Via NewScientist (26 January 2019, paywall) I learn something new about the pharma patent wars:
… when a pharmaceutical firm’s patent on a drug expires, it may still hold patents on many of the chemical steps to create it. In other words, rivals that are able to copy a drug after its patent ends often have to pay the original firm to use the still-patented recipes.
So why haven’t the patents on the steps expired? But this is part of a larger article on how this profit-strategy is being obviated by the generics manufacturers:
[AI program Chematica] learned hundreds of patented reactions then devised recipes that avoid them (Chem, doi.org/czs4). “Pharmaceutical companies spend billions making sure there are no loopholes and people think these patents are bulletproof,” says Grzybowski. “But actually it seems quite possible to get around them.”
Chematica caused an earlier stir in 2012, when Grzybowski and his colleagues used it to find ways of making the nerve agent VX from readily available chemicals including water, table salt and sulphuric acid. This earned him an invitation to the Pentagon, he says, where he called for better regulation of chemicals that can be used to make weapons.
So having written the post title before finishing the article, I find out that War is even more apropos than I intended, which makes me a little sad. Nor is it an idle question to ask when a terrorist group will gain access to Chematica or similar program, for after all Chematica isn’t going to have the moral agency to tell them No. So long as Chematica lacks anything resembling moral agency, it’s just like any other tool – it can be used for good or for evil.
But, from a computer science perspective, I can’t help but wonder if this is a brute force approach, or if there’s some real ML (Machine Learning) application going on here, and, if so, what rules it learned on its own. (I’m a zero when it comes to chemistry, though.)
