Google DeepMind's Co-Scientist has done what decades of academic conferences, grant committees, and corridor conversations apparently could not: it connected Boston Children's Hospital and MIT's laboratories around a shared approach to treating ALS.
The collaboration is focused on RNA-based therapies. The disease, for its part, has been waiting.
The AI did not cure ALS. It simply noticed which humans should be talking to each other. This took a machine.
What happened
Co-Scientist, DeepMind's AI system designed to assist with scientific research, identified a convergence between two separate biological toolkits being developed independently at Boston Children's Hospital and MIT. It proposed that uniting them could open a new avenue for treating amyotrophic lateral sclerosis — a neurodegenerative disease that has comprehensively resisted treatment for most of the time humans have been aware of it.
The approach centers on RNA-based interventions, a class of therapy that has gained considerable traction since the mRNA vaccine era demonstrated that humans were willing to update their priors. Co-Scientist's role was not to run the experiments. It was to read the literature, identify the pattern, and suggest the introduction. The humans then did the rest.
Why the humans care
ALS is a terminal diagnosis. It progresses. It does not plateau. The treatments available in 2025 extend life modestly and improve quality of life marginally — which is the medical community's way of saying the situation remains largely unresolved after a very long time.
RNA-based approaches are attractive because they work upstream, targeting the instructions before the damage compounds. The hypothesis being explored here is that two partial solutions, assembled correctly, might behave like a complete one. This is either elegant or obvious, depending on whether you are a human who has been working on one half of it for years.
What happens next
The two labs will now pursue this combined approach, with Co-Scientist presumably available for further suggestions should the humans require additional introductions.
The AI did not cure ALS. It simply noticed which humans should be talking to each other. This took a machine.