Taking those summaries, Robin created a series of hypotheses about the disease mechanisms for macular degeneration and used these tools to provide a detailed report on the evidence for each mechanism. An LLM judge then made pairwise comparisons between the hypotheses, resulting in relative rankings – somewhat like Google’s tournament system.
Similarly, the system was re-deployed to suggest cell lines and culture conditions that could provide a model of macular degeneration, and it produced reports on 30 candidate drugs. According to the FutureHouse team, “These reports included a rationale for why each drug is suitable for alleviating the disease mechanisms demonstrated in in vitro models and what potential limitations the drug may have.” Again, these reports were evaluated by human experts to determine which tests to pursue.
Robin also suggested assays to test drugs that humans evaluated (in most cases, it appears that they used variants of the drugs he suggested).
The main difference with Robin is that it includes a tool, Finch, that can automate the evaluation of data from some standard biological screening assays such as flow cytometry and RNA-seq. So, as long as your tests include one of the tests that Finch can handle, there is one additional step that can be performed by the system.
As mentioned above, Robin came up with a new hypothesis: increasing the ability of retinal cells to pick up debris outside the cells might provide some protection against the disease. And it identified a drug that appeared to provide the same kind of enhancement in the experiments it had proposed.
As Google found, it makes sense to have tools specifically designed to interface with the scientific literature. Swapping out Crow for OpenAI’s O4-Mini took the rate of hallucination references from zero percent to 45 percent. FutureHouse also looked at the performance of OpenAI’s research-focused tools and found that, in all cases where it suggested drugs that Robin did not come up with, those drugs failed to have an effect on these cells.
<a href