Skip to content

From the Wall Street Journal:

OpenAI’s new artificial-intelligence project is behind schedule and running up huge bills. It isn’t clear when—or if—it’ll work. There may not be enough data in the world to make it smart enough. The project, officially called GPT-5 and code-named Orion, has been in the works for more than 18 months....

I get that competition is stiff in the AI biz and that vast sums of money are involved. But it's a sign of how sky-high our expectations have gotten that taking 18 months for a major upgrade is considered something of a crisis. Hell, routine major releases of Windows take twice that long.

I suspect we should temper our optimism a little bit anyway. My super-duper-oversimplified history of AI goes like this:

  1. Neural networks
  2. Deep learning
  3. Transformers
  4. ???

I figure we're still one major innovation away from true AI. There's no telling when we'll get that, and in the meantime AI will make spectacular progress but won't quite make it to the promised land of AGI. That's still a little ways away.

According to a new report commissioned by Waymo and conducted by SwissRe, Waymo driverless cars are about ten times safer than cars driven by people:

The report is based on 25.3 million miles driven by Waymo cars in four US cities (San Francisco, Phoenix, Los Angeles, and Austin). This was before Waymo began freeway driving, which may affect the results. Check back next year for more on that.

Hardworking Iowa senator Joni Ernst delivers a bombshell:

ZOMG! And this is the result of "investigations," so you know it must be serious. I decided this demanded an investigation of my own, and after literally minutes of work I got the goods:

As it turns out, the Department of Energy reported 15,365 employees last May. Of those, 3,171 are remote field workers without offices. The others averaged 60% of their hours in person. This compares to an estimated 67% of in-person hours for US senators.

Merry Christmas.

Derek Thompson writes in the Atlantic about a mysterious new "health wave" sweeping the United States:

In May 2024, the U.S. government reported that drug-overdose deaths fell 3 percent from 2022 to 2023, a rare bright spot in a century of escalating drug deaths. In June, the National Highway Traffic Safety Administration reported that traffic fatalities continued to decline after a huge rise in 2020 and 2021—and that this happened despite a rise in total vehicle miles traveled. In September, the U.S. government announced that the adult-obesity rate had declined in its most recent count, which ended in August 2023. Also in September, FBI analysis confirmed a double-digit decline in the national murder rate.

This is maybe a little overstated? Three of these things aren't health-related, per se, and both murder and traffic fatalities are obviously just reversions to the mean after a brief pandemic spike.

The only true health-related metric here is obesity, and the news here is a little more ambiguous than Thompson lets on:

The CDC's obesity data is a little fuzzy right now, but it shows a couple of things.¹ First, obesity is down, and has been since 2018. But severe obesity continues to rise, reaching a record 9.7% last year. This doesn't suggest some kind of health epidemic is at work.

For what it's worth, it also doesn't suggest that the obesity decline is likely the work of Ozempic and its ilk. The decline began around 2018, before the Ozempic craze was widespread, and it apparently hasn't affected severe obesity, which you'd expect to be the thing it affected the most. So the explanation probably lies elsewhere.

I dunno. Is it possible that we lost weight during the pandemic? I would have thought the opposite, with so many workers hanging around at home where snacks are everywhere. But who knows? Maybe we all just worried the weight off.

¹Older CDC data showed an obesity rate for 2017-18, but that's missing from the newer data, which has only 2017-2020. I'm not sure why. The data is here and here if you're interested.

If it's inflation day that means it's also personal income day. Adjusted for inflation, disposable personal income increased at an annual rate of 1.3% in November:

After accounting for inflation, disposable personal income is up 10% since the start of the pandemic.

Core PCE is The Fed's Preferred Measure of Inflation™, so let's focus on that this month. On a month-over-month basis, it came in at 1.4% in November:

PCE is a volatile series, so it's best not to get too excited about monthly ups and downs, including this one. Nonetheless, core PCE has been on a steep trend downward ever since it peaked in 2021 and there's no reason to think that's going to change. It remains a little elevated thanks to inflation in services, which is mostly caused by catch-up wage growth, but that's natural and won't last much longer. The Fed needs to stop obsessing over this.

On a conventional year-over-year basis, headline PCE clocked in at 2.4% and core PCE came in at 2.8%.

Today the Border Patrol reported a total of 94,190 Southwest border encounters in November. This was the first time the number has dropped below six figures since Joe Biden took office.

Out of the total, about 47,000 were caught trying to cross the border illegally. The rest were asylum seekers who presented themselves at border stations or gave themselves up after crossing between stations.

I forgot to mention that I'm out of the hospital and back home. In fact, I came home yesterday. I was scheduled to stay until today, but I made a nuisance of myself and bullied them into releasing me early.

Why? This was my third time getting a treatment with a risk of causing Cytokine Release Syndrome and I've never suffered even a hint of it. I just seem to be one of the lucky ones. This time around, there were exactly zero problems. No side effects; no fever; excellent counts and stable vitals the entire time; and I felt great. So I came home. The odds of a last-second collapse seemed minuscule and I was tired of twiddling my thumbs in the hospital bed.

All that's left now is to see if the Tecvayli works. That is, I'm sure it will work, but the question is how long it works. A key date is roughly August 2026, when I'll be two years out from my prostate cancer treatment. That means I'll once again be eligible for clinical test treatments if there are any promising ones around. It will be a few months before we have a better handle on all this.

Our story so far:

  • Democrats negotiate with Speaker Mike Johnson on a CR to keep the government open for another three months.
  • After a bit of minor pressure from Elon Musk, Johnson reneges on the deal.
  • The new deal is: Fuck you. We get everything we want, you get nothing.

Now, I'm no Master of the Deal™, like our president-elect, but no Democrat in their right mind could possibly vote for this. It's an open admission that you'll roll over for anything. And sure enough, only two did. It didn't even get close to a majority, let alone the two-thirds it needed to suspend the rules.

So what happens next? Beats me. If Mike Johnson had any spine at all, he would have ignored Musk, announced that a deal is a deal, and put the original deal to a vote. It might have gone down depending on how many Republicans rejected it, but at least it would have been an honest effort. But that's not what he did.

At this point, it's hard to see Democrats voting for anything remotely like the current bill, and with Trump now fully bought into the "fuck you" version of the bill it's hard to see how Republicans could ever go back and vote for the original deal. I suppose that negotiations start all over again and the government will probably shut down for a while. Perhaps it will still be shut down during Trump's inauguration. That would be something.