MIT Technology Review to host talk with student who designed viral genomes using AI

MIT Technology Review will hold an online roundtable on 16 October with Samuel King, a bioengineering PhD student at Stanford and the Arc Institute, who in 2025 used a generative model to propose complete viral genome sequences. The event, open only to subscribers, will stream live at 18:30 UK time and be moderated by James O'Donnell, the magazine's AI reporter.

King's 2025 work did not demonstrate the creation of artificial life; it stopped at an earlier stage. A large language model produced DNA sequences that describe entire viral structures — essentially computational blueprints that have not yet been synthesized in a laboratory. The Arc Institute, where King conducts his research, specializes in large-scale biology and computational tools, making the integration of generative models with genomic design a natural progression for the group.

His selection for the magazine's 2026 Innovators Under 35 list places him among researchers identified for long-term impact potential rather than immediate commercial breakthroughs. The list tends to highlight research directions over finished products, and in this case the signal is clear: AI-assisted genomic design is regarded as a trajectory worth tracking.

The decision to place the conversation behind a paywall reflects MIT Technology Review's strategy in recent years — using deep content as a retention mechanism rather than a distribution tool. Two critical reviews appear alongside the event listing: one on the absence of genuine reasoning in large language models, the other on recursive AI self-improvement being further off than commonly assumed. The juxtaposition is deliberate; the magazine frames King's achievement within current limitations, not as evidence those limitations have been overcome.

According to the event description, the discussion will centre on "new ways of seeing biology," a broad framing that leaves room for questions about experimental validation, safety, and the gap between a computational blueprint and a functional viral particle. King will need to explain how the model selects sequences, which biological constraints it respects, and what prevents it from generating sequences that are statistically plausible but functionally meaningless. Concrete answers will carry more weight than any headline.