This criticism, like most criticisms of AI, addresses the wrong part of the problem. The problem is harm to other parties. Typically customers and employees of corporations.
A solution is strict legal liability. Corporations must be strictly liable for harms. That liability should be higher when AI is involved. Such liability may not be waived by contract, forced into arbitration, or devolved upon a third party service.
Then we let the plaintiff's bar and the insurance adjusters price that risk.
Something like this turns on in the EU in October 2026.[1]
The books exist, but those books are read by few outside military leadership.
The good ones come from leaders who have been there, and they skip over the basics. Being able to talk to a chatbot that can get people past the classic military mistakes could make a force far more effective. Maybe. It's not going to create a Giap or a Rommel, but it might keep a force from repeating classic military mistakes.
There's a brutal little book, "The Defense of Duffer's Drift".[1] It's about classic small-unit mistakes, written from the point of view of someone who has to dream about the same battle over and over. Each time, his plan seems reasonable. But he loses and gets killed. He finally gets it right, after quite a few tries. It's to hammer home the message that there are many ways to screw up an operation. If you don't know the classic mistakes, you're going to make them.
A more modern critic is The Angry Staff Officer. This is a currently serving US Army officer who writes, with a biting wit, about tactics, both real and fictional.[2] He's a good read.
The classics for revolutionaries/terrorists are, of course, Mao and Marighella. Mao is philosophical. Marighella is nuts and bolts.
Poland and the Baltics know they're next on Putin's list. So they're glad to help Ukraine. Better than fighting the Russians on home territory.
Taiwan and Japan, though, aren't in Russia's line of fire. The cooperation with Ukraine is because Ukraine figured out how to stop Russia. Taiwan has a real threat, and if they can build defenses that mean China faces a years-long war, there probably won't be one.
"The guerrilla must move amongst the people as a fish swims in the sea" - Mao
Guerrilla forces have to be popular, or at minimum, have the local population cowed. They're useful for kicking an oppressor out of your own country, but not for conquest.
Russia does not have anywhere near enough aircraft to do that. If they tried, they'd lose more of their air force.[1] Russia does not have much of an aircraft industry left. For one thing, much of the USSR's aircraft industry was in Ukraine.
- Why not fire off a litany of missiles?
Nuclear? Russia so far has not wanted to cross that threshold and start WWIII.
Non-nuclear, they have to build them.
- Why not conscript the bulk of their manpower reserves and actually invade in earnest?
Because Russia is running out of manpower reserves. The draft in Russia now covers all males from 18 to 30. In that age range, men can no longer leave Russia. Russia has 1.4 million military casualties so far, with about 400,000 deaths. Currently, the casualty rate is higher than the new recruit rate.
Conscript soldiers serve for one year, but there is heavy pressure to sign up as a "volunteer" with no time limit.
The Russian tradition of throwing recruits into battle with a huge casualty rate continues.
It worked in WWII, but cost 20 million lives. It hasn't been working in Ukraine. The Ukrainian claim is that the survival time for a new soldier on the Russian front lines is four hours.
This is probably exaggerated.
It was policy for a while that the draft wasn't enforced too strictly in Moscow and St. Petersburg. This was to avoid generating political opposition to the war. That seems to be over.
The main point of this paper is that rear area bases are too vulnerable now.
This paper is from West Point, and is the view from the Army side.
It's a big problem for air forces, too.
The US has a tradition of large, supposedly secure bases in the rear. The USAF has relied on this since WWII. It's not working any more. Iran has been hitting US air bases. In the last day, they've hit US bases in Qatar, Jordan, Kuwait and Bahrain. That's just this round. There were previous hits a few months back. It's not publicized much, but it's no secret. It's hard to protect an air base against drones. Air bases are big, everyone knows where they are, airplanes are hard to hide from drones, and are vulnerable to small explosive charges. As Russia keeps finding out as Ukraine hits their air bases. There have been hits well inside Russia. Blast-resistant hangars in Crimea have been attacked. Don't leave a door open.
China is going in for airbase hardening in a big way.[1] This is sometimes called a "concrete sky" program. The USAF is way behind in the Pacific. Too many planes parked out in the open, or in weak hangars.
Active defenses against drones and missiles do work, but they just thin them out. Some get through.
If the attacker has enough manufacturing capacity, the defenses can be overwhelmed. Ukraine is currently building about 7 million drones a year.
Look at 1950s aircraft. That was the decade of really weird aircraft, as people figured out how jets were supposed to work. Supersonics. VTOL aircraft. The X-planes. Rocket-assisted takeoff. After that, more was known about what worked, and designs became more similar.
Ukraine and Taiwan quietly cooperate in weapons development and production, especially drones.[1] They both have big, aggressive neighbors. Ukraine knows how to fight them, and Taiwan can make electronics in quantity. Ukraine is starting to get cooperation from Japan, too, but that's in an earlier stage.[2] With Taiwan, it's serious.
The paper doesn't mention Ukraine at all, yet it's all about the kind of war that Ukraine is fighting.
This isn't about radioactivity at all. It's about the millions of pounds of mercury used at the Y-12 plant at Oak Ridge,[1] resulting in a lot of low-level mercury contamination.
For some problems, yes. Formal specification is particularly useful in two cases. 1) The problem is simple but an efficient implementation is hard or bug-prone. Examples are garbage collection, file systems, sorts, databases, and tree updating.
2) The inverse of the problem is simpler than the forward operation. Examples include matrix inversion and parsing.
Dumbing down of science museums happened decades ago. The Smithsonian, the Henry Ford Museum, and the Science Museum in London were made more "accessible". They used to focus on historically significant artifacts, without much explanation. Then they switched to heavy explanation of the basics.
The Science Museum in London still had Maudslay's lathe on display. Maudslay's lathe is the direct ancestor of all modern lathes, with its slide rest and the leadscrew drive. No previous lathe was like that. So it's one of those rare artifacts that changed the world. People walk by it to get to see James Bond's car. That was the strength of that museum - they had some of the first examples of key advances.
The Henry Ford Museum used to have really obscure items. "Capacitor, Cornell-Dublier, 1938". A combination television camera and kinescope projector for rotating disk television. Then they added explanatory exhibits and pushed aside the small obscure stuff.
That was probably for the best; it was like a really well funded antique shop.
The kids just want to see the locomotives, anyway.
The Smithsonian was once called "The Nation's Attic". Now, most of the obscure stuff has been moved out to storage in favor of more understandable exhibits. They used to have a history of clock escapements, with a working model of each type, kept wound and running. No more. The Arts and Industries Building had even older stuff. Their stamp collection had examples of all US stamps.
www.animats.com