Figure AI CEO on Solving General-Purpose AI for Humanoid Robots

The Core Challenge: General-Purpose AI

• 💡 Figure AI CEO Brett Adcock identifies general-purpose AI as the primary challenge for humanoid robots, stating it's 10 to 100 times harder than manufacturing the robots themselves.
• 🧠 The goal is to build a horizontal AI stack that can perform any human task in unseen locations, allowing users to simply talk to the robot for instructions.
• 🎯 This problem is currently unsolved and requires significant "hill climbing" to achieve true autonomy.

Figure AI's Strategic Approach

• 🚀 Figure AI is tackling this by refactoring its autonomy and AI stack to use end-to-end deep learning and neural nets from scratch, believing it's the only way to scale.
• 🛠️ They are also developing high-scale manufacturing capabilities in parallel at their Bot Q facility in California, handling final assembly and testing.
• ✅ The company aims for full autonomy with low human intervention rates in commercial markets before launching products at scale.

Real-World Deployment Advantage

• 📈 Figure AI currently has robots working in commercial environments for 10-hour shifts, with one running autonomously for almost six months.
• 📊 This real-world deployment provides valuable lessons learned and insights, helping to build the right technology stack and refine performance.
• ⚡ The data generated from these deployments creates an exponential learning curve, giving Figure a significant advantage over competitors still in lab settings.

Future Vision and Scalability

• 🔮 Figure AI anticipates seeing their robots in the commercial workforce within the next one to two years, with efforts to reduce costs and increase reliability.
• 🏡 They are also working on deploying robots in homes, with early tests involving tasks like laundry and dishes, aiming for a language-conditioned, pixel-space understanding.
• ⚠️ The company emphasizes that products will only be launched at scale in homes or the workforce once they are confident in safety and functionality, avoiding premature or tele-operated deployments.