Note: With publication of my book on AI, The God Test, 2.5 weeks away, The Earthling will undergo a temporary change in character. Immersed as I will be in desperate self-promotion, there will be less time for the sustained focus I need to write the longish pieces that have traditionally appeared in this space. So this week I give you a handful of AI updates, followed by some highlights from the recent work of NonZero Network members. We’ll see what next week brings. As for podcasts: They’ll continue more or less as usual, with perhaps a bit more emphasis on AI. Though on balance the content flow won’t be much reduced, we’ll still do what we’ve done in years when we took June entirely off: give paid subscribers a bonus month.
Save the date: The discussion of AI on the NonZero Discord has been picking up lately, and, in support of that momentum, there will be a livestream discussion of AI this Monday, June 8, at 12 pm US Eastern Time (9 am PT), hosted by NonZero’s Nikita Petrov along with Jeremy Eliosoff, head of Pause AI Canada and an active member of the NonZero community. The conversation will range from policy and politics (e.g. Bernie Sanders’s recent proposal to give the public a 50 percent stake in AI companies) to philosophy and spirituality (e.g. the Pope’s recent encyclical on AI) to issues raised by some of the news items below. You can join the livestream discussion at this link (on the Streamyard platform; no need to install any software—it works in the browser). If you haven’t yet checked out the NonZero Discord, you can join here.
There was big news in the AI world this week. “Anthropic Urges Global Pause in AI Development,” read the headline on the Wall Street Journal’s homepage. It wasn’t exactly true news, but it was big.
What Anthropic had actually said was, “If it were possible to effectively slow the development of this technology to give ourselves more time to deal with its immense implications, we think that would likely be a good thing.” But, Anthropic continued, that’s not possible now, because there’s no “global coordination mechanism” to ensure compliance. So “The Anthropic Institute will conduct research—in collaboration with many others—and take actions to help build the systems that a credible slowdown or pause would require.”
So a headline more accurate than the Journal’s headline would have been: “Anthropic says world not ready for AI pause.” Nonetheless, the Journal’s spin was the most common spin on the story.
The company’s ruminations about a global pause came at the end of a long post about how rapidly the pace of Anthropic’s AI research is accelerating as each generation’s most advanced model plays a bigger role in developing the next generation of models—thus moving the process of innovation toward “recursive self-improvement.” Highlighting this theme probably won’t hurt the chances of Anthropic’s coming debut on the stock market being a big success.
The Financial Times reported that “Anthropic is helping the US National Security Agency deploy its powerful Mythos AI model for offensive cyber operations, embedding engineers inside the agency despite an ongoing legal battle with the Pentagon.” Mythos is the not-yet-released model that is famously good at detecting and exploiting security vulnerabilities. FT quoted “a person close to Anthropic” saying that “the best way to build a good defense is to build a good attack…If [Mythos] is not used to build attack agents, adversaries will find a way to do it.”
Anthropic revised its “responsible scaling policy” (RSP) in a way that AI safety advocates said amounted to weakening its safeguards against such misuses of AI as building a bioweapon. This was not the first weakening of the RSP, and it led Dan Hendryks, head of the Center for AI Safety, to issue a kind of “RIP, RSP” post on X: “What a waste of a few years RSPs were.” If it’s any consolation: The heads of Anthropic, OpenAI, and Google’s Deep Mind joined others in the tech community in urging Congress to tighten restrictions on ordering synthetic DNA online.
Wired magazine reported that “Meta has quietly embedded face-recognition technology for its smart glasses into an app downloaded to millions of phones.” Apparently Meta had begun shipping “face-recognition code to users’ phones while publicly describing it as something the company was still ‘thinking through.’”
NZNet Highlights
The NonZero Network is a group of Substack writers and podcasters who have a diversity of ideological perspectives but who all try to make sense of the world in a compelling and honest way. Paid NonZero subscribers can get a 50% discount on a subscription to any NonZero Network newsletter.
At Foreign Exchanges, Derek Davison looks at conflicting reports about negotiations between Iran and the United States and speculates about the thinking in Tehran:
With Trump seemingly unable to make a deal over the past few days there’s been less focus on the extent to which Iranian leaders are interested in an agreement, particularly if getting one would require them to make significant concessions like committing to the elimination of their highly enriched uranium. One of the problems that seems to recur in these deliberations is the fact that the US is demanding that Iran take steps that are not easily reversible while it’s only offering concessions that are easily reversible. That plus a massive trust deficit may be coloring internal Iranian deliberations. However, with Iranian inflation reportedly hitting 77 percent year-on-year last month, a rate not seen since World War II, Iranian officials have an economic incentive to conclude at least a preliminary agreement.
SINICA
On Kaiser Kuo’s Sinica podcast, retired Marine major Eddie Conger talks about why he has built the country’s largest K-12 Mandarin program in Texas, one of America’s least China-friendly states:
From the Marine’s perspective, before we ask our sons and daughters of this country to give their life in a conflict, we should do everything on Earth to make sure that we are communicating with whoever is sitting across the table from us. And we can’t hide from China. It’s 1.4 billion people. It’s the second largest economy in the world. They are going to move forward… What we are trying to do is make sure that we have competent… future leaders to be able to engage with each other. Because if we can do that, I don’t think there’s a problem on Earth that the United States and China, if we actually work together, couldn’t solve.
Asked by conversation partner John McWhorter if he’s feeling the effects of age, Glenn Loury muses:
I might be a little bit slow on the mathematical proof of the theorem in the Econometrica paper, which requires a lot of concentration and short-term memory, in the sense that you have to keep the various derivations and whatnot in mind, and you have to follow the math…But my preferences change as well. I mean, I actually have lost interest in a lot of the technical stuff. Somebody sends me a paper. It’s 40 pages. I read the first three or four pages. I sort of get the idea. And then I encounter this forest of gibberish which has to be deciphered. And I used to dive in with relish to such a task. And now I say: Oh, my God, do I really need to do this? Do I care? And the answer often is I don’t really care, not enough to wade through this thing. So there are changes afoot, that’s for sure. But I’ll let others be the judge about whether or not I’m sharp enough.
On the Global Dispatches podcast, Alexandra Bell, president of the Bulletin of the Atomic Scientists, tells host Mark Leon Goldberg why some of America’s longtime allies are considering getting their own nuclear weapons. The reason, she says, isn’t just that in the age of Trump they are less confident that America’s “nuclear umbrella” would be deployed on their behalf:
Poland, Germany, Japan, South Korea, Canada—multiple countries have talked about this…Partly this is happening because the U.S. is doing things as a government, as an administration right now, to make people doubt their commitment that the U.S. would actually follow through on its extended deterrence guarantees. But that’s not happening in a vacuum. There are countries like Russia, like China, that are expanding their nuclear arsenals, expanding their delivery system capabilities that are making countries feel insecure and like they may need to get their own deterrent because they can’t trust any of these larger nuclear-armed states.
In a video essay, Peter Beinart argues that many Jewish Americans refuse to look at the reality of what’s happening on the ground in Gaza and as a result confuse outrage with antisemitism:
What they do is they look at the people who are enraged at Israel, or they look at Zohran Mamdani, or they look at people who are boycotting, and they don’t understand where this anger comes from. But if they would just pay attention to what Israel is actually doing, they would understand where this anger comes from. They could see it as something other than pathological and antisemitic, and they themselves might actually start to feel that kind of anger themselves.
Paul Bloom asks why publishers and Substack writers who want to boost their revenue aren’t following in the footsteps of baseball legend Billy Beane, who used sabermetrics to remake the Oakland Athletics:
Here’s what I think: People have looked, and there’s nothing interesting there. Book publishing and Substack just don’t work like baseball, and there’s nothing that HarperCollins Billy Beane could tell you that you don’t already know.
On the Glenn Show, John McWhorter tells Glenn Loury why he feels good about a black woman playing Helen of Troy:
I think the idea of extending our imaginations and really getting past race by in some ways, in some contexts, like this one, where Helen of Troy is a fictional character, pretending that somebody with that appearance could have been in that place and time, in that story, as long as she is beautiful, the way Helen of Troy was supposed to be—I think that it’s a healthy thing. It gives black actors more work. I think a lot of people are of the opinion that: How come she can be Helen of Troy, but if you cast Sidney Sweeney in a film of “Raisin in the Sun,” there would be people who would go blow up a building? And as far as I’m concerned, that does need to happen at some point in the future. We have to get to a point where there could be a white “Porgy and Bess,” if we’re really going to get past race.
Banners and graphics by Clark McGillis.
If you can’t join the Monday meetup as a participant but want to watch/listen live, here is the link for that: https://streamyard.com/watch/CgCCuR9tvGxQ