Figure's Humanoid Robots: Revolutionizing the Workforce and Home

The Rise of Figure's Humanoid Robots

In just 31 months, Figure has gone from a cold start to shipping its first humanoid robot - an extraordinary feat in the challenging field of robotics. Founded by Brett Adcock, who previously had success with companies like Archer Aviation, Figure is taking on one of the most difficult engineering challenges: creating capable humanoid robots powered by artificial intelligence.

Why Humanoid Robots?

Adcock believes humanoid robots are the ultimate deployment vector for artificial general intelligence (AGI). As he explains:

"We really need to figure out a way to give AGI a body. I think it's a negative, almost dystopian future if we figure out how to solve AGI and it lives in a server somewhere, more intelligent than all of humanity, but has to ask humans to do anything in the physical world."

A humanoid form factor allows for a single robotic platform that can theoretically do anything a human can do, without hardware changes. It also provides an ideal embodiment for neural networks to learn from and transfer skills across a variety of tasks.

Rapid Hardware Iteration

One of Figure's key advantages is the speed at which they iterate on their robot designs. Adcock notes:

"We are designing a new hardware platform every 12 to 18 months. By the time I filed the C Corp, we had the robot walking in under 12 months."

This rapid iteration cycle allows Figure to continually improve their robots' capabilities. They recognize that the first iterations of any new hardware will have flaws, so they push to evolve the design as quickly as possible.

Vertical Integration

Figure has taken a fully vertically integrated approach out of necessity. As Adcock explains, "There's no supply chain for humanoid robots." The company designs and manufactures nearly every component in-house, including:

• Motors and actuators
• Sensors
• Battery systems
• Structural components
• Kinematics
• Firmware and embedded systems
• Operating systems
• Middleware
• Control systems
• AI and machine learning software

This vertical integration gives Figure full control over the entire robot system, allowing for tighter integration and faster development cycles.

Commercial Deployments

Figure has already begun deploying their robots in real-world industrial settings. Their first major customer is BMW, where Figure robots are working in the Spartanburg, South Carolina plant every day. The robots assist with tasks like placing sheet metal on fixtures - a common job in manufacturing that requires precision and consistency.

Adcock notes that the Figure robots are performing these tasks "fully autonomously at the speeds we need to basically hit high performance, with no human intervention, no faults, no failures." This demonstrates the potential for humanoid robots to take on repetitive industrial tasks reliably.

Figure has also signed a second major customer in the logistics industry, deploying robots within 30 days of beginning work with the company. This rapid deployment showcases the flexibility of Figure's robotic platform.

The Path to Consumer Robots

While industrial applications are the near-term focus, Figure has its sights set on eventually bringing humanoid robots into the home. Adcock believes consumer robots could be available for around $20,000 to $30,000 - comparable to leasing a car.

At that price point, Adcock envisions multiple robots per household, taking on mundane chores:

"I wake up every morning and help unload the dishwasher and pick up kids' toys. I never want to do any of that ever again. We really haven't had a lot of innovation in the home for almost 50-70 years. We have the same appliances, same stuff."

He imagines a future where you can simply talk to a robot, have it understand context and your preferences, and take care of household tasks autonomously.

Building the Team

A critical factor in Figure's rapid progress has been assembling an exceptional team of engineers and operators. Adcock brought over much of his business team from previous ventures, providing continuity and allowing him to focus primarily on product and engineering.

The company culture emphasizes hard work and a shared vision:

"The entire culture at Figure, even at Archer when I built the initial team, was very deliberate. We have the culture deck, we have the master plan, we have things laid out that are really unique. We're in Silicon Valley but almost like the anti-Silicon Valley. You have to work every day in the office, we work five to seven days a week, we work really hard."

This intense work ethic has allowed Figure to make rapid progress on an extremely challenging engineering problem. Adcock notes they've assembled "hundreds of the best engineers in AI robotics in the world."

The Three Key Challenges

When founding Figure, Adcock identified three major technical hurdles that needed to be overcome to make general-purpose humanoid robots viable:

1. Hardware: Build incredibly complex, reliable hardware that can match human speed and range of motion. Most existing bipedal robots struggle with basic locomotion.

2. Neural control: Develop neural networks that can control the full humanoid system, rather than relying on traditional robotic control algorithms. This allows for more flexible, generalizable behavior.

3. Generalization: Create AI systems that can understand novel situations through natural language, then execute appropriate actions. This is key for robots to be useful in unstructured environments like homes.

Adcock believes Figure has made substantial progress on all three of these challenges in a remarkably short time frame.

Helix: Figure's AI Brain

A major breakthrough came with the development of Helix, Figure's large-scale AI system for controlling their robots. Helix is a vision-language-action model that allows the robots to understand and act on natural language commands in novel environments.

A demonstration video shows two Figure robots working together to put away groceries based on a simple verbal instruction. Critically, the robots had never seen these specific grocery items before, showcasing their ability to generalize.

Adcock describes this as "probably the most important AI update for robotics in human history." He believes AI agents like Helix will power all robotic systems in the future.

The Helix system emerged from relatively little training data - just 500 hours. This efficiency suggests rapid improvements may be possible as more data is collected.

Interestingly, the robots developed behaviors like making eye contact during handoffs without being explicitly programmed to do so. This emerged naturally from the training process as an efficient way to coordinate actions.

Figure 3 and the Path Forward

While Figure 2 robots are already deployed commercially, the company has completed design work on Figure 3 - described by Adcock as a major leap forward:

"Figure 3 is just like 90% cheaper, it's smaller, it's less mass, it's got better sensors, its hands, head and feet were designed for neural nets. It's a completely new design."

He considers it the most significant engineering achievement of his career so far. Figure 3 is intended to be the model that scales to widespread deployment across both industrial and eventually consumer applications.

The company is pursuing parallel tracks for workforce and home robots. Workforce applications are easier to deploy and monetize in the near term. Major companies have already expressed demand for over 100,000 robots if they were available immediately.

However, the home environment presents a far greater challenge due to its unstructured nature and safety considerations. Adcock likens it to the difference between highway and city driving for autonomous vehicles.

Nonetheless, Figure's rapid progress with systems like Helix has accelerated their timeline for home robots. Adcock now believes they may begin alpha testing in homes (starting with employees) as soon as this year.

He sees the primary bottleneck as collecting enough training data to handle the diversity of home environments and tasks. But with the core AI capabilities demonstrated by Helix, it may just be a matter of scale:

"We just feel data-bound in the home. We think if we just increase the dataset that we trained Helix with by a couple orders of magnitude, it would probably [be capable enough]."

The Implications of Capable Humanoid Robots

If Figure succeeds in creating generally capable humanoid robots at scale, the implications for society could be profound:

• Workforce transformation: Robots could take on a huge variety of manual labor jobs, potentially displacing human workers but also filling crucial gaps in industries facing labor shortages.

• Productivity gains: 24/7 robot labor without fatigue or errors could dramatically increase economic output in many sectors.

• Domestic assistance: Affordable home robots could free up significant time for families by handling household chores and maintenance tasks.

• Eldercare: Robots could help address the growing need for in-home care for aging populations.

• New interaction paradigms: As robots become commonplace, humans will need to develop new social norms and ways of interfacing with intelligent machines in daily life.

• Ethical and safety considerations: Widespread deployment of powerful robots will require careful governance to ensure they are used responsibly and safely.

Conclusion

Figure's rapid progress in humanoid robotics represents a potential inflection point in the field. By combining advanced hardware engineering with cutting-edge AI systems like Helix, they are pushing the boundaries of what robots can do in both industrial and domestic settings.

While significant challenges remain, particularly for home robots, the pace of improvement suggests capable humanoid robots may become a reality far sooner than many expected. As Adcock puts it, we may be witnessing an "iPhone moment" for robotics - the emergence of a transformative new technology that reshapes society.

The coming years will be critical in determining whether Figure can scale up production, continue improving their AI systems, and successfully navigate the complex regulatory and social factors involved in deploying robots widely. If they succeed, it could usher in a new era of human-robot collaboration across all facets of life and work.

Article created from: https://www.youtube.com/watch?v=hHA4-nEBer8