Limitations of ChatGPT | Stephen Wolfram and Lex Fridman

TL;DR

Large language models have limitations in deep computation, but excel at tasks humans can do off the top of their head. Humans will still have a role in making choices and defining objectives for AI systems.

Transcript

what do you think are the limitations of uh large language models just to make it explicit well I mean I think that deep computation is not what large language models do I mean that's just it's a different kind of thing you know the outer loop of a large language model if if you're trying to do many steps in a computation the only way you get to do... Read More

Key Insights

• 🤕 Large language models struggle with deep computation and are better suited for tasks that humans can do quickly and off the top of their head.
• 💄 Humans still have an important role in defining objectives and making choices for AI systems.
• 🧑‍🏫 AI tutoring systems can benefit from large language models combined with computational language to help teach humans more effectively.
• 💡 The ability to quickly generate ideas and possibilities using large language models can be concerning, as it highlights the potential for harmful or damaging ideas.
• 🏑 AI systems can potentially automate tasks that were previously specialized areas, reducing the need for extensive specialization in certain fields.
• 🥡 The collective intelligence of the species could trend towards being more generalist and philosophical, with AI systems taking over more specialized tasks.
• 🤨 The influence of large language models on human decision-making raises concerns about the control AI systems may eventually have over society.

Questions & Answers

Q: What are the limitations of large language models?

Large language models struggle with deep computation and lack the ability to efficiently perform tasks that require multiple steps. They rely on spooling out words, which can be an inefficient way to perform complex computations.

Q: Can large language models automate tasks humans do well, but faster?

Yes, large language models can automate tasks that humans do well off the top of their head. However, they may not always get them right, as they think through problems similarly to how humans do.

Q: How can large language models generate arbitrary code bases?

Large language models can generate arbitrary code bases by spooling out words to represent computations. However, this approach is not efficient and may not produce desired outcomes. It is more effective to provide clear and formal instructions in a computational language.

Q: What role do humans play in defining objectives for AI systems?

Humans play a crucial role in defining objectives for AI systems. While AI can achieve specific objectives, the determination of those objectives comes from human input. Humans have the ability to make choices and select which possibilities and inventions to pursue.

Summary & Key Takeaways

• Large language models excel at tasks that humans can do quickly and off the top of their head.

• Deep computation is not the strength of large language models, as they rely on spooling out words as a means of computation.

• Humans still play a crucial role in defining objectives and making choices for AI systems to follow.