News The GOP wants to deregulate AI — provision in Budget Reconciliation bill blocks state governments from meddling for 10 years

[CelicaGT]

>Manufacturing is best done with purpose built robots, designed and configured to do that specific task. There is little to no demand for a general purpose humanoid robot in this role.

It depends on the task(s), but generally speaking, general-purpose robots can be reprogrammed/retrained to handle a wide range of tasks, and can be more cost efficient than having multiple purpose-built robots for multiple tasks.

A robot imbued with AI & sensors can also learn on its own how to best do a task, assuming the task is ill-defined, which allows flexibility and adaptability. Purpose-built robots by definition don't have that trait.

A humanoid robot in particular can be more effective by virtue of its form, since the world is designed for humans. For example, it can conceivably make use of existing human tools, as opposed to needing proprietary tools made for it. This allows versatility.

In manufacturing, tasks are very specific and the motions and tools to complete the task are integrated into the design making this type of automation far, far more efficient than any human, or humanoid robot could ever be. Most of the "Arms"we use have 360 degrees of rotation in each joint, and full axis movement around the workpiece. That's just not something that can be done with a humanoid form. This is well a well studied and solved problem.

 

[USAFRet]

And for the other side of the magical AI....

https://gizmodo.com/its-breathtaking-how-fast-ai-is-screwing-up-the-education-system-2000603100

 

[Eximo]

“Technical maturity of humanoid robots is still uneven,” said Chen. “Lower limb technology is mature (e.g., walking, dancing), while upper limb technology lags due to challenges in manipulation and generalization. Software is still scenario-specific: most robots are trained for specific tasks and lack generalized adaptability.”

"upper limb technology lags due to challenges in manipulation and generalization."

"most robots are trained for specific tasks and lack generalized adaptability.”

That right there is the main takeaway. You are talking about a generalized robot, that is what they say isn't ready.

A humanoid robot with high precision is going to be expensive, likely several times the cost of traditional automated manufacturing. Stability, and all that too. Arm and hand precision is great, but while balancing and lifting, etc...better to have a fixed arm with known precision and accuracy.

Mass production doesn't mean mass adoption. These are early models and they are probably going to end up as show pieces to generate additional funding for more R&D. It is cool idea, just not something that is going to happen any time soon. The AI we have available now just isn't up to the task.

 

[CelicaGT]

>In manufacturing, tasks are very specific and the motions and tools to complete the task are integrated into the design making this type of automation far, far more efficient than any human, or humanoid robot could ever be.

As said in the first sentence of my response, it depends on the task.

Obviously, there are roles for both types of bots, GP and PB. But your blanket statement that "there's little to no demand for a GP humanoid robot in [manufacturing] role" is pigeonholing. I note you didn't refute any of my counterpoints re: advantages to GP robots.

>Most of the "Arms"we use have 360 degrees of rotation in each joint, and full axis movement around the workpiece.

One observation is that the above bots aren't just designed for the task, but the task flow was also designed with the bots functionality in mind. Basically, you are saying that the square peg is no good since it doesn't fit into your round hole.

I'll defer to your assessment on your specific manufacturing process. But I doubt you can generalize that to ALL manufacturing processes as per your statement.

BTW, "general-purpose" doesn't necessarily equate to "humanoid."


>"upper limb technology lags due to challenges in manipulation and generalization."
>"most robots are trained for specific tasks and lack generalized adaptability.”

>That right there is the main takeaway. You are talking about a generalized robot, that is what they say isn't ready.

You are missing the forest for one stunted tree. You take one problem to be solved and conflate it to declare the whole venture as unworkable. Did you even read the article? How about just the title, "Humanoid Robots Ready for the Masses: Report"?

I don't need to refute any points specifically because I refute the entire premise that there is any efficient or cost effective use for a humanoid robot in the manufacturing space. There is no task in manufacturing that cannot be done better by a task specific machine. The humanoid form is good at a great many things but it only has one elbow, one wrist and no stable base to work upon. These processes require a level of flexibility and dexterity that just cannot be achieved by the humanoid form. That is why we built different, better forms. Engineering a general purpose, humanoid form machine for this type of work is a step backwards. Also keep in mind that many tasks done by humans in a factory are often because of political or economic reasons. There's lots of cases where we can automate the entire thing. Have you even seen these things in motion? I have and I can tell you it's no contest.

Now, I imagine that in the industrial scene in general there could be tasks where it's not cost effective to engineer a specific solution, tasks that are too dangerous or high risk or even just infrequently done. THESE situations I can see a strong use case for a general purpose robot.

 

[USAFRet]

>And for the other side of the magical AI....
>https://gizmodo.com/its-breathtaking-how-fast-ai-is-screwing-up-the-education-system-2000603100

AI is upending a lot of systems, not just the schools.

It's facile to blame AI, as per your "magical AI" snark. But AI is just a tool, and how good or bad the result is depends on how it's used. The upheavals we see now comes from the systems not yet adapted for AI use--and we have to adapt, because it's not going away. Pooh-pooh'ing it with snark and putdowns is just more Internet noise.

I'm not poohpooing the AI tool.

Rather, the less than clueful that see it as the all knowing giver of all information.
Yes, I know some people like this.

A coworker was taking some college classes. Specifically, biology. "P, are you going to use your fave AI tool?"
'Yes, 1000%'

"So basically, you're not learning anything of biology. All you're doing is learning how to manipulate the AI generator into giving you a semi (but not really) rational response."

'Yep.'


Now...whether he needs to know anything about biology for his future career is not relevant. I just know that he isn't learning any of it now. Just filling a square on his grad matrix.

 

[CelicaGT]

>And for the other side of the magical AI....
>https://gizmodo.com/its-breathtaking-how-fast-ai-is-screwing-up-the-education-system-2000603100

AI is upending a lot of systems, not just the schools.

It's facile to blame AI, as per your "magical AI" snark. But AI is just a tool, and how good or bad the result is depends on how it's used. The upheavals we see now comes from the systems not yet adapted for AI use--and we have to adapt, because it's not going away. Pooh-pooh'ing it with snark and putdowns is just more Internet noise.

Pretty much this. As much as I disagree with even calling it AI, it is proving to be a useful tool in many industries. I feel a lot of the pushback stems from having it shoved down our collective throats in a mass effort to get a return on the investment. Chat bots and digital assistants were already a financial flop, injecting an LLM isn't going to move that needle much.

In regards to research and fact finding, like any computer system these models are subject to "Garbage in -> garbage out". What I mean by that is the LLM doesn't know anything, so they often feed false or incorrect information to the user. I mean, just scroll through Reddit and other user forums that these tools are training on. There's some pretty low quality stuff there and it shows in the results. Also one of the reasons its use is banned in my country's education system. They do use AI in school, but not for completing work. They use it to learn how to use it, and to do so responsibly.

(Edit: Spelling, probably missed more than what I fixed)

 

[USAFRet]

I mean, just scroll through Reddit and other user forums that these tools are training on. There's some pretty low quality stuff there and it shows in the results.

We see some "replies" in here that were AI generated.

Clueless user, that does not really know the AI generated answer he spouted was wrong and misleading.

 

[CelicaGT]

We see some "replies" in here that were AI generated.

Clueless user, that does not really know the AI generated answer he spouted was wrong and misleading.

I've seen it too. Unfortunately those less experienced don't know any better and therein lies the core issue.

 

[Rogue Leader]

I could care less about flying cars. I need a Rosie, from the Jetsons.

This. Was chatting with my neighbor the other day, another techie guy like myself. I don't need AI to organize my life, I don't need self driving cars, I don't need any of this stuff they say they are developing.

Give me a robot that while I'm at work every day cleans all the bathrooms, mops the floor, wipes the counter, some stainless steel cleaner on the appliances, and then shuts up and goes and hides in the closet. That would be LIFE!