How Benchmark Invests in AI

In today's episode, we discuss Benchmark's investment philosophy in the AI era with general partner Eric Vishria, and Sergiy Nesterenko, the Founder and CEO of Quilter. Sitting in for Erik Torenberg as host is AI Scout and @CognitiveRevolutionPodcast host Nathan Labenz. They cover the questions that Benchmark asks to determine whether a new company is solving an enduring or temporary problem, and the reasons why Benchmark hasn't invested in a foundational model to date. They also discuss the innovation behind Quilter's groundbreaking use of reinforcement learning to automate integrated circuit board designs. They delve into the importance of thinking beyond 'co-pilots' to fully automated AI solutions, and explore the balance of research and engineering in the AI space.

--
SPONSORS:

--
LINKS:

Automating Circuit Board Design Using Reinforcement Learning w Sergiy Nesterenko, Founder of Quilter:

Chris Dixon’s The Idea Maze:

--
TIMESTAMPS:
(00:00) Intro
(01:13) Eric's Investment Thesis at Benchmark
(03:35) The "AI and Done" Approach
(07:15) Diverging from Large Language Models
(09:27) The "Idea Maze" Framework
(12:40) Disruptive Innovation
(16:00) Exploring the Data Landscape
(16:45) Data Limitations and Reinforcement Learning
(18:06) Overcoming the Sparse Reward Problem
(19:17) Human Heuristics and Intermediate Rewards
(21:12) Extrapolation vs. Interpolation
(24:43) Trusting the Reward Signal in Circuit Board Design
(25:52) Physics as an Unambiguous Oracle
(27:11) User-Defined Search
(29:45) Compute Allocation
(31:53) Different Physics Simulation Techniques
(34:13) The Diffusion Model Approach
(38:07) Advantages of Reinforcement Learning
(40:08) Timeline and Bottlenecks
(41:43) Unsolvable Problems and Gradual Progress
(42:03) The Future of Circuit Board Design
(50:22) Benchmark Portfolio
(47:31) Challenges and Opportunities in Each Category
(53:34) The Gap Between Research and Engineering
(53:52) Call for Startups
(54:27) Wrap