Humanoid robotics has witnessed sluggish development for nearly two decades, but a groundbreaking collaboration between Figure AI and OpenAI is giving it a much-needed boost. The result? The most awe-inspiring humanoid robot video to date – a game-changer in the field.
In a recent video update, Figure AI showcased their Figure 01 robot equipped with a new Visual Language Model (VLM). The introduction of this technology has completely transformed the robot, elevating it from a mundane machine to a full-fledged, futuristic marvel with capabilities resembling those of the iconic C-3PO.
In the video, Figure 01 stands confidently behind a table adorned with a plate, an apple, and a cup. Positioned to the left is a drainer. To demonstrate its newfound capabilities, a human counterpart faces the robot and curiously asks, “Figure 01, what do you see right now?”
Within seconds, Figure 01 responds with a voice that remarkably resembles a human, despite lacking a physical face. An animated light synchronized with its voice reveals a detailed description of the items on the table and even discerns the features of the person in front of it.
This alone is impressive, but there’s more.
The human then queries, “Hey, can I have something to eat?” Much to everyone’s amazement, Figure 01 promptly replies, “Sure thing,” and with flawless precision, picks up the apple and hands it to the individual. The fluidity and dexterity of its movement leave onlookers, like myself, in awe.
But the true revelation comes when the human deliberately scatters some crumpled debris in front of Figure 01. The human then questions the robot, “Can you explain why you did what you just did while you pick up this trash?”
Without hesitation, Figure 01 offers an explanation while placing the paper back into the bin. It states, “So, I gave you the apple because it’s the only edible item I could provide you with from the table.”
My initial skepticism was vanquished by this demonstration. This was more than just a cleverly orchestrated act. This was an advancement that defied expectations.
Speech-to-speech reasoning lies at the heart of Figure 01’s capabilities. The robot utilizes OpenAI’s state-of-the-art multimodal model, VLM, which comprehends both images and texts. The system engages in an entire voice conversation to craft its responses, distinguishing it from OpenAI’s GPT-4, which focuses on written prompts.
Moreover, Figure AI has pioneered “learned low-level bimanual manipulation” in their creation. By combining precise image calibrations in real-time with their neural network, the robot achieves remarkable control over its movements – down to a pixel level. At a rate of 10hz, onboard images are processed, resulting in 200hz generation of 24-DOF actions, covering wrist poses and finger joint angles.
Figure AI is adamant that every action depicted in the video is the result of system learning, refuting any claims of teleoperation or puppeteering. While it remains challenging to fully verify these assertions without personal interaction and independent inquiries, the implications are undeniably profound.
Could this be the hundredth seamless execution of Figure 01’s routine, accounting for its fluency? Or are we truly witnessing an unprecedented feat? Whether it be a testament to tireless practice or an extraordinary leap forward in humanoid robotics, the only suitable response is one of astonishment.
This remarkable achievement not only foreshadows a future where humanoids can comprehend their surroundings, communicate, and respond like never before but also prompts us to ponder the endless possibilities lying ahead.
Frequently Asked Questions
1. How does Figure 01’s Visual Language Model (VLM) enhance its capabilities?
Figure 01’s VLM enables it to understand both images and text, allowing for speech-to-speech reasoning and comprehensive responses.
2. Is Figure 01 teleoperated or pre-programmed?
No, Figure 01’s actions are based on system learning and are not controlled by teleoperation.
3. What is “learned low-level bimanual manipulation”?
It is a technique developed by Figure AI that combines precise image calibrations with neural networks to achieve precise control over the robot’s movements.
4. Does the video accurately represent Figure 01’s abilities?
While it is difficult to ascertain the video’s authenticity without personal interaction, the demonstrations showcased are captivating and raise exciting possibilities for humanoid robotics.
Sources:
– Figure AI: [URL]
– OpenAI: [URL]
1. How does Figure 01’s Visual Language Model (VLM) enhance its capabilities?
Figure 01’s VLM enables it to understand both images and text, allowing for speech-to-speech reasoning and comprehensive responses.
2. Is Figure 01 teleoperated or pre-programmed?
No, Figure 01’s actions are based on system learning and are not controlled by teleoperation.
3. What is “learned low-level bimanual manipulation”?
“Learned low-level bimanual manipulation” is a technique developed by Figure AI that combines precise image calibrations with neural networks to achieve precise control over the robot’s movements.
4. Does the video accurately represent Figure 01’s abilities?
While it is difficult to ascertain the video’s authenticity without personal interaction, the demonstrations showcased are captivating and raise exciting possibilities for humanoid robotics.
Sources:
– Figure AI: Figure AI
– OpenAI: OpenAI
The source of the article is from the blog bitperfect.pe