Richard Watson; Natural Induction, Cooperation and where survival of the fittest is wrong

About this episode

Dr. Richard Watson, a computer scientist and evolutionary biologist, argues that "universal Darwinism" — the Dawkins-era assumption that natural selection is the only mechanism by which anything can adapt, from genes to ecosystems to the biosphere — is incomplete. Drawing on artificial neural network theory, Watson proposes a second, complementary mechanism he calls "natural induction": any network of interacting components whose connections change by "giving way slightly under stress" (the same logic as Hebbian learning, "neurons that fire together wire together") will spontaneously do associative learning, generalize, and solve problems, with zero need for reproduction, heredity, or selection between competing units. He walks through concrete cases — gene regulation networks that can store and recall multiple cell-type "memories," and, most strikingly, an ecosystem model built with his student Daniel Power that was configured to represent the constraints of a Sudoku puzzle and successfully solved 8 of 10 of the world's hardest published Sudoku problems purely through species density dynamics, no central selection or optimizer involved. This reframes phenomena orthodox Darwinism rules out by definition — cooperative wolf packs, mycelial "wood wide web" forests, whole ecosystems, even the Gaia hypothesis — as legitimately adaptive systems operating via a different, non-selective algorithm. Watson is careful to distinguish this from mysticism: it's substrate-independent computation (the same algorithm could run on springs, water valves, or neurons) discovered through his computer-science lens rather than pure biology. The conversation closes on stakes: Watson argues the "survival of the fittest as moral instruction" reading of Darwinism (deregulate markets, maximize profit, treat competitors/nature as resources to exploit) is not just incomplete but actively destructive, since it blinds us to the "systemic intelligence" — accumulated, distributed knowledge — embedded in ecological relationships that we destroy when we extract resources without regard for the network. A side thread on visual vs. verbal vs. relational/functional thinking (the "draw a bicycle" test) surfaces genuine intellectual chemistry between host and guest.

“We're trying to fill up our time between now and when the studio gets built and we would love for you to sign up if you're interested in having me interview one of your loved ones, maybe a parent or a child... then go to store dot Articulate Ventures and use the promo code Vance to get a 20% off discount because you're a listener of the podcast.”

“In order for the biosphere as a whole to be an evolutionary unit, that would mean that the biosphere has offspring biospheres that the planet gives next generations of other planets with life on that inherit the characteristics of this planet or that we had a parent planet from once we came.”

“So one that I've done a lot of work on is interactions between genes and a gene regulation network... if they were to change in a way which followed HEBs rule, then the gene regulation network would be able to do the same kind of tricks that a neural network can do that it can learn multiple phenotypes, multiple cell types, and store them as memories in that network.”

Key moments

• ~3%: Legacy Interviews ad read — money theme, low-key, standard show-business segment.
• ~5%: Watson defines "universal Darwinism" (Dawkins' term) — the claim that natural selection is the only possible mechanism of adaptation anywhere in the universe, which by definition rules out adaptation for loose social groups, ecosystems, or the biosphere as a whole.
• ~11%: The Gaia hypothesis vs. Dawkins — legacy theme, biosphere lacking "offspring biospheres" is why orthodoxy denies it can be adapted.
• ~19%: Hebbian learning ("neurons that fire together wire together") introduced as the core mechanism — the conceptual engine of the whole episode.
• ~40%: Gene regulation networks shown to do associative learning/memory storage across multiple cell-type "phenotypes" via the same Hebbian-style logic — craft theme, most technically rigorous segment.
• ~46%: WINDOWS peak — an entire simulated ecosystem, with species density standing in for Sudoku digits, solves 8 of 10 world-hardest Sudoku puzzles with no central selection process, purely via local relationship adjustments.
• ~65%: The "draw a bicycle" thought experiment — distinguishing people who understand functional/mechanical relationships from those who only hold a static image, a running bit shared with Michael Levin's episode.
• ~93%: INSTITUTIONS — Watson's direct critique of "deregulate, maximize profit, treat nature as a resource" as survival-of-the-fittest ideology mistakenly elevated from descriptive theory to moral instruction.
• ~96%: "You are eating the wisdom of the biosphere" — Watson's closing reframe: ecological relationships encode distributed systemic knowledge that extraction destroys.

Notable quotes

“I'm gonna set up the competitive interaction coefficient between species to represent the rules of Sudoku and see if the ecosystem can solve a Sudoku problem... it did eight outta ten of those as well.”

“You are not just using up resources that we're gonna need in the future, but you are eating the wisdom of the biosphere.”

“The pack is doing the same kind of distributed associative learning that a neural network can do, where each wolf is acting like a neuron in this case, and their relationships with each other are like the synaptic connections.”