Two AI-based scientific assistants succeed with drug recycling projects

Taking this overview, Robin then developed several hypotheses about the mechanisms of macular degeneration and used these tools to provide a detailed report of the evidence for each mechanism. The LLM judge then compared the two between hypotheticals, which resulted in similar rankings – just like Google did.

Likewise, the system was also used to identify cell lines and culture that can provide a model of macular degeneration, and prepared 30 reports that require treatment. “These reports had all the reasons that prove why any drug is necessary to reduce the disease process represented in the in vitro model and the restrictions that the drug can cause,” according to the FutureHouse team. Again, these reports were analyzed by human experts to determine which tests to proceed with.

Robin also gave experiments to test the drug, which people tried (in most cases, it seems that they used the different types in question).

The main difference with Robin is that it includes a tool, Finch, which can analyze data from other biological analysis tools, such as flow cytometry and RNA-seq. So, as long as your test includes one of the tests that Finch can handle, then there’s still some work to be done with the system.

As above, Robin came up with a new idea: Increasing the ability of eye cells to take debris out of the cells can provide protection against the disease. And it identified drugs that seemed to provide the same kind of stimulation in the trials they administered.

As Google discovered, having tools designed to support scientific writing is essential. Crow’s exchange for OpenAI’s o4-mini took the number of referrals from zero percent to 45 percent. FutureHouse also looked at the performance of the OpenAI research tool and found that, every time it reported that the drug that Robin did not come up with, the drug failed to have an effect on these cells.