motor engineering. It's well known that high ratio gearboxes for robots are a headache. Back driveability doesn't work, and tiny teeth are fragile. Comments on this go all the way back to Feynman writing about his time spent engineering automatic gunnery aiming systems in WWII. This new discovery is that gearbox problems mess up a machine learning system. It's trying to track gearbox noise and…
Why is a pancake motor too large for fingers? Human motors aren't placed on the fingers either...
This is what a pancake motor looks like.[1] Thin and flat. Wrong form factor for fitting into a wrist.
Strings ("tendons") have their own problems. Wear at bearing points, mostly. You need an opposed pair, too. So there's a motor, a winding drums, and something like a spring to maintain tension. Or two motors and two winding drums, which gives better control.
That's why snake robots have never caught on. They're good painting robots, because they can get into tight spots. But repetitive cycling wears the tendons at the same place on each cycle.
There was a good application - Tesla's snake charging robot.[2] All the complicated stuff is in the pedestal, and you can surround that with concrete to prevent damage if it's hit by a car. The snake part has no motors, just discs and strings, plus the connector and probably a cell phone camera at the end. It can be made easily replaceable, like a gas station hose.
But some people thought it was too creepy.
Re: The Robotic Dexterity Deadlock
#48Earlier quoted context omitted.
Did you read the article? It's all AI tells. The tone may as well be a fax machine
I didn't reply to the comments talking about the AI tells. I replied to the comment that is making a bad argument. It doesn't matter to me whether the article is or isn't LLM assisted.
Is the not X it's Y in large frequency not an AI tell? I think they're pointing out it has a tell. Humans use those but not that often.
Re: The Robotic Dexterity Deadlock
#49That's a nice piece of motor engineering. It's well known that high ratio gearboxes for robots are a headache. Back driveability doesn't work, and tiny teeth are fragile. Comments on this go all the way back to Feynman writing about his time spent engineering automatic gunnery aiming systems in WWII. This new discovery is that gearbox problems mess up a machine learning system. It's trying to track gearbox noise and…
(I am not a mechanical engineer) Are capstan drives a solution to some of these gearing problems?
capstan drives help with backlash (though they are in generally less stiff than gears, so it's not without tradeoffs), but they don't really help with any of the other issues.
Re: The Robotic Dexterity Deadlock
#50Earlier quoted context omitted.
I didn't reply to the comments talking about the AI tells. I replied to the comment that is making a bad argument. It doesn't matter to me whether the article is or isn't LLM assisted.
Is the not X it's Y in large frequency not an AI tell? I think they're pointing out it has a tell. Humans use those but not that often.
> Is the not X it's Y in large frequency not an AI tell?
I doubt it. The AIs are statistical models, if they've picked up a habit of saying "not X it's Y" then that is probably the most likely thing for humans to say when they are explaining something. The whole training process is about making what the AI says statistically indistinguishable from what humans do; the only way to pick up that it is an AI is either because the model is badly fitted (which, in fairness, many are, they're still working out the ideal weights) or because it isn't grounded in reality. It isn't reasonable to say "oh this looks like AI" based on small phrases like that, AIs use the same phrases humans tend to. It is where they should be doing the best job of fitting in with us.
"This uses really common phrasing ergo it is AI" is a bad case to be trying to make.