No Priors Ep. 86 | With Sarah Guo & Elad Gil

Detailed Summary

Introduction (0:00 - 0:47)

• Sarah and Elad welcome listeners to "No Priors," joking about its popularity within their families.
• They set the stage for a discussion on recent AI advancements and their implications.

Google Releases NotebookLM (0:47 - 5:20)

• Elad expresses fascination with Google's NotebookLM, recalling a similar journaling tool prototyped by a Stanford student.
• NotebookLM allows users to upload documents and generate AI-driven discussions, effectively turning information into an interactive podcast-like experience.
• Sarah highlights NotebookLM's utility as a pre-built RAG (Retrieval-Augmented Generation) interface for personal data.
• They note that while similar RAG indexing tools existed, NotebookLM's elegance and voice integration are significant advancements.
• The discussion emphasizes how AI can automate information discovery and integrate multimodal audio, making AI capabilities more accessible and engaging for end-users.
• They observe that many early consumer AI apps (ChatGPT, Perplexity, Midjourney) started as utilities rather than social apps, focusing on information and content generation.

Integrating AI into Consumer Apps and Gaming (5:20 - 9:11)

• Elad and Sarah discuss the potential of AI in entertainment, particularly gaming.
• They revisit the long-standing concept of intelligent NPCs (Non-Player Characters) that can interact with players in a human-like manner, potentially blurring the lines between human and AI players in games.
• Sarah mentions a demo showcasing AI integration into gaming for coaches and NPCs, with the ability to trigger agentic actions with low latency.
• Elad highlights the potential for AI to generate infinite game levels, similar to how franchises like Angry Birds iterate on core mechanics with new content.
• They foresee AI enabling users to create powerful, AI-based expansions and modifications (mods) for games, extending beyond publisher-created content.
• The conversation touches on how AI could modernize generational trends like journaling, leveraging new technology for personal expression.

Future of AI Companionship and Procreation (9:11 - 14:45)

• Elad shares a conversation with an AI researcher who believes his son might have an AI romantic partner, sparking a debate on the implications for human reproduction and societal values.
• Sarah and Elad discuss the "dinner question" of what proportion of future generations will date AI bots.
• Elad raises concerns about artificial wombs becoming mainstream, fearing a dissociation of people from human interaction and society.
• He draws a parallel to the "Clone Wars" from Star Wars, questioning the importance of human connection.
• They reflect on how the COVID-19 pandemic and lockdowns demonstrated the negative effects of human isolation.
• Sarah suggests that the premium on human-like communication will decrease due to the infinite volume of AI-generated communication, leading to a future where "Bots email Bots."

OpenAI O1 Model Improves on Iterative Reasoning (14:45 - 18:06)

• The discussion shifts to OpenAI's recently released O1 model, which focuses on longer-term planning and iterative reasoning.
• O1 can scale compute usage at test time, making it better at tasks requiring deeper thought, such as math problems, code, and crosswords.
• They acknowledge mixed reactions from the industry, with some questioning its immediate superiority over GPT-4, while others view it as a new scaling law and a significant step forward in AI capabilities.
• Elad compares O1's current stage to GPT-2, suggesting it's an early proof point that will rapidly improve with more compute and data.
• Sarah notes that O1's potential will be fully realized as labs integrate tools and function calling with its planning and iterative reasoning abilities.
• Elad shares his personal experience using O1 for brain teasers, highlighting the challenge of guiding the model to consistently correct answers.

Sarah and Elad Reflect on Nobel Prizes Going to AI Researchers (18:06 - 21:23)

• They discuss the recent Nobel Prizes, noting that for the first time, AI researchers were recognized.
• The Nobel Prize in Physics was awarded for pioneering work in AI and AI models, while the Chemistry Prize went to Demis and others for protein folding applications of AI.
• Elad acknowledges the world-changing nature of the work but questions if the physics award directly advanced the field of physics itself, suggesting it recognized the application of physics principles to AI.
• Sarah, with her biology background, finds the chemistry award for AI in protein folding more obviously merited due to its profound impact on predicting protein structure and function.
• They humorously speculate on who from the AI community might win a Nobel Peace Prize, suggesting an "all AI Nobles" category.

Jobs and Businesses at Risk of Disruption (21:23 - 27:18)

• Elad discusses job functions most at risk from AI, focusing on roles involving repetitive "email jobs" like customer support and success, where AI can augment or replace human labor at scale.
• He highlights AI's advantages in multilingual support and 24/7 availability.
• Sarah adds that AI will disrupt not just existing SaaS companies but also change how people are utilized.
• They discuss how AI can impact niche vertical software companies, similar to those acquired by Constellation Software (e.g., campground management, window design software).
• Sarah suggests that end-users might be able to generate more of this software themselves, or new generations of web hosting companies will emerge as web development becomes more AI-driven.
• They anticipate AI attacking both the "people aspect" and the "tools aspect" of work, as users trade off infinite control (like Adobe) for cost and speed (like Canva).

AI-Durable Companies (27:18 - 29:01)

• Elad introduces the concept of "AI-durable" companies, where AI's impact is minimal, making them robust against the coming wave.
• He uses the example of a railroad due to its physical infrastructure and regulatory barriers.
• In the software world, Rippling is cited as an AI-durable company due to its comprehensive bundle of services, despite having recently rolled out an AI product.
• Sarah agrees that while systems of record companies might seem durable, some are more vulnerable if they are merely "data stores pretending to be something more" without a strong ecosystem.
• They conclude that companies with complex business process maintenance and customization, especially in the enterprise, remain durable, though AI can attack the services spend associated with them.

Notable Quotes

"Notebook LM kind of feels like something special to me from the perspective of you can upload different documents or information and it'll basically turn it into two AI posts effectively two AI Bots using voice discussing it as if it's a podcast." — Elad Gil

"I think that will have an obvious impact of being less concerned about society." — Elad Gil

"I think the premium put on humanlike communications is is going to decrease right just by volume of how how possible that is." — Sarah Guo

"I think the more interesting thing is where you're augmenting or replacing people at scale." — Elad Gil

"You almost ask what is AI really good at and there's going to be a big shift as we see real-time voice models sort of kick in." — Elad Gil

"I think there are going to be companies that I I call like AI durable the ones where AI doesn't matter that much." — Elad Gil

"I think it's more vulnerable than that I just think Rippling is especially robust but I do think there's some areas which are basically like a data store pretending to be something more." — Sarah Guo

"Well I mean listeners you heard it here first like Clone War artificial WBS it's all we got the episode in which you find out Sarah and a lot are Lites good the future his hero is just not really distributed that's the takeaway as always." — Elad Gil