I've mentioned several times that the bulk of the scientific evidence points to a natural origin for the COVID virus, but you may be wondering just what that evidence is. First off, there's this:
The early cases of COVID in Wuhan clustered almost entirely around the Huanan Market. That's a helluva coincidence in a city of 11 million if the virus actually originated somewhere else. Note that the virology lab is eight miles away.
Second, there's the fact that there were actually two COVID lineages in the early days and both of them cluster around the Huanan Market. It's possible to conceive of one lab worker acquiring the virus, traveling directly to the market, and then having the virus spread from there. Unlikely but possible. However, the odds of two lab workers doing the exact same thing with two separate lineages is literally a one-in-a-million longshot. It's far too remote a coincidence to be believable.
Third, there's no genomic evidence of a non-natural origin. The famous furin cleavage site, for example, is found in other viruses. It isn't that hard for this short string of nucleotides to mutate its way into another virus.
There's more, and one of the best guides to the evidence is Michael Worobey, a professor of evolutionary biology at the University of Arizona. A journal article he co-authored in Science is here. A good NPR interview with him is here. And a one-hour lecture he gave a few days ago is below:
UPDATE: I originally located the Wuhan virology lab in the wrong place. It's about eight miles from the Huanan Market, not 15 miles.
Today Freedom House delivered its report on 50 years of political rights and civil liberties around the world. The United States managed to avoid another year of decline, but has still fallen precipitously over the past decade:
Currently the United States is rated just above Poland and just below Mongolia.
At the risk of being a homer, this seems . . . not right to me. Have we really gone from 93 to 83 on the freedom scale in the past decade? Or, perhaps to put it more accurately, have we really gone from 7 to 17 on the autocracy scale? Our propensity for autocracy has more than doubled?
Maybe so! But even with all the crap going on these days in the Republican Party, Fox News, and the Supreme Court, I have my doubts about this.
The response of the punditocracy to ChatGPT has entranced me. I mean, here we have a tool that, judged by ordinary standards, is absolutely remarkable. It's not playing chess or Go or Jeopardy! It's a computer program that produces high-school level text on pretty much any subject you throw at it—and will likely produce college-level and then PhD-level text in a few more years.
That's incredible. And yet, many people take a look at ChatGPT and claim to be underwhelmed. They stroke their chins and explain to us that Large Language Models are nothing like the human brain and are merely algorithms that predict text based on some previous text. Nothing to be impressed by.
Really? The implication here is that a crude text prediction algorithm can produce essays that are remarkably human-like. What does this say about human brains and the algorithms we use?
This gets to the core of my take on artificial intelligence. One of the reasons I'm convinced that it's coming soon is that—apparently—I have a much less generous view of the modern human mind than most people do. The unappetizing fact is that our intelligence is built primarily on simpleminded algorithms that rely on things like crude pattern matching and unreliable induction, all resting on the foundation of our ancient lizard brain. We very seldom produce anything very original and, what's worse, modern research has made it plain that we often have no idea why we do the things we do. We think we know, but we don't. Our self awareness is extremely unreliable.
But mine is obviously not a universal view. Today, for example, Noam Chomsky and two other researchers say this about machine learning models like ChatGPT:
We know from the science of linguistics and the philosophy of knowledge that they differ profoundly from how humans reason and use language. These differences place significant limitations on what these programs can do, encoding them with ineradicable defects.
This almost makes me weep. What do these guys think about the human brain? Isn't it clear that it too has significant limitations and ineradicable defects? There's hardly any other conclusion you could possibly draw from 10,000 years of recorded human civilization.
Then there's this about machine learning programs:
Their deepest flaw is the absence of the most critical capacity of any intelligence: to say not only what is the case, what was the case and what will be the case — that’s description and prediction — but also what is not the case and what could and could not be the case. Those are the ingredients of explanation, the mark of true intelligence.
The authors go on to talk about theories of gravity, and it's true that ChatGPT has not independently recreated Newtonian dynamics or general relativity. (And it never will since, oddly, one of ChatGPT's current weak spots is arithmetic.)
But I don't understand why the authors think that causal explanation, as opposed to simple description, is flatly impossible not just for ChatGPT, but for the entire universe of similar computer models. There's an implicit assumption here that the only way to think in sophisticated terms is to do it the way we humans do. But that's not right. In fact, we humans think very poorly, which is hardly surprising since our brains were built by blind forces of natural selection that eventually produced a machine that was pretty good at things like gossip and hunting in groups but not much else. We have since figured out how to use this machine for solving differential equations and writing sonnets—but only barely. No one should be surprised if we build AIs that work entirely differently and can think far better and more efficiently than we do. When we want to fly somewhere, after all, we don't build airplanes that flap their wings to take off.
Moral of the story: our brains really aren't that great. They're a couple of notches better than a chimpanzee's brain, and this allows us to produce some remarkable stuff. But this brain also requires massive training to read simple text, do simple arithmetic, overcome its desire to kill anything coded as a threat, and just generally get through life with even modest levels of rationality. Can we produce something better than this? I sure as hell hope so.
I always assumed that Rupert Murdoch founded Fox News, put Roger Ailes in charge, and then pretty much ignored it after it was up and running. He had black-tie fundraisers to attend, wives to divorce, and all the other accoutrements of the modern billionaire. Why waste time overseeing the daily ops of a cable news station?
But no. One of the surprising things (to me) about the documents released in the Dominion lawsuit is that Murdoch was very intensely interested in FNC, attended editorial meetings regularly, and talked to the CEO and others on nearly a daily basis:
Murdoch emerges in the documents as an extraordinarily engaged and active figure at the network in the weeks after the 2020 election, not to mention a political junkie and pundit of daily news developments, large and small.
....“Horrible,” he declared in early December 2020, after Axios reported that Trump was considering a grand finale rally to be held in Florida on Inauguration Day, to take attention away from Biden as he took the oath. Trump’s behavior, Murdoch wrote, was making it more difficult for Fox to “straddle the issue” of the election.
The most frequent recipients of Murdoch’s steady stream of missives included his son Lachlan, the CEO of Fox Corp., as well as Scott. But other emails went to a wide array of Murdoch friends, from the New York Post’s former editor to the Australian owner of television stations in Afghanistan.
What this means is that Murdoch can't pretend he was a hands-off guy who never really knew how Fox was covering the post-election news in 2020. He knew precisely what they were saying; he repeatedly acknowledges that he thinks Trump lost fairly; and that all the Big Lie reporting coming from Fox was, at the very least, over the top.
He also made it clear that, yes, he could have influenced this but didn't. That's because he was mostly worried about losing his audience, and thus some money, not about whether his network treated people and news topics fairly.
This is no surprise, but it's a bit of a surprise that he's so open about it. I guess he puts a lot of faith in American libel laws being strict enough to save his ass. We'll see.
The Wall Street Journal has a seemingly insatiable appetite for trend stories that just aren't true. Here's the latest:
Women’s Return to the Workforce Piles Momentum on a Hot Economy
American women are staging a return to the workforce that is helping propel the economy in the face of high inflation and rising interest rates. Women have gained more jobs than men for four straight months, including in January’s hiring surge, pushing them to hold more than 49.8% of all nonfarm jobs.
Blah blah blah....
I could swear that last night the headline was about service sector jobs, but maybe not. It doesn't really matter, though. Here is hiring growth since 2021:
No matter how you slice it, men have been returning to the workforce at a higher rate than women (though only slightly higher in the service sector). Nor have women been outgaining men for the past four months. By my count it's been a grand total of two months.
But maybe I made an arithmetic mistake. Or maybe the Journal is using a different measure of employment. Who cares, really? Even if women have been outpacing men for four consecutive months, that's meaningless. It's four months. And the longer term trend shows very clearly that employment has been growing almost identically for both men and women.
I really don't get it. This happens over and over. Do the Journal's editors think their readers desperately want trends to latch onto, so they invent them wherever they can? Or what?
WAIT! I was using employment levels. The Journal is using nonfarm payroll employment. I had to go out to three decimal places to replicate their findings, but I finally did it:
By this measure, women's employment has been growing faster than men's for four consecutive months. Their share of the total has been 49.8% the whole time, but if you go out to more decimal places it looks like this:
September: 49.778%
October: 49.796%
November: 49.803%
December: 49.813%
January: 49.817%
January was a close call, with women increasing their share of employment by only 0.004%, which is way, way below the survey's standard error of 0.2%, but let's count it anyway. So it's four months in a row after all!
And just in case you think I'm trying to cheat you by snarking about just the past two years, here's the same chart going back to 2015:
Women are nowhere close to the share of employment they had before the pandemic, and by any reasonable measure their share has been below trend and dead flat for the past two years.
I've been wondering for a while what lessons we should take from the COVID pandemic. We've had three years to collect data, after all, and another pandemic is likely to come along someday.
Nobody with real expertise seems to have written about this, so I figure I'll take a crack at it. If nothing else, maybe we'll learn something just from all the different ways that people call me stupid.
Note that this is generic advice. Obviously it will change depending on the viral characteristics of the next pandemic.
Things we should be doing now
Improving ventilation in interior spaces like schools, churches, and so forth.
Installing far UV lamps in places where they'd be effective.
Improving our emergency vaccine production ability.
If a pandemic occurs, things we should emphasize less
Cleaning hands and surfaces. As far as I know, there were virtually no cases of COVID from touching a contaminated surface.
Worrying about the outdoors. With a few isolated exceptions, it's mostly OK to go outside normally with or without a mask.
Closing schools. The bulk of the evidence suggests that with proper safeguards it's safer to keep schools open than to close them.
Things we should do differently/more of
MAGA crackpottery aside, masks work. But we should be stockpiling N95 masks and insisting on them right out of the gate if a new virus starts to spread.
Superspreader events were a major cause of large scale spread, and there's no real way to address this except by shutting down indoor events where lots of people sit near each other. This includes movies, live theaters, sports events, and, I'm sorry to say, churches. The virus doesn't care why you're packed in together, and superspreader events hurt everyone, not just the people who attend them. Needless to say, these rules can be relaxed or made more specific as more becomes known about the new virus.
Encourage outdoor seating at restaurants and, where possible, allow it to happen with a minimum of red tape.
Do everything possible to approve and distribute home testing kits as soon as possible.
Social distancing . . . I'm not sure about. What's the evidence for or against it in places like malls and supermarkets?
It's common these days to measure the impact of the COVID pandemic by looking at total excess deaths as a more accurate measure than deaths attributed directly to COVID. So here are excess death rates for most of the large European countries plus the United States. Do you notice anything interesting?
Here's a similar chart. The countries are in a slightly different order, but the same interesting thing is there:
House Oversight Committee Chair James Comer (R-KY) said Tuesday morning that it was a “mistake” that the administration didn’t go through with bombing drug labs in Mexico after then-President Donald Trump suggested it in 2020.
Okey dokey. It sure is good that the adults are back in charge in the House.
You might be tired of us Californians going on and on about our recent storms, as if a few feet of snow is some kind of otherworldly miracle. And it's true that if I wanted pictures of snowy peaks I could just hop in my car and drive north to Shasta or east to the Rockies and take all the pictures I want.
But I don't feel like doing that right now, and I do have snowy peaks right here at home. So here's a triptych of snowy peaks in the San Gabriel mountains after our megastorm last weekend.
March 1, 2023 — In and around Diamond Bar, California
The early cases of COVID in Wuhan clustered almost entirely around the Huanan Market. That's a helluva coincidence in a city of 11 million if the virus actually originated somewhere else. Note that the virology lab is eight miles away.
