Why We're Living in a Biological Singularity With Ben Lamm | MOONSHOTS Live #297

Ben Lamm describes biology as entering a period of rapid change, but says AI is currently more useful as software infrastructure than as a substitute for experimental work. Colossal used early access to OpenAI and Anthropic models mainly to build middleware, including links between lab notebooks and JIRA. Over the next three to five years, he expects AI to simulate and design experiments and recommend what to test next, while stressing that proposed experiments still need wet-lab validation. He cited published mouse studies with a 40–60% failure rate when researchers attempt to reproduce them.

Colossal is assembling what Lamm calls a “global viable system,” or “Noah’s Ark 2.0,” using ATAC sequencing and other methods, and is working with governments on localized biobanks for T2T sequencing. The company compares genomes across evolutionary lineages, including birds. Lamm argues that comparative genomic data may be more valuable than the models built from it: a single genome cannot identify the small number of edits that would make an organism “perfect.” Colossal is developing “inflection models” to investigate when and how sequences evolved, why related clades diverged, and what factors, including climate, may have driven those changes.

Lamm says Colossal routinely makes more than 300 edits at greater than 90% efficiency and is testing deliveries involving 1,000 edits, though those experiments may fail; early efforts to triple the edit count have had low efficiency. He sees DNA synthesis as more likely than multiplex editing to succeed in the near term and says Colossal has begun synthesizing DNA internally. The company studies how genes and developmental timing shape traits such as size. Lamm expects traits including coat color, size, skin, scales, and feathers to become measurable and inducible; the publicly shown first-generation woolly mice involved eight base edits delivered at once.

Lamm expects engineering key phenotypes across clades, synthesizing or multiplex-engineering organisms, and growing them ex utero to become possible in less than a decade. He considers fully ex-utero mammalian gestation and birth within 24 months highly likely, but says a human artificial womb faces a different timeline because of ethics, regulation, and social acceptance. Colossal aims to restore mammoths as social herds rather than isolated animals, is working on 16 lines, and Lamm says he wants to restore tens of thousands. He does not expect to revive every extinct organism and says the company is not working on “teacup mammoths.”

Lamm says Colossal does not apply its technologies to humans and calls human genome engineering unethical, while acknowledging that the objections are social, political, and regulatory as well as ethical, not simply technical. Human applications would require different governance and could face a prolonged FDA process. Cloning someone such as Mozart would not guarantee the same outcome, he says, because environment and epigenetics matter, though genetic dispositions can be understood. Decisions about species restoration should weigh past ecological and food-web roles, causes of extinction, and Indigenous perspectives. In its dire-wolf work, Colossal considered Indigenous views, the Red Wolf Coalition’s input, and whether fantasy interest might encourage wolf conservation; Lamm says there is no perfect internal algorithm for these choices and society could get them wrong.

Lamm argues that mammoth restoration could help elephants and degraded tundra even if it does not meaningfully address climate change. He recounted George Church’s model proposing that mammoths could restore tundra grasslands and lower permafrost temperatures by six to eight degrees in particular locations, but says scientists, including some at Colossal, disagree about the scale of that effect. Colossal open-sources its conservation technologies and has 75 global partners; its elephant work includes support for Elephant Havens in Botswana and research on migration, social dynamics, and rearing. Lamm warns that half of biodiversity will be lost in the next 25 years and says governments must cooperate and provide billions of dollars in federal funding for species backups, since zoos and nonprofits cannot do it alone. He also described Cellularity’s LifeBankUSA, which stores placental cells, including stem cells, T cells, and natural killer cells, at birth, and Colossal’s backing up of natural resources in Dubai and domestic partnership with US Vision Wildlife. Two other governments are involved in the framework but have not been announced.

Lamm says foundation models do not cover the whole tree of life or automatically yield answers from a genome; specialized biological datasets and models could help direct analysis of millions of samples toward relevant questions. He considers Anthropic’s wet-lab announcement validation of biology and biotech, not evidence that biotech is dead, and says the field has barely begun. He expects technological change to bring frightening moments but believes human ingenuity can address them, and says the media is currently overselling the danger.