I just finished hearing the CEO of Anthropic say the following on one of today’s “talking heads” programs:

The biggest dilemma is how we get China to agree to join the agreement on regulating the pace of AI development.

but I think he’s wrong. The dilemma that’s bigger than that is how we prevent a country we support that’s engaged in an existential conflict against a larger foe from developing their own LLM AI that breaks all the rules when it means their survival as a country.

That’s also the problem with Megan McArdle’s more modest and seemingly more doable agreement which she describes in her most recent Washington Post column:

Sure, I can imagine China and the United States coming to a rapid-fire agreement for an AI pause. But then, I can also imagine getting a MacArthur “genius grant,” which seems about as likely.

So back to my question about the murderbots. Could we at least not build the literal, physical murderbots, such as fully autonomous drones that can operate without humans in the kill chain? Controlling their development would be a very hard problem, as all arms control is. But it seems more manageable than completely stopping AI advancement.

Consider her proposal in the context of Ukraine. Should we actively discourage even prevent Ukraine from developing their own “murderbots” if they make the difference between their victory or defeat against the Russians?

The bigger dilemma may be how you construct an AI arms-control regime that remains rational to obey when one of its members believes violating the rules is necessary for national survival.

There’s an even more elementary problem. Suppose Anthropic, OpenAI, xAI, and Google agree on rules limiting what their LLM AIs may do. What prevents a well-financed startup from developing an LLM AI that ignores those rules?

If the prohibited capabilities are useful, that company’s product may be more competitive precisely because it breaks the rules. The agreement would then have accomplished something rather peculiar: it would have imposed competitive disadvantages on the companies that agreed to it while creating an opportunity for companies that did not.

Preventing that requires something considerably more ambitious than an agreement among AI companies—or even an agreement between the United States and China. It requires preventing *anyone capable of developing the technology* from defecting when there is an economic or strategic advantage in doing so.

Now consider Ukraine. If we cannot be confident that a startup will refrain from breaking the rules for money, why should we expect a country to refrain from breaking them when its leaders believe national survival is at stake?

When we woke the morning after leaving Ketchikan, the ship was already docked in Juneau. After showering and breakfasting we left the ship and spent the rest of the morning sightseeing.

Neither my wife nor I were particularly interested in the tourist-y downtown area so we immediately headed off to the Tongass National Forest. We caught our first glimpse of a glacier while walking in the Tongass National Forest near Juneau. You can see it in the photo at the top of the post.

We continued walking until we reached a very scenic waterfall.

Following that we returned to the ship, got lunch, and rested for the remainder of the day.

James Joyner has a fine round-up of posts and articles relating to the commemoration of the anniversary of the terrorist attacks on September 11, 2001. I’m going to devote this post to a few ramblings of my own on that subject.

There have been many changes as a consequence of 9/11. Many of them are institutional; fewer have improved our actual security. Key issues include

• American military presence in the Middle East
• Islamist terrorism
• Intelligence
• Airport security
• Surveillance
• Federal bureaucracy
• War powers
• Fiscal consequences
• National unity

The reality is that we solved the specific problem that occurred on September 11 rather better than we solved the general problem it exposed. It is now extraordinarily difficult to hijack four airliners simultaneously and use them as guided missiles. Cockpit doors, passenger behavior, intelligence and screening make a literal replay of the events of 9/11 unlikely. The absence of another comparable attack over 25 years shouldn’t simply be waved away as luck.

It still remains the case that a small, determined organization could exploit the enormous complexity and openness of modern American society, while American institutions had great difficulty recognizing a threat that did not fit their existing categories. That problem hasn’t gone anywhere.

American military presence in the Middle East

Before the September 11, 2001 the U. S. is estimated to have around 25,000 troops in the Middle East; now it has double that. Our presence in the Middle East is now more greatly institutionalized and dispersed.

Islamist terrorism

After 2001 roughly a quarter million people have been killed in Islamist terrorist attacks.

Intelligence

Our intelligence gathering has been reorganized greatly over the last 25 years.

Airport security

Airport security and screening is much more visible and intrusive than it was 25 years ago. Although around $4 billion is collected per year in federal security fees associated with each departure, we spend an additional $8 billion per year on airport security and screening.

Surveillance

One of the greatest changes on the domestic scene over the last 25 years is in the area of surveillance of people in the U. S. Authorities and institutional arrangements created or expanded after 9/11 have become normal parts of American government while technology has made surveillance vastly easier.

Federal bureaucracy

The creation of the Department of Homeland Security following 9/11 consolidated 22 federal agencies and departments. GAO has documented persistent problems integrating their acquisition, financial, information technology, and personnel systems. That is broadly consistent with what I predicted at the time: consolidation would add another bureaucratic layer rather than actually simplify the federal government’s security apparatus.

War powers

The post-9/11 period greatly normalized presidential military action without anything resembling the notional congressional declaration-of-war process.

Fiscal consequences

Federal post-9/11 war expenditures and obligations have been about $8 trillion, including military operations, homeland security, veterans’ care and other costs.

National unity

The increased national unity in the immediate aftermath of the attacks was fleeting. The lamented blog American Future documented that the New York Times started running editorials against George W. Bush about six weeks after the attacks.

I think a lot of the actions of the TSA are more theater than producing security. For two years I unknowingly carried an X-Acto knife in my computer bag while flying. I didn’t discover it because airport security found it. I discovered it myself. During those two years the bag had repeatedly gone through airport screening. If a security system designed in response to 9/11 could repeatedly miss a boxcutter carried by someone who wasn’t even trying to conceal it, exactly what did all of that screening accomplish?

We changed our institutions much more than we changed ourselves.

When I read Francis Fukuyama’s op-ed in the Washington Post advocating what he calls a “realism of means” one sentence took me aback:

The United States has always seen the promotion of liberal democracy as part of its national identity.

because it is objectively untrue. Since the Founding Americans have regarded our own political system as liberal and republican/democratic and have often regarded that system as an example for other peoples. Actively promoting liberal democracy abroad as an objective of American foreign policy has largely been a feature of post-World War II American views. Woodrow Wilson advocated it but was unable to establish his conception of America’s international role as a durable national consensus. John Quincy Adams’s 1821 formulation that America “goes not abroad, in search of monsters to destroy” captures the distinction almost perfectly.

There is another weakness in Dr. Fukuyama’s “realism of means”: he proposes no limiting principle. He provides no mechanism for preventing his “realism of means” coupled with idealism of ends resulting in overextending our means to accomplish those ends. Dr. Fukuyama mistakes one American political tradition for the American national identity. More importantly, his “realism of means” contains no principle for deciding when desirable ends cease to be American interests. Realism can certainly tell us that our means are inadequate to an end. But unless it is permitted to constrain our ends as well, it cannot prevent the accumulation of objectives that eventually exceeds our means. John Quincy Adams understood that problem two centuries ago: America could wish liberty well everywhere without assuming responsibility for securing it everywhere.

I also wanted to comment on Karl Rove’s most recent Wall Street Journal op-ed. In the op-ed Mr. Rove assesses the state of the midterm elections as they enter their final phase. Mr. Rove opens by characterizing the Democratic candidates’ overall strategy:

Democrats appear to believe it’s sufficient simply to blame President Trump for the economy and offer hazy promises to do something about housing, grocery, healthcare and gasoline prices. No need to explain what Democrats would do, how it would work, or what it would cost. Just trash Mr. Trump and chant “affordability.”

citing the rather small advantage Democrats seem to be showing in the opinion polls, and continuing with the Republicans’ overall strategy:

Still, the Republican campaign strategy doesn’t look much better. It’s good news that Mr. Trump finally announced Friday that he’ll open his record $1 billion campaign war chest. He says he’ll spend up to $500 million from his political-action committees now and keep the balance, perhaps for the 2028 election. But it’s already late and could prove too little.

This close to Election Day, advertising airtime sells at a premium. While candidates and party committees by law can buy ads at the lowest unit rate a TV station charges for any nonpolitical client, a PAC has to buy what the market will bear. If Mr. Trump’s MAGA Inc. plans to spend $10 million on the Texas Senate race across the next two weeks—as its Federal Election Commission filings indicate—Mr. Talarico can probably match it by spending $4 million to $6 million.

It all reminds me of the late Mayor Daley’s wisecrack, “No matter how it looks now someone will win the election.”

I have no idea whether the Democrats will take control of the House or the Republicans will retain it. The Polymarket and Kalshi prediction markets presently favor Democratic control of the House, but the actual shift in seats they imply is comparatively modest. That would hardly be an unusual result for a midterm election.

In reaction to President Trump’s proposal that Americans receive a $5,000 “dividend” I think an actual dividend is the last thing we should want. In general in a company dividends are paid from profits, i.e. revenue – expenses. The U. S. government is presently running at a loss, i.e. we are “borrowing” to pay our bills.

In the instances in which companies running at losses (expenses exceed revenues) pay dividends it is generally borrowed and believed to be a temporary move. Over the last century the federal government has operated at a surplus in 17 nonconsecutive of those years. Five of those were in the 1920s so that’s hardly a temporary condition.

But the bad news gets worse. For decades the government’s commonly cited consolidated deficit included Social Security’s cash flow. That mattered particularly after the 1983 reforms, when Social Security accumulated substantial surpluses: those surpluses made the government’s overall fiscal position look better even while the rest of the federal government continued to run deficits. Look instead at the government’s on-budget balance excluding Social Security and the record is considerably worse. The federal government has been financing its ordinary operations with borrowing for almost the entire modern history of Social Security.

The irony is that a $5,000 “affordability dividend” could actually make the affordability problem worse. Borrowing more than a trillion dollars and distributing it as cash does not produce another trillion dollars’ worth of houses, food, automobiles, medical care, or other goods and services. It increases the number of dollars available to bid for the same supply of them. To the extent that additional demand outruns additional production, the result is higher prices. And to the extent that the additional federal borrowing raises interest rates, financed purchases become more expensive as well.

I wanted to take note of a single sentence in Walter Russell Mead’s latest Wall Street Journal column:

The alleged Russian sabotage attempt on Germany’s Leipzig/Halle airport failed to trigger a rush of Euro-patriotism.

because it reflects a widespread misconception.

There has never been “Euro-patriotism”. There have been France, Germany, Italy, etc. pursuing their own parochial interests. France has pursued subsidies for French farmers. Germany has pursued the profits of German banks and manufacturers. Other less developed economies e.g. Greece have pursued the increased consumption that unlimited borrowing enabled. And so on.

There has never been “Euro-patriotism” and I see little reason to expect it to emerge now that establishment political parties are fighting for their political lives against “far right-wing” upstarts. Don’t infer European solidarity from coincident national interests. European unity exists when countries continue acting together after their national interests diverge and doing so becomes costly.

The first stop in our cruise was in Ketchikan, Alaska, a small town of 8,000 people at the southernmost tip of the Inside Passage coastal route for ships and boats. Historically, it was a fishing community but today it is largely devoted to tourism and government services.

We weren’t interested in the very tourist-y downtown area and immediately made our way to the “fish ladder”, the rapids used by salmon to jump their way upstream to spawn, pictured below.

We must have seen a dozen salmon leaping the rapids in the few minutes we spent watching the fish ladder. We then made our way from the fish ladder to the beginning of our guided nature walk in the Tongass National Forest, the largest intact temperate rainforest in the world. The walk was one of the highlights of our trip.

Our very well-informed trail guide introduced us to the flora and fauna of the rainforest in the course of a roughly two mile walk. Sitka Spruce. Western Hemlock. Various berries. Devil’s Club, a spiny bush. Much to our delight we encountered four black bears along the trail maintaining a safe distance. We were told that we may have seen more bears on our nature walk than those who took the bear walk.

It rains quite a bit in Ketchikan—to the extent that they measure rainfall in feet per year not inches. 12 or more feet of rain per year is typical. Apparently, we brought good weather with us. The temperature varied from the 50s to the 70s in the course of the morning and it was one of Ketchikan’s few rainfree days.

Our visit to Ketchikan was quite brief, just the morning. After our nature walk we caught a bus at the Totem Heritage Center (Ketchikan has the largest collection of standing totem poles) and returned to the ship which departed around noon for Juneau.

Can anyone envision an end condition for the war in Iran other than the end of President Trump’s term? I can’t. I think that the war in Iran is likely to continue as long as anyone in Iran has drones or missiles and as long as there are targets within striking range which may be a very long time, indeed. Is there one person with the practical authority to say, “The war is over; cease firing,” and have every organization and commander capable of launching a missile or drone actually stop? I don’t believe there is.

Iran in 2026 is not Japan in August 1945.