Despite reports from corporations I get cold feet once I embrace generative AI Due to cost and accuracy issues, it is obvious that the AI era is just starting on this planet of robotics.
Today, figurea $2.6 billion robotics startup founded lower than two years ago by former employees of Boston Dynamics, Tesla, Google DeepMind and Archer Aviation. presented his first collaboration with latest investor and partner OpenAI, maker of ChatGPT, and it's undeniably impressive.
Figure co-founder and CEO Brett Adcock used his account on the social platform X to release a video demo of a full-size humanoid Figure robot, Figure 01 (pronounced “Figure One”), demonstrating its abilities to interact with a close-by human and its surroundings, and showing the robot following the person's commands , she locates and passes an object (on this case an apple), describes what it does and converses with the person (albeit with a rather delayed response time than we might expect in a typical person-to-person conversation), and discover, plan, and perform helpful tasks alone (on this case, picking up trash and putting dishes in a drying rack).
In a scene straight out of a sci-fi movie, the video begins with the human saying, “Hey Figure One, what are you ?” The robot replies, “I see a red apple on the plate in the course of the table , a drying rack with cups and a plate, and also you stand next to it along with your hand on the table.”
“Great, can I eat something?” the person asks.
“Sure,” says Figure One, fastidiously reaching for the apple, grasping it, and handing it to the human – understanding that the apple is the one edible object in front of him, without the human even specifying.
The video also shows Figure picking up the trash and putting the plate and cup within the drying rack.
A brand new model is emerging? OpenAI VLM
Adcock posted in a thread on X that “Figure's onboard cameras are fed right into a large vision language model (VLM) trained by OpenAI,” even though it is unclear whether this can be a version of GPT-4, OpenAI's flagship LLM, which is the subscription version of ChatGPT (Plus) supports, reminiscent of GPT-4V, whether it is a fine-tuned version of such an existing model or whether it is a totally latest model. We've reached out to OpenAI for more details on the collaboration and this demo and can update as soon as we hear back.
In a powerful way ExplanationAdcock also noted, “The video shows end-to-end neural networks. There is not any teleop. Additionally, this was filmed at 1.0x speed and shot repeatedly.” In other words, the video was not sped up, as previous demo videos of humanoid robots often have been, to point out smoother movements, and there was also behind the scenes no human to remotely control the robot's movements.
Where figure goes from here
Figure's demo video appears to be a big advance in humanoid, general-purpose robotics interaction – it shows a robot interacting fairly naturally with an individual, obeying them, intuitively recognizing what they need, and doing so far more easily than many previous examples other corporations and company researchers.
However, it’s in fact still only a demo, and a prototype at that. It will likely require significantly more work to organize such a robot for industrial use and sell it to corporations and/or individuals. Still, Adcock, including his today's X-thread, has openly stated: “Our goal is to coach a world model to operate humanoid robots on the billion unit level.”
And Figure's website features Adcock's first-person perspective:Master plan“Figure's goal is to develop general-purpose humanoids which have a positive impact on humanity and create higher lives for future generations.” These robots can eliminate the necessity for unsafe and unwanted jobs – and ultimately give us happier and more purposeful lives make possible.”
But Adcock continues: “Our company journey will take a long time – and would require a masterful team dedicated to the mission, investing billions of dollars and technical innovation to attain mass-market impact.” We are high risk and faced with extremely low possibilities of success.”
He also vows: “We won’t use humanoids in military or defense applications, nor in roles that involve harming humans.”
The advances that Adock and Figure are showing today based on OpenAI are more likely to challenge competitors in the sector of humanoid robotics reminiscent of: B. put significantly more pressure on Tesla with its Optimus project And agility, a humanoid robotics startup working with Amazon on success roles. This also comes as an increasing number of corporations enter the space, including Hugging Face (which just hired a former Tesla Optimus scientist to steer its newly announced open source robotics project) and yesterday's announcement of a Startup called Physical Intelligence.
