Can a Superintelligent Robot defeat Gordon Ryan in submission grappling? | John Danaher

TL;DR

Can a robot with the same physical abilities as a human defeat a skilled grappler like Gordon Ryan? The question explores the possibility of formalizing grappling as an engineering discipline.

Transcript

let me ask you a completely ridiculous question uh but it's a fascinating one for me from an engineering and a scientific perspective when i look at a sport really any problem one way to ask how difficult is this problem is to see how can i build a machine that competes with a human being at that problem you can look at chess you can look at soccer... Read More

Key Insights

• #️⃣ The number of options in complex games like chess and grappling is astronomically high, making it impossible for any computer to calculate all potential moves. Heuristics help reduce the number of options and simplify decision-making.
• 🤯 Machine learning has shown rapid progress in AI, surpassing human capabilities in chess and go. This suggests that a mind inside a robot that can defeat a human opponent in grappling may be feasible in the future.
• 😤 The combination of human insight and computational power can be a powerful strategy, as seen in cyborg chess teams. This partnership has proven to be superior to both humans and computers alone.
• 👏 The use of AI in military applications raises concerns about unintended consequences and the need for proper supervision and oversight to prevent catastrophic events caused by software bugs.
• 🤑 The development of AI systems in grappling and physical spaces faces challenges in replicating the rich sensory experience of humans and understanding complex movements and interactions.

Questions & Answers

Q: Can a robot defeat a skilled grappler like Gordon Ryan?

In theory, if the robot has the same physical capabilities as a human but superior computational powers, it could potentially defeat a human opponent. However, the challenge lies in developing a mind inside the robot that can match the grappling expertise of someone like Gordon Ryan.

Q: What are the approaches to building AI systems?

One approach is rule-based, where a set of rules is created for different conditions or patterns. Another approach is machine learning, where a system learns from observations and adjusts its expectations based on outcomes. Machine learning has shown promise in achieving success in complex tasks like chess.

Q: How do heuristics play a role in AI systems?

Heuristics, or rule of thumb, can significantly reduce the number of options and simplify decision-making for AI systems. They provide general guidelines for action and are especially useful in complex domains like grappling.

Q: Are cyborg chess teams still competitive with machine learning systems?

Machine learning has surpassed human-machine pairs in chess competitions. Although cyborg chess teams initially showed promise, machine learning has become dominant. Humans now focus on competing with other humans and relying on their own expertise and computational powers.

Summary & Key Takeaways

• The question raises the possibility of building an AI system or robot that can overpower a skilled grappler like Gordon Ryan.

• Two approaches to building AI systems are discussed: rule-based systems and machine learning. Both have their challenges and limitations.

• The use of heuristics, or rule of thumb, can significantly reduce the number of options and make AI systems more competitive.

• Cyborg chess teams, where humans and computers work together, have shown superiority over both humans and computers alone.

• The increasing capabilities of AI, especially in machine learning, may eventually lead to the creation of a mind inside a robot that can defeat a human opponent.