Jim Keller: Most People Don't Think Simple Enough | AI Podcast Clips | Summary and Q&A

February 10, 2020
Lex Fridman
YouTube video player
Jim Keller: Most People Don't Think Simple Enough | AI Podcast Clips


Understanding a concept deeply is more valuable than following a set of recipes, especially when building or designing.

Install to Summarize YouTube Videos and Get Transcripts

Key Insights

  • 🍞 Building something special requires simple thinking and understanding, with a balance of deeper understanding and executing recipes.
  • 🥪 Having a deep understanding allows for a wider scope of knowledge and enables experts to view things differently.
  • 🍳 Unpacking understanding can prevent mistakes, but constantly seeking deeper understanding can hinder progress.
  • 💻 Building a computer involves constantly optimizing and rewriting it, with a tendency to rewrite every 5 years instead of 10.
  • ⚙️ Incrementally improving existing projects can lead to diminishing returns, requiring a new project to reach the next level.
  • 📈 Short-term limitations can block long-term success, and leaders who achieve long-term success are willing to redo things when necessary.
  • 📚 Marketing often wants promises of improved performance, but computer architects know there will always be outliers that are slower.
  • ️ Balancing the desires of short-term objectives with long-term objectives is challenging and requires strategic decision-making.


so what have you learned about the human abstractions from individual functional human units to the broader organization what does it take to create something special well most people don't think simple enough all right so do you know the difference between a recipe and understanding there's probably a philosophical description of this so imagine y... Read More

Questions & Answers

Q: Why is understanding a concept deeply important in building and design?

Understanding a concept deeply allows for a more holistic and comprehensive approach to problem-solving. It enables individuals to recognize patterns, connections, and potential solutions that may not be apparent when following a set of recipes.

Q: How does executing recipes differ from achieving deeper understanding?

Executing recipes is efficient when it aligns with an individual's goals, but it can be limiting when faced with complex problems that require a broader perspective. Deeper understanding allows for a more flexible and adaptive approach to problem-solving.

Q: How do individuals and teams balance deeper understanding and following recipes?

Finding the right balance between deeper understanding and following recipes depends on the objectives and context. Some individuals or teams may prioritize one over the other, while others may develop a combination approach depending on the specific situation.

Q: Why is it important to reassess and potentially rewrite computer architectures?

Rewriting computer architectures periodically is crucial to make significant progress. Incremental improvements reach a point of diminishing returns, and a fresh perspective can lead to more efficient and effective designs.

Q: How do short-term and long-term goals impact decisions to rewrite or optimize computer architectures?

Short-term goals, such as quarterly business objectives, often prioritize incremental optimization. However, leaders who prioritize long-term success understand that short-term limitations can hinder overall progress. They are willing to invest in rewriting and redesigning to achieve better long-term outcomes.

Q: How do marketing expectations and technical limitations clash when building new computers?

Marketing teams often expect that every new computer will be faster in every aspect compared to its predecessor. However, computer architects understand that performance results have a distribution, and some outliers may be slower. Balancing these expectations is challenging, as individual customers may prioritize specific aspects of performance over others.


In this video, the speaker discusses the importance of deep understanding and the limitations of following recipes or executing predefined steps. He emphasizes the need for individuals to achieve deeper understanding in their respective fields rather than just following a set of instructions. The speaker also talks about the challenges faced when balancing deep understanding with practical execution and how this impacts the development process. Additionally, he touches upon the importance of purpose and fundamental questions in building and designing things.

Questions & Answers

Q: What is the difference between a recipe and understanding?

A recipe provides a specific set of steps to follow in order to achieve a desired outcome, while understanding allows for a deeper comprehension and knowledge about the subject matter. Recipes are limited in scope and do not provide the same level of insight as understanding does. For example, when making bread, a recipe will only provide the steps to make bread, but deep understanding includes knowledge of biology, supply chains, physics, and more.

Q: Why is deep understanding important in building and designing things?

Deep understanding allows individuals to have a broader view of the subject area and consider various factors and elements that may affect the outcome. When building and designing, following a stack of recipes will only provide limited results, while deep understanding enables individuals to approach problems from different perspectives and come up with innovative solutions. It helps in recognizing when deeper understanding is required and when execution of predefined steps is sufficient.

Q: Does deep understanding apply to every stage of development?

Yes, deep understanding is relevant at every stage of development. It helps individuals make informed decisions, intervene when necessary, and avoid doing the wrong thing. At every juncture, there are numerous possible outcomes, and deep understanding helps in navigating through them. However, it is important to strike a balance between deep understanding and practical execution, as constantly unpacking everything for deeper understanding can impede progress.

Q: Is achieving deep understanding an unsolvable task?

Yes, it can be considered an intractable problem due to the complexity and unpredictability of human beings. Every piece of information given to individuals can lead to numerous possible outputs, and these outputs interact with each other in unique and sometimes hilarious ways. Determining what to tell individuals, when to intervene, and when to let them figure things out themselves is a complicated task that cannot be computationally solved.

Q: What areas does deep understanding delve into?

Deep understanding can cover various fields such as fundamental questions, purpose, biology, supply chains, physics, and more. It involves going beyond surface-level knowledge and gaining insights into the core aspects of a subject. For example, in building a computer, deep understanding would require considering fundamental limits, rethinking and rewriting the implementation, and continuously seeking improvement.

Q: How often is it necessary to throw out everything and start from scratch?

The need to start from scratch arises more often than people might think. In computer architecture, it is recommended to rewrite the entire system every three to five years in order to make significant progress. While it may seem scary and counterproductive, the process of rewriting often results in a faster implementation and a simpler design. By continually refactoring and rewriting, progress can be made beyond the limits of incremental improvements.

Q: What is the typical frequency of rewriting computer architectures?

The speaker mentions that there is a tendency to rewrite computer architectures approximately every ten years. However, he argues that it would be more beneficial to rewrite them every five years, as this allows for greater progress and innovation. The x86 64 architecture, for example, was co-authored by the speaker 20 years ago and has been extended multiple times since then, but changes in the implementation, rather than the instruction set, have been less frequent.

Q: Are continuous rewrites and frequent changes in computer architecture considered unconventional?

Yes, continuous rewrites and frequent changes in computer architecture can be considered unconventional, especially for companies focused on short-term objectives and immediate profitability. Companies and individuals driven by quarterly business objectives might approach continuous rewrites with fear and resistance, as they disrupt the status quo and require significant resources. However, for long-term success and innovation, it is essential to overcome short-term limitations and embrace change.

Q: How do short-term and long-term objectives impact the decision to rewrite a system?

Short-term objectives, such as quarterly business goals, often prioritize incremental improvements rather than rewriting systems from scratch. This approach aims for short-term success and immediate profitability. On the other hand, long-term objectives recognize that short-term limitations can hinder long-term success. Leaders of successful companies often make the decision to rewrite systems when necessary, even if it goes against short-term objectives. This requires balancing the perspectives of different stakeholders and considering both short-term and long-term consequences.

Q: What challenges arise when deciding to rewrite systems and embrace change?

The decision to rewrite systems and embrace change can be challenging due to the conflicting interests of different stakeholders. For example, marketing teams often demand that the new computer is faster on every single task, which is not always feasible. Computer architects understand that while the new computer may be faster on average, there will be outliers where performance may suffer. Balancing these conflicting interests and managing customer expectations can be a complex task.


In order to create something special, it is important to think simple and achieve deep understanding rather than merely following predefined steps. Deep understanding allows individuals to approach problems from multiple perspectives and find innovative solutions. Striking a balance between deep understanding and practical execution is crucial at every stage of development. While deep understanding may not have a computationally-solvable solution, it is essential for making informed decisions and achieving long-term success. Continuous rewrites and frequent changes may be necessary in order to overcome short-term limitations and drive innovation, even if they may cause temporary disruptions. Leaders must consider both short-term and long-term objectives when making these decisions and manage conflicting interests of stakeholders.

Summary & Key Takeaways

  • Building and designing things often involves executing a stack of recipes, but this approach has limited scope.

  • Achieving a deeper understanding of the subject matter allows for a more comprehensive view and better problem-solving.

  • Balancing deeper understanding and following recipes is crucial, and it varies for individuals and teams.

Share This Summary 📚

Summarize YouTube Videos and Get Video Transcripts with 1-Click

Download browser extensions on:

Explore More Summaries from Lex Fridman 📚

Summarize YouTube Videos and Get Video Transcripts with 1-Click

Download browser extensions on: