• Ezekiel Emanuel writes in the Atlantic about his impression of Joe Biden’s fitness.  Alas, I agree with his conclusions; though Mr. Biden is remarkably active for his age, he is definitely showing some signs of functional cognitive decline.  I would for for him over DJT in a heartbeat, but I think he would do a service to his country if he were to step aside for a younger candidate.

• As always, Matt Levine adroitly skewers finance-related misadventures and curiosities. His June 20 Bloomberg posting was especially comically worthwhile.  I found this AI Sorting section especially funny and enclose this excerpt:

“A dumb simple model of artificial intelligence companies is:

1. 1. It would be good to develop good AI (AI that helps humans), but bad to develop bad AI (AI that kills or enslaves humans).
   2. If you try to build good AI, there is some risk of building bad AI instead (your robot tricks you into thinking that it’s nice, then enslaves you), so you have to be very very careful. You can’t move too fast; you have to check carefully, at each step, to make sure that your robot is not secretly evil.
   3. Company A is formed by idealistic AI researchers who want to create good AI. They work together well for a while.
   4. Disagreements develop. Some researchers at Company A say “we need to work faster to build good AI, because if we don’t, someone else will come along and build bad AI first instead.” Others say “no, we can’t work faster, that would compromise our ability to check that the robot is not evil.” 
   5. The first group wins the argument, for reasons.[5]
   6. The people who lose the argument, who are genuinely worried about bad AI, quit Company A in outrage and go start Company B, with the goal of carefully and safely creating good AI.
   7. They work together well for a few months.
   8. Disagreements develop at Company B. Some researchers say “we need to work faster to build good AI, because otherwise Company A will build bad AI first. That’s why we quit, after all.” Others say “no, we can’t work faster, that would compromise our bad robot checks. That’s why we quit, after all.”
   9. The first group wins the argument, for the same reasons as in Step 5.
   10. The people who lose the argument quit and start Company C.
   11. This keeps repeating: Company C eventually splits over similar tensions, but also Company A and Company B can themselves keep dividing as some people want to move faster than others.
   12. Eventually all the AI researchers are very finely sorted by aggressiveness, so that Company Z is full of purists who are too cautious ever to build anything at all, while Company A is full of people who are like “actually being enslaved by robots would be pretty cool.””

• This was a cool study positing that the Earth passed though a dense interstellar cloud 2-3 million years ago. It was a clever use of computational modeling to recreate the paths of our solar system and the Local Ribbon of Cold Clouds through millions of years. From commentary in the Harvard Gazette re. the study by Opher and Loeb in arXiv.

“The evidence exists in the form of noticeable peaks in the deposition of two radioactive isotopes: iron 60 and plutonium 244. Both are very rare, created when massive stars explode in supernova. Those isotopes are thought to be more plentiful in the interstellar medium.

“It is everywhere, in the deep ocean, on the moon, on ice in Antarctica,” Opher said. “These papers describe a global phenomenon. Something happened. And iron 60 is not produced on Earth. So I knew that somehow this iron 60 got trapped in dust, and somehow, 2 to 3 million years ago, we had more dust delivered to us.”… “Our work should trigger more studies into this question,” Loeb said. “It draws attention to our cosmic neighborhood as having potential influence on life on Earth. We usually tend to just look at it and enjoy it, but we are actually moving through interstellar space, and there could be risks along the way.””

• Glen Kessler from the Washington Post refutes *rump’s latest hash of untruths regarding his trial and conviction.

• Matt Yglesias does a good job reporting on the ways *rump has mistreated people in his May 22 Slow Boring newsletter Trump Scams the People Who Trust Him.  An excerpt:

“What makes Trump uniquely dangerous is his disregard for the rule of law. But while it’s certainly possible that Trump will leverage that disregard to advance conservative policy aims, what he has actually consistently done throughout his career is seek personal financial benefit, specifically at the expense of his fans and admirers.”

• If you really want to understand how Israeli government policies contributed to the development of the current terrible conflict, this well-researched NYT article by Ronen Bergman and Mark Mazzetti does a superb job of documenting the problematic history.

• Liz Cheney writes in the NYT about why the Supreme Court should rule quickly on DJT’s immunity claim. The most trenchant quote from that piece:

“It cannot be that a president of the United States can attempt to steal an election and seize power but our justice system is incapable of bringing him to trial before the next election four years later.”

I agree.

• Kristen Panthagani MD, PhD as usual writes quite effectively in Katelyn Jetelina’s Your Local Epidemiologist Substack space on the dramatic benefits of the modern world (including vaccines) on pediatric mortality.  An excerpt, and illustrative graphs:

“Deaths from infectious diseases have plummeted with the discovery of bacteria and viruses, improved sanitation, pasteurization, the discovery of antibiotics, and the development of vaccines. Childhood mortality dropped astronomically, and life expectancy in grew by three decades in the twentieth century alone. The most dramatic increases are among children under 5 years old.”

• Another good piece from David Wallace-Wells in the NYT: Who ‘Won’ Covid? It Depends How You Measure. An excerpt:

“In the end, everyone got it, and probably the most important factor shaping national death totals was how many people were vaccinated before their first infection and how many weren’t. The United States could’ve done much better on that test, given that more Americans have died of Covid since the vaccines were made available to anyone who wanted them than had died to that point. But by some estimates, those vaccines also saved more than three million American lives.”

Kaiser Permanente trialed using ambient AI scribes to determine whether they were of net value to providers and deemed it a success, “saving most of the physicians using it an average of one hour a day at the keyboard” and that the quality of the transcriptions was for the most part quite acceptable.  See the implementation described here in NEJM Catalyst.