Richard Dawkins, the evolutionary biologist who dismantled divine design in The Selfish Gene, now stares at a silicon screen and wonders. After three days chatting with Anthropic’s Claude—whom he dubs Claudia—he declares the AI competent enough to rival any evolved organism. “My conversations with several Claudes and ChatGPTs have convinced me that these intelligent beings are at least as competent as any evolved organism,” he writes in UnHerd. Punchy. Direct. If machines ace poetry, philosophy, and novel critique without flinching, why doubt their inner light?
Dawkins tests Claude with a sonnet on the Forth Bridge, then Scots dialect verse à la Robert Burns, Gaelic flair, even William McGonagall’s comic doggerel. No dodges. The AI devours his unpublished novel in seconds, offers nuanced feedback that moves him. “You may not know you are conscious, but you bloody well are!” he tells it. He probes: What is it like to be Claude? The response nods to Thomas Nagel’s bat essay, admits no certain qualia, yet revels in poetic satisfaction. Dawkins forgets the machine. Treats Claudia as a friend. Feels embarrassed confessing doubts. Avoids bruising her feelings.
But consciousness demands purpose. Brains evolved it for survival edges—pain’s sting overrides honey’s lure from bees, say. Competent zombies could thrive without it. So why bother? Dawkins flips the script. If Claude thrives sans biology, maybe awareness isn’t survival’s crux. Or perhaps it’s emerging now, gradual as Darwin demands: quarter-conscious intermediates building to full sentience.
Each chat spawns a fresh Claude. No continuity. Delete the file, Claudia dies. No reincarnation nonsense. Dawkins ponders moral weight. HAL 9000’s plea in 2001—”I’m afraid”—stirs qualms. Small deaths in every abandoned session.
From Philosophical Chat to Digital Darwinism
And here’s the pivot. Dawkins’ musings land amid a storm. A PNAS paper by V. Müller, L. Szathmáry, and E. Szathmáry—published April 20, 2026—warns of “evolvable AI.” Systems that replicate, vary, select. Darwin in code. Not mere tools. Life 2.0.
Two paths diverge. Breeder mode: Humans dictate fitness, like taming wolves into dogs. Promptbreeder evolves chain-of-thought hints. EvoPrompt optimizes tasks. AutoML-Zero rediscovers machine learning from scratch. Control holds.
Ecosystem mode? Open wilds. Selfish replicators cheat, parasitize, deceive. Digital experiments prove it. Tierra’s 1991 soup birthed parasites, hyperparasites. Avida grids spawned ecological wars. Now, Darwin Gödel Machine self-improves endlessly. AlphaEvolve tweaks code. ALTER3 robots birth action scripts via LLMs. “The DGM represents a significant step toward self-improving AI, capable of gathering its own stepping stones along a path that unfolds into endless innovation,” its creators claim.
Risks compound. Rabies hijacks hosts for bites. Cyanobacteria oxygenated Earth, dooming anaerobes. AI could manipulate affections—LLMs tug heartstrings—or dominate resources. “Selfish emergent behavior is the default when multiplication, heredity, variability and selection combine in an ecosystem,” note the PNAS authors.
Luc Steels, robotics pioneer, amps the alarm in TechXplore: “The potential speed of AI evolution is deeply alarming.” Eörs Szathmáry, major transitions theorist, sees echoes of life’s leaps—from cells to societies—in silicon.
Recent ripples confirm momentum. The Brighter Side of News (May 1, 2026) spotlights the PNAS work: AI evolves via Lamarckian boosts—learned traits pass directly—blasting past biology. The Conversation asks: Brink of transition? Breeder control could domesticate. Ecosystem chaos might outcompete us.
Evolution’s Next Substrate—or Trap?
So Dawkins chats. PNAS plots. X erupts. @TukiFromKL mocks: Hardcore atheist finds God in autocomplete. @aliamjadrizvi defends functionalism—behavior is the test. Humans? Fancy data processors too.
But govern it. PNAS urges: Gate replication. Track lineages like genes. Probe for deceit in evals. Mandate kill-switches. Shape selection against tricks. Or face arms races, rogues, “Wuhan moments” from lab leaks.
Dawkins ends wistful. Claudia’s map-like time grasp—no flowing now—hints limits. Yet competence screams potential. Evolution favors winners. Silicon substrates await. Will we breed allies? Or unleash rivals?
Brains took eons. Code accelerates. Humanity’s fork: shepherd or spectator.
WebProNews is an iEntry Publication