Boston Dynamics CEO on ChatGPT | Robert Playter and Lex Fridman

TL;DR

Transformer neural networks have come a long way and offer exciting advancements, but there are concerns regarding disinformation. However, the risks associated with these models differ from their use in robotics.

Transcript

what do you think in general about the all the exciting advancements of uh Transformer uh neural networks most beautifully uh Illustrated through the large language models like gpt4 like everybody else we're all you know I'm I'm surprised at uh how much uh how far they've come um I'm a little bit nervous about the there's anxiety around them obviou... Read More

Key Insights

• ⌛ Transformer neural networks, exemplified by models like GPT-4, have undergone remarkable advancements in recent times.
• 🌥️ Concerns exist regarding the potential misuse of large language models, such as spreading disinformation through social media.
• ✳️ The risks associated with transformer neural networks differ when applied in robotics, where physical reality provides a reliable means of verification.
• 🤖 Language models like GPT offer exciting possibilities for enhanced communication with robots, expanding the ways humans can interact with them.
• 👨‍🔬 The Boston Dynamics AI Institute, spearheaded by Mark Raibert, aims to pursue unadulterated research and tackle long-term problems in the field of artificial intelligence.
• 💦 The AI Institute and Boston Dynamics will work together, with the Institute focusing on pushing the forefront and conducting research beyond the next five years.
• 🚥 Mark Raibert's vision for the AI Institute is to have the freedom to pursue difficult over-the-horizon problems, similar to the multi-decade challenge of legged locomotion.

Questions & Answers

Q: What are the advancements and risks associated with transformer neural networks?

Transformer neural networks like GPT-4 have made significant advancements, but there are concerns regarding the potential use of these models for spreading disinformation. The risks differ when these models are used in robotics, with physical reality providing a means of verification.

Q: How do language models improve communication with robots?

Language models like GPT allow for increased communication bandwidth with robots such as Spot. By enabling verbal interactions and integrating with vision algorithms, the variety of ways to communicate with robots is greatly expanded.

Q: What is the Boston Dynamics AI Institute?

The Boston Dynamics AI Institute is a separate organization led by Mark Raibert, the founder of Boston Dynamics. Its mission is to pursue unadulterated research in the field of artificial intelligence, focusing on pushing the forefront and tackling difficult long-term problems.

Q: What distinguishes the research conducted by the AI Institute from Boston Dynamics?

While Boston Dynamics focuses on commercialization and research with a shorter time horizon, the AI Institute's goal is to work on over-the-horizon problems, tackling complex challenges beyond the next five years.

Summary & Key Takeaways

• Transformer neural networks, like GPT-4, have made impressive advancements, but there is anxiety surrounding their potential use for spreading disinformation.

• When used in robotics for manual tasks, the risks associated with these models differ. Physical reality provides a built-in mechanism for verifying if the robot is performing the task correctly.

• Language models like GPT open up possibilities for improved communication with robots, such as Spot, allowing for verbal interactions and increased bandwidth through various interfaces.