In brief
In London, Basecamp Research stores a century’s worth of human genes to train AI that speaks biology, echoing the promise of tools like AlphaFold. A feverish dispute between mathematicians over OpenAI’s Navier–Stokes claim exposed the risk that AI could absorb and manipulate confidential research. Across biology, physics and chemistry, publishers warn of a dangerous flood of AI‑generated “slop” – fabricated citations, phantom papers and hallucinated results – threatening the integrity of peer review and the future of funding. While harnessing AI can unlock new insights, the moment it produces claims without human guardianship, it risks swamping science with noise that could be mistaken for truth. The debate therefore marks a watershed moment: AI must be combined with rigorous oversight to preserve both progress and trust.
In the cool silence of a former watch factory in London, a bank of freezers hums while holding more than 100 billion genes, the data that Basecamp Research hopes to feed into the first AI that directly speaks the language of biology.
The company’s ambition echoes that of tech giants that have turned waves of text into powerful language models, but it carries a new set of risks. A heated exchange between mathematicians in New York last month over a claim that OpenAI’s models had solved the long‑standing Navier–Stokes equations revealed a potential for AI to indirectly absorb confidential research and influence results that have never been published.
Meanwhile, the relentless wave of AI‑powered breakthroughs is reshaping both biology and physics. Google deep‑learned the structure of proteins with AlphaFold in a single year, yet the algorithm also produced millions of crystal structures that the scientific community is still debating for authenticity. Similar “AI slop” – hallucinated papers, phantom citations and fabricated data – is now a growing concern for publishers, who are scrambling to add detection tools and tighter checklists to keep peer review reliable.
From DeepMind’s dazzling 800‑year leap of knowledge to university labs uncovering sperm‑egg interactions with AI assistance, the technology offers a powerful shortcut—but it also demands clear accountability. Critics warn that an entirely algorithm‑driven research pipeline could erode the very trust on which science rests, while proponents argue that the very human oversight needed to vet AI outputs is what keeps discovery honest.
The AI‑in‑science debate is a watershed moment: the next decade will decide whether these algorithms become trusted partners or become the very threat that turns the scientific method on its head.
