Metal
What? No… Elon wouldn’t fake something! Right? Right… ?
That’s actually so funny how it puts its hands up in the air right before falling. It really looked exasperated.
I think it realized who it was working for, had an existential crisis and quit… LOL!
If his bots were that capable he’d be doing a lot more crazy stuff.
Yaah, I think he’s pretty bughouse as is.
Strange times we live in.
Sendex, is one of the best older channels for AI on YT; the guy that literally wrote the book on creating models from scratch.
He stopped posting regular content on AI model stuff after describing how to prompt AI to turn nearly anything into an interactive personalized tutorial.
Anyways, he bought one of the cheapest humanoid robots that can actually be purchased ($60k+) and posted about it on his channel. It is a no nonsense view of the state of this technology without any commercial interests or sponsorship of any kind, not even send-free-stuff bribery.
The colossal gap in software between motor controls integration and real world spacial interactions is probably a couple of decades away, even with an ever accelerating developer productivity growth. We will likely need local very large models to make this work. Stuff like a 5090 GPU should already be at something like 96 GB of VRAM, if we lived in an honest world. We would probably need 2-3 times that for a model that is fast enough for real time spacial interactions.
Silicon is stone age technology trying to replicate the greatest and final technological age to come; when biology is a fully constrained engineering corpus, long after the age of scientific discovery has ended. Humanoid robots is asking the Romans or Han Chinese to produce somewhere between a piston ICE and jet engine. A human brain is only somewhere around 100 Hz in a few different internal clocks, but massively parallel for the cost of caloric energy orders of magnitude less than present primitive tech, and is fully integrated into an elemental recycling system that is sustainable for billions of years. All present rare minerals are measured in thousands of years or less before they are fully depleted and outer space is the only option. Once there in space, fully integrated biological systems are even more essential for tiny sustainable terrarium like ecosystems.
There is nothing more toxically capitalist than the ignorance about our primitive state of science and place on the timeline. Biology is only in its infancy. One day, biology will be as easy to manipulate and program as Python, but that is many orders of magnitude more complicated and encompassing than anything humans have accomplished so far.
It is cool to see people working on the framework of robotics within a century of Asimov coining the term, but Daneel is a product of two scientists of Aurora, not Earth, and Musk is no Susan Calvin. I believe infinite cheap or free energy is a fallacy that defies the technological superiority of evolution thus far. It neglects the creation of infinite waste, finite resources, and heat.
The thing about this con is that that’s actually a viable product so… Why fake it? Are they too stupid to see useful that is hazardous work?
No it’s not viable, Elon is a grifter. This isn’t Atlas, whose autonomy and stability has been in development for three decades.
Musk knows it’s easier to raise funds by showing off advanced puppets, because most of the work is the autonomy and stability.
Elon’s most famous line is “we’ll have it next year” - that’s where his money comes from, he’s been saying that about Tesla fully self driving for six years or more, and still hasn’t got there. He’s even said it about having people on Mars.
The man is a grifter, no idea with him is viable, as it would end his main source of income, it would end the grift.
Flase promises and under delivery is his business model. It’s how he raises money.
he’s been saying that about Tesla fully self driving for six years or more
Basically 12 years according to wiki:
Since 2013, Tesla CEO Elon Musk has repeatedly predicted that the company would achieve fully autonomous driving (SAE Level 5) within one to three years
And some, myself included, didn’t think Tesla will ever achieve it. Not without LIDAR.
Its not though. They have had robots that can mirror humans for more than a decade. They don’t use them for hazardous work for one REALLY simple reason:
They are worth more than humans.
Seriously. Anything that is so dangerous that it is prohibitively expensive to safely do with humans just gets exported to the third world where humans aren’t worth anything. Why use million dollar robots when you can can just use brown guys in sandals for a nickel an hour instead?
Take auto battery recycling. That is a dangerous job that requires a lot of PPE and a ton of environmental mitigation efforts for the factory doing it. You have tons of vent hoods and lots of training and tons of inspections and oversight. Of course, that is only if you want to recycle batteries in a country that has workers rights, which is why no one recycles batteries in those countries. Instead, those batteries all get shipped to Africa where dudes break them apart with hammers in a field and cook out the metals in open pits and dump the acid into their water supply. Sure, those dudes are going to be dead in a 3 years, but who cares? Battery companies make a ton by getting the cheap recycled materials back and selling new “recycled” batteries.
There’s a difference between a viable product and the most profitable option even under capitalism.
You don’t understand capitalism if you seriously think anything but the most profitable option is viable.
A viable product is something with a customer. This doesn’t have one.
I think there’s a difference here on viable product vs practical product.
The human form is so powerful in a labor setting not because the human form is the absolute best, it’s because a person can be autonomous with very little direction. With robots you have to meticulously program them for every single movement and timing, and coordinating the dozens of joints a humanoid robot would have just isn’t worth the practical effort. Far cheaper, easier, and faster to build a robot with the exact number of joints you need for the job at hand.
