AI- hurting black people by the millions...

At home with my son during a recent flare up in the local conflict, I am less willing to leave the house with him. Some of my fellow US citizens are probably wondering why I don't leave the house and board a plane with him. I live in Israel, but I still exercise my USA levels of caution and paranoia about violence- and that is why I can't come back. It's an example of flawed thinking on the part of my fellow Americans to assume I'd be better off in New York City than Tel Aviv, really. In the USA, more than once I came home to find a neighbor shot and bled to death between my car and my apartment complex door. In all cases the police tape was up by the time I got home. In the USA, I personally was the victim of crimes ranging from violent to white collar ( identity theft after my wallet was stolen, what a MESS!). In the USA, in a leading city, at one point I lived in an area where someone was shot on a more or less daily basis. But today oddly, I'm afraid to go buy toilet paper with my son, lest we be the victims of violence, in a country I have never had violence perpetrated against me specifically. As the victim on one terrorist attack already (in the USA, 9/11) I can not deny the hardship of them, but let's face, it they are statistically less likely than mundane problems that may end up killing me. Yet it's hard to make these calculations rationally.

When I am being rational, what I completely expect to do me in, in the end, is a medical error, and that is why I stay out of the USA. Ever since the time I had an unexplained hypertensive urgency, and I was discharged by an emergency room doctor after only being tested for street drugs, and sent to psychiatry on the premise that if I thought I must be a doctor I must be delusional, I realized, racism could actually literally end up killing me.

So months later, I remember writing to some of my international colleagues in every space I could that AI algorithms would likely end up worsening the racism in the US medical system. Turns out it had already happened.

BY MY LOWEST ESTIMATES BAD HEALTHCARE AI PROGRAMS HAVE EFFECTED OVER 7 MILLION AFRICAN AMERICANS. Thats well over 2000 times the number of people effected by the bombing I survived on September 11th decades ago. Now certainly, algorithms, even ridiculously bad ones, alone can not kill people. We are talking about code, not bullets...but if the fatality rate from bad medical decisions is even one in 2000...well, you do the math.

I place a link below. I may update with more analysis- but suffice it to say, it happened...and I know this won't be the last time. In order to combat the problem, I'm building something like a Jupyter notebook to teach docs the basics of coding when my kid naps...it will take months at the rate I am going...but someone must do this. In the meantime If nothing else doctors need to understand that code is not produced by objective people on clean data, it is in many cases messy dirty data is now 'encoded' into algorithms, so to speak metaphorically. They need to learn how to do adversarial testing of algorithms- it doesn't require code, it just takes brains.

https://science.sciencemag.org/content/366/6464/447