E21: VC Insights on Investing in Artificial Intelligence with Sarah Guo and Elad Gil of No Priors Podcast

Episode Summary

Executive Summary: Sarah Goa and Elad Gil argue AI is creating a new software era where value will split between incumbents and startups, with the most exciting opportunities in vertical applications, tool use, voice/media, social products, and agentic workflows. They emphasize product-market fit over defensibility early on, caution against overreliance on waitlists, and predict major shifts in human-AI interaction, education, and regulation.

Main Topics: Software 3.0 and the new AI startup landscape (Priority: 5/5): Sarah frames AI as 'software 3.0': a wave of unexpectedly powerful software businesses enabled by exponential ML capabilities, new business categories, and novel UX paradigms beyond the chat box. Where value will accrue: incumbents vs startups (Priority: 5/5): Elad argues each tech wave distributes value differently; AI will likely split value, but still leave substantial room for de novo startups across applications, tooling, and new interfaces, unlike the first decade of AI which mostly benefited incumbents. High-potential application areas (Priority: 5/5): Both guests highlight voice, dubbing, translation, social products, consumer companions, B2B workflow automation, retrieval/memory, and media generation as especially promising areas with strong demand signals. Human-AI interaction models and agents (Priority: 4/5): The conversation explores co-pilots, supervision modes, delegated agents, and task-specific workflows. They stress that interaction will be multimodal and context-dependent rather than a single universal chat UI. Product-market fit, waitlists, and launch strategy (Priority: 5/5): Elad and Sarah criticize overly curated waitlists and slow launches, arguing real usage matters more than hype. They stress that many AI demos fail to convert, while the strongest products show organic, sustained demand. Societal impacts, education, and ethics (Priority: 4/5): They discuss deep implications for kids, education, values, and who controls AI-mediated information. Questions include RLHF bias, cultural differences, and what it means if bots increasingly shape learning and behavior. Regulation and long-term risk (Priority: 4/5): Short-term regulation may entrench incumbents and slow innovation, especially amid an AI-influenced election cycle, while long-term risks include species-level competition with AI/AGI.

Key Arguments: AI is enabling a new class of software businesses that can attack previously underserved categories like copywriting, illustration, law, and media transformation. The current AI wave will not be purely an incumbent wave; startups can capture meaningful value, especially in applications, interfaces, and workflow layers. Voice, dubbing, synthetic media, and translation show unexpectedly strong demand and are attractive because they convert one media form into another cheaply. Many AI products should be built around specific workflows, memory, retrieval, and action-taking rather than generic chat interfaces. Tool use and automation will become much more powerful as models reliably interact with software systems, fill forms, run code, and execute tasks. Social AI products should not merely recreate Twitter/Facebook; the most promising ideas will create new interactions that feel native to generative systems. Consumer adoption in AI is often masked by hype; real signals are sustained usage, paying customers, and organic demand rather than closed waitlists. Early-stage AI products may appear weird or toy-like, but that is often where the best startup opportunities begin. Defensibility should be assessed later; early on, the key questions are whether people care, use the product, and whether the market is large enough. AI will likely reshape education and parenting by creating highly personalized tutors/companions, raising major questions about values and control. Overregulation is a major near-term concern because it could lock in incumbents and distort markets before the technology’s benefits are fully realized.

Data Points: Conviction VC fund size: $100 million - Sarah Goa’s AI-focused venture fund launched to invest in software 3.0 companies. AI startup value split (estimate): 80/20 incumbent/startup - Elad Gil suggests AI may resemble a differential split where incumbents capture most value but startups still capture a significant share. First internet wave value split (historical comparison): 80% startups - Used as a contrast to explain how value accrued in the first internet wave. Mobile wave value split (historical comparison): 80% incumbent value - Used as a contrast to explain how value accrued in mobile. Crypto value split (historical comparison): 100% startup value - Used to show that different technology waves distribute value very differently. AI first decade outcome: Mostly incumbent value - Elad characterizes early AI (CNNs, RNNs, GANs) as a wave where incumbents captured most of the gains. Company scale benchmark: ~20 people - Sarah says there is now empirical proof that a one-to-ten-billion-dollar company can be created with about 20 people. Waitlist warning sign: 10 years ago - Elad describes an old AI company whose fake waitlist masked manual human operations in the background. Time horizon for better AI bots: Less than 6 months - Elad predicts bots that can actually do work for users are very near-term. Short-term optimism horizon: 5–10 years - Elad says he is very optimistic about AI’s near-term effects on health, education, and global costs over this window. Long-term AGI concern horizon: A few decades - Elad describes himself as a long-term doomer about eventual species competition with AI/AGI. Election risk probability: 1-in-5 chance - Elad estimates a meaningful chance AI becomes a major issue in the next election cycle.

Pivotal Quotes: "I think it's shorthand for just believing that there's a very unexpected new set of software businesses emerging that can be very important." — Sarah Goa: Defining 'software 3.0' and why she launched an AI-focused fund. "I think the next great company starts off looking like a toy." — Elad Gil: Why weird, non-obvious behavior is often an early signal of breakthrough startups. "The real thing you're trying to figure out is like, can I make something that just creates so much of its own demand?" — Elad Gil: On evaluating AI startups and distinguishing real traction from hype.

Implications: Listeners should expect AI to reshape products, workflows, education, and media through specialized agents and new interaction modes. For founders, the best opportunities may be in weird, narrow, high-demand use cases rather than generic chat, while policymakers risk slowing innovation if they regulate too early.