Brett Adcock, the CEO and founder of Figure Robotics, declared after the previous year that “2024 will be the year of Embodied AI.” In the present day, a humanoid robot exists that is capable of speaking with you in addition to doing our jobs!
OpenAI’s partnership with Figure is undoubtedly the first of its type, even as we eagerly await GPT-5. Things are now gradually but steadily coming together. OpenAI models offer advanced linguistic and visual intelligence. Conversely, figure neural networks provide quick, low-level, and agile robot operations.
Embodied AI
Adcock felt that he would have dependable humanoid hardware well ahead of dependable real-world neural nets operating on robots when he founded the AI robot business Figure in 2022. But as recent developments in AI have shown, they are inextricably linked to the creation of dependable humanoid robots.
In their most recent demo video, Figure not only completes modest tasks but also communicates with you. Unlike competing robot manufacturers, the humanoid can have end-to-end neural network conversations thanks to ChatGPT integration.
“We’ll be scurrying to acquire dependable hardware, enormous training sets, and developing production procedures for billions of pieces.” Speaking of the improvements he hopes to implement at Figure, Adcock remarked, “I can’t imagine a more exciting future.”
Sam Altman, the CEO of OpenAI and one of the prominent creators of Figure, has expressed great optimism about the direction robotics may go. He stated, “Finally, for the first time that I’ve ever seen, really exciting new platforms are being built on the physical hardware side.”
Adcock’s goal for robotics is to create an entirely “embodied” robot that can perform manual labor and even have human-to-human conversations. This is very different from what Boston Dynamics’ Atlas or Tesla’s Optimus can do.
Source: X
The stylish Optimus from Tesla is capable of amazing feats like yoga. In contrast, Atlas was most recently spotted somersaulting at a factory. China’s Unitree Robotics created the humanoid H1 robot, which recently achieved a walking speed of 7.4 mph and is capable of reaching 11 mph.
Unitree’s Robotics created the humanoid H1 robot Video Link: https://youtu.be/83ShvgtyFAg
2024 is predicted to be the year of robotics due to the impressive advancements made in this field. Although humanoids are one type, general-purpose robots are also becoming more and more common. A few models that assist in completing simple and low-skill tasks were introduced at the beginning of this year by Google DeepMind-led robotics initiatives.
Innovations and Advancements in Robotics Training
The difficulty of gathering and training robot data is still present, which contributes to the slow rate of robotics advancement. On the other hand, creative approaches to learning are also developed concurrently, such as imitation learning.
A Stanford PhD student named Chen Wang is leading a project that makes it possible for wearable technology to gather robot data without the need for a real robot. Robots can be trained with 3D data using the portable hand motion capture equipment called DexCap.
Karol Hausman, a Stanford professor and former Google DeepMind researcher, recently revealed the opening of his new company, Physical Intelligence. The company’s goal is to create foundational models that are capable of controlling any robot in any situation.
Hausman stated, “We’ll concentrate on making algorithmic advancements, training very large models, and whatever else is needed to bring AI into the physical world. We’ll collect robot data at a scale never seen before.”
Proprietary systems, such as the full self-driving (FSD) function seen in Tesla vehicles, are used to train Optimus.
AGI Discussions, Once More
A comment made by user X on the demo video of Figure 1. Source: X
Moreover, with discussions about AGI emerging every two weeks, the most recent demonstration of Figure 01 has undoubtedly sparked a lot of debate over whether robots are the route to AGI. Given that the topic of robotics is “Embodied AI,” the path to AGI may involve creating a fully humanoid population.
It’s interesting to note that roboticist and businessman Paolo Pirjanian founded “Embodied, Inc.”, a firm that produced Moxie robots, an AI-powered robot designed to assist and engage with children.
