It seems that Musk has a one more reason to hate OpenAI.
Figure AI startup, which announced cooperation with OpenAI two weeks ago, has already demonstrated its first results. In the new video, Figure 01 is already fully interacting with humans — it perceives requests by sight and hearing, and provides full answers.
«I see an apple on a plate in the center of the table, a drying rack with cups and a plate, and you standing next to the hand on the table», — This is how Figure 01 describes what he is shown.
Next, the human asks the robot for something to eat, and Figure 01 holds out an apple. Yes, there are some delays in processing instructions, but the robot is generally following the instructions perfectly.
«I gave you an apple because it was the only, uh, edible item I could give you from the table», — says Figure 01, and this human «um» adds even more realism to the conversation.
For the last trick in Figure 01, you ask what to do with the dishes. The robot replies that it needs to be moved to the dryer, and when asked to do so, it turns the glass upside down and places it in the right place. Figure 01 does the same with the plate, very accurately placing it on the ledge in the dryer.
Vision and speech are handled by GPT-4Vision and ChatGPT technologies, respectively, while hearing is provided by WhisperJAX. Figure AI created the robot’s skeleton and the models that set it in motion.
It seems that such progress has long been expected from Tesla with its Optimus. The company is quite quickly launched prototypes — until now, we have seen one assembled model that barely moved, and another that stood on a stand and waved to the audience. At the same time, the promo that followed the presentation featured walk more confidently and interact with the environment.
However, the last demonstration that Elon Musk posted on X, «illuminated» the engineer who controlled the robot’s movements.
Figure 01 doesn’t use any gimmicks in the video, says co-founder Brett Adcock.
«The video demonstrates end-to-end neural networks. No teleoperation», — wrote Adcock on X. «In addition, it was taken at a rate of 1.0x and continuously».
- Figure AI — is a Sunnyvale, California-based company with 50 employees that develops autonomous general-purpose humanoid robots driven by AI. According to the company, its «creations» are designed to address labor shortages, perform undesirable or dangerous tasks, and support the supply chain on a global scale.
- Recently, Figure AI raised $675 million in another round of funding — including Amazon founder Jeff Bezos, Nvidia, OpenAI, and Microsoft. Currently, the company is estimated at $2.6 billion.
- In August, the startup announced a commercial agreement under which its first humanoid robot, Figure 01 will operate at BMW’s only plant in the United States.