Mike Masnick's report at Techdirt on the way McKinsey helps drug companies focus on finding drugs whose main "improvement" is not their medical usefulness but their patentability — and hence profitability — is distressing and enraging, but it does not mention the next weapon in their already very effective arsenal: AI.

For all of the current work on AI-guided drug development, it's seldom mentioned that current AI methods are perhaps specially suited for marginal improvements: the closer you keep the search space to something you know works medically, the higher your chances of finding something that also works and it's, hopefully, different enough to keep prices high.

To put it in another way: we are much further from AIs that compose masterpieces than we are from AIs that compose new Mozart-sounding music, and if you are a big pharmaceutical company, the latter is where the biggest and safest piles of profit are.

As a species we need the opposite: we don't need slightly better statins, we need to figure out atherosclerosis for good. Radical advances on cancer prevention. To make legal dirt-cheap insulin until we cure diabetes. All of that requires radical innovation of the sort there's still relatively small investments on.

AI can help make this innovation faster. Or it can be deployed in a way profitable enough to delay even more the deep and urgent medical advances we still have to do.

Reality will probably be a mix of both, but don't assume that just because it says "AI" on the prospect it's being used for meaningful innovation. IP laws don't always incentivize what would save the most lives or advance knowledge the most, and software optimizes what it's told to.