Skip to content

What Most People Get Wrong About Evolution | Akarsh Kumar

Machine Learning Street Talk (MLST)

47m
Transcript

Original podcast

Where to listen

Summary

Written by a model from the transcript; not yet reviewed.

This episode of Machine Learning Street Talk features guest Akarsh Kumar discussing the automated search for artificial life (ASL) using foundation models, contrasting statistical intelligence with regularity-based intelligence, and exploring how evolution serves as a powerful optimizer. The discussion examines how artificial life simulations like Conway's Game of Life, Lenia, neural cellular automata, Boids, and particle life can be parameterized and searched using vision-language models to uncover open-ended complexity and emergent phenomena.

The guest and host explore why the path of learning matters more than the destination, analyzing how curriculum learning and sequential representation building act as strong forms of regularization. They explain that foundation models can act as subjective critics over simulated universes, allowing researchers to explore computational spaces without hardcoding specific rules. The episode highlights how combining large language models with evolutionary loops solves discrete combinatorial problems like Core War programming games, demonstrating that evolution is far from random and can transform exponential search problems into linear ones.

  • Parameterize a space of simulations rather than hardcoding a single simulation to discover emergent life forms more easily. 2:07
  • Use foundation models as critics over unrolled simulations to evaluate subjective complexity without human intervention. 2:13
  • Leverage evolution combined with language model mutations to solve discrete combinatorial programming games effectively. 3:25

Chapters

  1. 0:00
    Political Ad Introduction

    The episode opens with a political advertisement segment that sets a tone before diving into the main discussion.

  2. 0:59
    Evolution is Not Random

    The host clarifies that evolution is a purposeful process, not a random one, and introduces the concept of artificial life.

  3. 2:13
    Foundation Models as Evolutionary Critics

    Discussion of how foundation models can serve as critics in evolutionary search for artificial life simulations.

  4. 4:20
    Open-Ended Search and Diversity

    Exploration of strategies to find diverse, open-ended simulations using evolutionary algorithms and foundation models.

  5. 41:37
    Artificial Life and AGI

    Connecting artificial life research to the broader goal of building general-purpose intelligence.

  6. 44:05
    Core War Evolution Experiment

    Description of an experiment evolving assembly programs in the Core War environment using LLMs.

  7. 44:40
    Map Elites and Core War Warriors

    Explains using Map Elites to navigate deceptive program space and create warrior sequences.

  8. 45:23
    Beating Human Warriors with LLM Evolution

    Shows that evolved warriors can defeat 96% of a dataset of human-crafted programs.

  9. 46:02
    Generalization and Behavioral Vectors

    Describes how later warriors become more general and exhibit reduced behavioral variance.

  10. 47:14
    Evolution vs Random Search in AI

    Clarifies that evolutionary algorithms are far from random and are powerful for combinatorial problems.

  11. 48:07
    LLM Mutations Outperform Random Mutations

    Highlights how LLM-guided mutations vastly improve success rates over random ones.

Highlights

  • 0:59

    Evolution Is Not Randominsight

    The guest clarifies that artificial life and evolution are anything but random, functioning instead to explore the space of all possible intelligences and life forms.

  • 2:07

    Searching Simulation Spacesstrategy

    Instead of hardcoding a single physics rule, researchers parameterize a simulation space and use foundation models to evaluate the outcomes automatically.

  • 2:21

    Islands of Open-Endednessdata

    Plotting 260,000 cellular automata rules reveals that most rules are not open-ended, leaving only a small island of solutions where the coolest simulations lie.

Quotes

  • “Artificial life is really about, like, understanding, like you said, life as it could be rather than life as it is.”
    Unnamed speaker · 1:06
  • “Evolution is probably the most powerful algorithm that we know to solve like discrete combinatorial problems.”
    Unnamed speaker · 4:12
  • “We need to understand the space of all possible intelligences, the space of all possible life forms that can be.”
    Unnamed speaker · 6:17
  • “The curriculum is not the only thing that matters because people have been doing curriculum learning in RL.”
    Unnamed speaker · 7:56
  • “And then you basically just optimize a program with an LLM as the mutation operator in an evolutionary loop to beat this program.”
    Unnamed speaker · 44:31
  • “We can beat 96% of them just by using an LLM to evolve them.”
    Unnamed speaker · 45:34
  • “If you just try to zero-shot this language, it's terrible.”
    Unnamed speaker · 45:43
  • “Evolution is anything but random and evolution is probably the most powerful algorithm that we know to solve like discrete combinatorial problems where you need to basically get pieces of the puzzle right before you get the whole thing right.”
    Unnamed speaker · 47:18
  • “The variance in the behavior of these warriors goes down as you increase the number of rounds.”
    Unnamed speaker · 46:44
  • HMP, the HouseMajorityPac.com 0:24–0:29
  • Apple Watch Series 12 0:33–1:02
  • House Majority Pac 48:53–49:25
  • Apple 49:26–49:53

About this episode

A lot of people in AI treat evolution as a dumb fallback, basically random search for when you can't take a gradient. Akarsh Kumar thinks that is wrong. Selection hangs on to partial solutions, so mutations only need to be useful about 1% of the time for the search to keep making progress. Akarsh is a PhD student at MIT working with Phillip Isola, works with Sakana AI, and is first author of the Fractured Entangled Representation paper with Kenneth Stanley, Jeff Clune and Joel Lehman. He tells Tim Scarfe why the path a learner takes may shape the structure of what it learns, and why that is a different way of looking at intelligence from the statistical one. Most of the conversation is about ASAL, the method he led for searching whole spaces of artificial worlds. Rather than predict what a rule will do, ASAL runs the simulation and asks a foundation model what happened. Mapped across all 262,144 Life-like rules, the most interesting worlds sit on one small island. Along the way: Game of Life, Lenia (with a clip from its creator, Bert Chan), neural cellular automata, Boids, Particle Life and the emergence of persistence.---TIMESTAMPS:00:00:00 Intro: artificial life, ASAL and Core War in four minutes00:04:19 Life as it could be, not just as it is00:05:46 Why the order you learn things in matters00:08:33 No shortcuts: Wolfram, novelty search and regularisation00:10:44 Kenneth Stanley: the path matters, not just the destination00:11:31 Statistical intelligence vs regularity-based intelligence00:13:49 Artificial chemistry and other possible universes00:15:59 Convergent patterns: shadows of the substrate?00:18:02 How Conway's Game of Life works00:20:49 Change one cell, change everything?00:22:31 Lenia (with Bert Chan) and neural cellular automata00:25:26 Boids, Particle Life and cell-like creatures00:28:47 Persistence, entropy and what life is00:30:15 ASAL: a foundation model as the critic00:34:04 Impostors inside the simulation00:36:50 From primordial soup to alien animals00:38:26 262,144 rules and the island of open-endedness00:40:17 From artificial life to AGI00:41:35 Core War: Game of Thrones, Turing edition00:43:22 LLMs as the mutation step: evolving warriors00:46:03 Evolution is anything but random---REFERENCES:paper:[00:04:19] ASAL (Kumar et al.)https://arxiv.org/abs/2412.17799[00:08:08] Assembly theory https://www.nature.com/articles/s41586-023-06600-9[00:09:54] FEP paperhttps://arxiv.org/abs/2505.11581[00:22:38] Lenia: Biology of Artificial Life (Bert Chan)https://arxiv.org/abs/1812.05433[00:23:20] Growing Neural Cellular Automata (Mordvintsev et al.)https://distill.pub/2020/growing-ca/[00:41:35] Digital Red Queen: Core War with LLMs (Kumar et al.)https://arxiv.org/abs/2601.03335[00:43:42] MAP-Elites https://arxiv.org/abs/1504.04909[00:44:55] AlphaEvolve https://arxiv.org/abs/2506.13131[00:47:03] AutoML-Zero (Real et al.)https://arxiv.org/abs/2003.03384book:[00:07:04] Why Greatness Cannot Be Planned (Stanley and Lehman)https://link.springer.com/book/10.1007/978-3-319-15524-1tool:[00:18:12] Conway's Game of Lifehttps://en.wikipedia.org/wiki/Conway%27s_Game_of_Life[00:25:27] Boids (Craig Reynolds)https://www.red3d.com/cwr/boids/[00:27:12] Particle Life (Tom Mohr)https://github.com/tom-mohr/particle-life[00:41:50] Core Warhttps://corewar.co.uk/other:[00:08:49] Computational irreducibility (Stephen Wolfram)https://www.wolframscience.com/nks/p737--computational-irreducibility/[00:10:46] MLST: Why Every AI Model Is an Impostor (Kenneth Stanley, FER documentary)https://www.youtube.com/watch?v=o1q6Hhz0MAg[00:28:30] MLST: Blaise Agüera y Arcas on life emerging from codehttps://www.youtube.com/watch?v=rMSEqJ_4EBk---LINKS:Akarsh Kumar: https://akarshkumar.com/ASAL project page and demos: https://pub.sakana.ai/asal/Digital Red Queen project page: https://sakana.ai/drq/RESCRIPT:https://app.rescript.info/public/share/Qfv3T0EVzqOXL_CYYeRr9Blv8HnNm4JsXBh7ByFjJbc

Transcript

0:00
47:54

10,033 words · 6 speakers

Topics

  • Artificial Life
  • Conway's Game of Life
  • Linear
  • LLMs
  • Foundation Models
  • LLM
  • Particle life
  • Core War
  • Red code
  • ASL

Page updated · transcript and summary are machine-generated