Progress on AI drug discovery

There's an upbeat article in The Economist: An AI revolution in drugmaking is under way (5 January 2026). 

It rehashes figures from a 2024 study:

AI-designed drugs are whizzing through the preclinical phase (that before human trials begin) in only 12-18 months, compared with three to five years previously. And the success of AI-designed drugs in safety trials is better too. A study published in 2024, of their performance in such trials, found an 80-90% success rate. This compares with historical averages of 40-65%.

I hope that holds up for later stages. I'm sure eventually we'll have a drug on the market that was partially AI-assisted, if not outright discovered. But right now the training data isn't there, and we don't know how some of the AI companies are doing.

But The Economist piece covers more than the molecules, and in some ways more exciting areas. First, there's the work on agents which "can generate a hypothesis about a disease, including testable predictions, and try to verify or falsify this with a literature search". I wonder if hallucinations might be helpful here, as a creative force (as long as I'm not paying for the costs of following up the crazy ones).

Second, there's clinical trial selection using data techniques. That is: finding individuals who would benefit from a drug, meaning "smaller—and thus faster and cheaper—trials". I thought that was cheating, but on reflection you want later trial phases to select those who the drug works on, and show that it does. You're not giving out a drug to the entire population, after all.

In the longer run, the possibilities for enhancing human health are enormous.