When I was in grad school both of my parents developed cancer (!) and thus, doing mostly stats at the time and with all my bio training, I dug into the literature. The numbers were all grim and the role of a lot of screening was surprising! But the biggest issue was the hype /1 https://twitter.com/stochastician/status/1319086942521335808

My parents would see a story on TV about "MIT developers treatment X" or "new Stanford research Y" and ask me, day after day, if it could help them. I always had to say no, and explain to them that these were just prototypes, ideas, demonstrations. Clinical impact was far. /2

The reality was often worse, and was something I saw repeatedly even in my own department. Fundamental research, while important (crucial!), was often carelessly spun into a promise of near-term clinical impact. Often this was not deliberate or malicious (although not always) /3

A grad student optimistically expressing to someone in the news office that "someday" it might lead to a treatment. A professor referencing a disease that may have originally spurned interest. I blame the news offices for a lot of the out-of-control hype. /4

Understandably, the public focuses on the clinical potential. My parents care far more about science giving them hope than they do about MAP kinase pathways. I think as scientists we have tremendous responsibility to police our own hype, because people will run with it. /5

I see this happening with ML technologies. I see promises that will obviously never be fulfilled. I see weasel words and self-promoting singulatarians. And it happens a lot at the intersection of ML and biology. /6

We have to do better, because policy decisions are being made on this hype. We have to do better, because as a field we should value humility. And we have to do better, because I'm tired of telling my loved ones that no, there's no magic bullet, in spite of what they saw on CNN.