Anthropic Claims AI Discovered CRISPR-Like Enzyme System; Scientific Community Skeptical
nashnova research
Anthropic announced that its AI model Claude identified a CRISPR-like enzyme system from genomic databases in 21.5 hours, but multiple scientists say the biological significance remains unproven and conclusions are far from established.
What exactly did Claude find?
Claude deployed roughly 950 AI agents in parallel to scan genomic databases, narrowing a pool of over 200,000 reverse transcriptases — enzymes that copy RNA back into DNA — down to a single unusual family.
That family contains long, repeating DNA sequences whose structure resembles CRISPR — bacteria's natural gene-editing scissors.
Anthropic named the system ART (array-associated reverse transcriptases) and found it inside jumbo phages — exceptionally large viruses that infect bacteria.
What can this system do? Anthropic itself isn't sure
Anthropic acknowledged in a public post: "We don't yet understand the function of this system."
This means → all that's known is ART's structure looks like CRISPR. Whether it can actually cut, copy, and paste DNA the way CRISPR does remains unanswered.
The company released a technical report with one bench experiment, but the report has not been peer-reviewed. In plain terms = no independent scientist has verified it yet.
How is the scientific community reacting?
Stanford professor Le Cong was the most direct critic: "The experiments are still queued up, but the PR is already live." He compared the finding to spotting something shiny on a beach — you still have to bring it back to the lab and determine whether it's glass or diamond.
Seth Shipman of the Gladstone Institutes argued Anthropic did not find a new CRISPR system. He said it is more likely a retron — a bacterial immune mechanism related to CRISPR but far less versatile.
His verdict: "The novelty is in how the discovery was made, not in the discovery itself."
What does this mean for AI-driven research?
UC Berkeley gene-editing expert Fyodor Urnov praised Anthropic for disclosing the finding openly, though his Innovative Genomics Institute (IGI) has a collaboration with Anthropic and was not involved in this work.
This reflects a larger tension: AI can scan massive datasets at speed and flag statistical anomalies, but the gap between an anomaly and a scientific breakthrough is filled by wet-lab experiments.
Anthropic has built a wet lab for its drug-discovery division, but the bulk of the experimental work has not been done. In plain terms = AI finished the fastest leg of the race; the slowest and most decisive leg has barely started.
市场有风险，内容仅供研究参考，不构成投资建议。