When Will a Robot Become UFC Champion | Russ Tedrake and Lex Fridman

TL;DR

Developing robots that can win in physical competitions against humans, such as wrestling or mixed martial arts, is a complex problem that we haven't solved yet.

Transcript

now you're a serious person uh i'm a little bit of an idiot and i'm going to ask you some dumb questions uh so i do uh i do martial arts uh so like jiu jitsu great i've wrestled my whole life so let me let me ask the question um you know like whenever people learn that i do any kind of ai or like i mention robots and things like that they say when ... Read More

Key Insights

• 🥋 Developing robots that can defeat humans in physical competitions like wrestling or mixed martial arts remains a challenging problem due to the complex nature of contact and game theory.
• 🛀 Advances in deep learning policies and controllers have shown progress in object manipulation but fall short in making quick decisions based on models during physical competitions.
• 🥋 The learning process in physical skills like martial arts involves both intellectual and distributed learning, which is different from deep learning in controllers.
• 🤪 The role of motivations, ego, confidence, and energy efficiency are crucial in the learning process for physical activities.
• 🤖 RoboCup, a robot soccer competition, has gained popularity and momentum, making it more likely to see a robot soccer team before a robot UFC champion.
• 👻 Creating rules for physical games that prevent exploitation of robot capabilities while still allowing fair competition is a challenging task.
• 👾 Machines have achieved victories against humans in digital games like chess and go but have not yet surpassed humans in physical games.

Questions & Answers

Q: How difficult is it to create a robot that can defeat a human in a wrestling match?

Developing a robot capable of winning a wrestling match against a human is a challenging problem, with complexities in contact, game theory, and decision-making based on models. Although there has been progress in manipulation, the dynamic nature of wrestling poses additional difficulties.

Q: What is the role of learning and expertise in physical activities like martial arts?

In physical activities like martial arts, learning involves both intellectual understanding and distributed learning through the body. Expertise brings about efficiency, awareness, and a refined model of mechanics. Motivations, ego, and energy efficiency also play essential roles in the learning process.

Q: Will we see a robot soccer team or a robot UFC champion first?

It is hard to determine which task is technically harder, but the popularity and momentum of RoboCup make it more likely to see a robot soccer team before a robot UFC champion. However, both problems are challenging and require careful rule-setting.

Q: Have robots beaten humans in physical games?

While machines have achieved victories against humans in games like chess and go, there hasn't been a case where a robot has defeated a human in a physical game. This is mainly due to the complexity of contact and the need for precise rule-setting to prevent exploitation of robot capabilities.

Summary & Key Takeaways

• Creating robots that can defeat humans in physical competitions is a difficult problem due to the complex nature of contact and game theory.

• While there have been advancements in deep learning policies and controllers for manipulating objects, making decisions based on these models quickly remains a challenge.

• The learning process and distributed learning involved in human physical skills, such as martial arts, differ from deep learning in controllers.