This Robot Only Needs a Single AI Model to Master Humanlike Movements
While there is a lot of work to do, Tedrake all the proof that so far so far that the approach that the approach that used to work for robots. “I guess it changes everything,” he says.
Gailing progress in robotic is more challenging words late, of course, with video clips that commercial humanoids performing complex tasks, such as Loading cooling whether to take out the trash with apparently convenience. YouTube clips can be deceitful, though, and humanoids tend to be teleoperated, carefully progressed in advance, or trained to do a single-controlled conditions.
The work of the new ATLAS is a big sign that start to experience Robots to experience the sort of exobotics in robotics that eventually lead to the general language models that gave us Chatgpt in the field of generative AI. Finally, such progress could give us robots that can work in a wide range of messages with messy ambients new skill – from welding pipes to make espressos – without expansion.
“It is definitely a step forward,” says Ken Goldberg, a robot work at UC Berkeley receiving some funding from tri, but was not involved in the Atlas work. “The coordination of legs and arms is a big deal.”
However, Goldberg says that the idea of rising robot behavior should be treated. Just as the surprising abilities of great language models can be redirected to examples in their training data, he says that robots may demonstrate more novel when they are real. He adds it is useful to know details of how often a robot succeeds and in what ways it fails to expire at experiments. Tri has been a transparent before the work it is done on LBMS and can release more data on the new model.
Whether the one simple scales that the data used to be used to robot models will be Ever-more emergent behavior remains an open question. At a debate Hold in May with the international conference of robotic and automatically in Atlanta, Goldberg and others warned that inggel methods will also play an important role.
Tedrake, for someone, is convinced that robotics nearest a branch point is someone who will enable more solution of the humanoids and other robots. “I think we need to put this robots out of the world and start doing truly work,” he says.
What do you think about Atlas's new skills? And do you think we are on my way to a chatgpt style traffic into robotics? Let me know your thoughts about Ailab@wired.com.
This is an edition of Will Knight's Ai Lab newsletter. Read previous newsletters over here.
