Month: December 2025

Posted on

Interesting articles, December 2025

Putin still thinks he can take over all of Ukraine, and once he’s done, he wants to start conquering other countries that used to be part of the USSR.
https://www.reuters.com/world/europe/us-intelligence-indicates-putins-war-aims-ukraine-are-unchanged-2025-12-19/

In the first attack of its kind, Ukraine used an autonomous suicide sub to blow up a Russian warship.
https://www.kyivpost.com/post/66281

The U.S. saber-rattling campaign to pressure Venezuela’s socialist government to quit is ramping up:

“The Pentagon’s own manual on the law of armed conflict points out that it is ‘dishonourable and inhumane’ to attack people incapacitated by shipwreck. Killing the survivors of an attack at sea is ‘literally the textbook definition of an unlawful order’, says Steven Lepper, who served as the air force’s second-highest-ranking lawyer.”
https://www.economist.com/united-states/2025/12/03/will-congress-rein-in-pete-hegseth-and-his-boat-bombing-campaign

“We’ve just seized a tanker on the coast of Venezuela, a large tanker, very large, largest one ever seized, actually.”
https://apnews.com/article/trump-tanker-seized-venezuela-maduro-0a148ba01684fc6ce1a228dd276732c0

U.S. jets meant to penetrate enemy airspace and scramble its defense radars have been deployed to Puerto Rico, within range of Venezuela.
https://www.twz.com/air/navy-ea-18g-growlers-have-touched-down-in-puerto-rico

‘China says US seizure of ships ‘serious violation’ of international law’
The roles will perfectly reverse if China ever tries invading Taiwan.
https://www.straitstimes.com/world/china-says-us-seizure-of-ships-serious-violation-of-international-law

Speaking of which, China started massive naval exercises around Taiwan simulating a blockade.
https://www.nbcnews.com/world/asia/china-military-exercises-pla-taiwan-blockade-trump-xi-justice-mission-rcna251464

With the cancellation of the latest U.S. warship program, China is now better and faster at building warships. Decades of missteps on our part led to this.
https://youtu.be/UBDfDvygFNg?si=OjghoUt0NxHiwSLS

There’s no task a battleship can do that a smaller warship or bomber can’t do more cheaply. Reactivating our old battleships would be a bad idea since they’d be prestige targets, and the loss of one of them would be a big blow to national pride.
https://youtu.be/KznZKdSkUbU?si=mMERi2HByCpfh1q7

Russia is using merchant ships to spy on Europe and to deploy drones over foreign military bases.
https://www.digitaldigging.org/p/they-droned-back

Though it was a technological marvel, the V-2 rocket was a gross misallocation of Germany’s wartime resources. The country would have fared better had it used the V-2 funds to build more boring synthetic fuel factories.
https://youtu.be/en9MBSmGWWo?si=NGSIMiaxXDnv4KlI

During the Korean War, the communists captured two U.S. F-86 Sabre fighter planes.
https://theaviationgeekclub.com/heres-how-soviets-were-able-to-steal-and-transport-an-f-86-sabre-to-moscow-during-the-korean-war/

“The Architects of AI Are TIME’s 2025 Person of the Year”
https://time.com/7339685/person-of-the-year-2025-ai-architects/

Google is now making large numbers of computer chips, reducing its reliance on NVIDIA. However, the latter still retains large advantages that will ensure its market dominance for the foreseeable future.
https://www.economist.com/business/2025/11/25/google-has-pierced-nvidias-aura-of-invulnerability

I agree with this: “Models keep getting more impressive at the rate the short timelines people predict, but more useful at the rate the long timelines people predict.”
https://www.dwarkesh.com/p/thoughts-on-ai-progress-dec-2025

After several months of slow AI progress, two of the authors of this year’s “AI 2027” report have pushed back their key deadlines by 1-2 years. Will the pattern hold?
https://blog.ai-futures.org/p/ai-futures-model-dec-2025-update

‘But it turns out that in [Winograd Schema Challenge], there are statistical associations that offer clues. Consider the example above. Large language models, trained on huge amounts of text, would have encountered many more examples of a roof being repaired than a tree being repaired. A model might select the statistically more likely word among the two options rather than rely on any kind of commonsense reasoning.’
https://www.sciencenews.org/article/ai-understanding-reasoning-skill-assess

Tesla’s Optimus robot can run almost as smoothly as a human…
https://youtu.be/N9LuzalIVGg?si=xmq1udI1TBJsS-4R

…BUT: “A new video surfacing from a Tesla demonstration in Miami this weekend shows the Optimus humanoid robot taking a nasty fall. But it’s not the fall itself that is raising eyebrows, it’s the specific hand movements the robot made on its way down, which strongly suggest it was mimicking a remote operator frantically removing a VR headset.”
https://electrek.co/2025/12/07/tesla-optimus-robot-takes-suspicious-tumble-in-new-demo/

The head of a Chinese robot company let his premier model, the “T800”, kick him down.
https://youtu.be/8UrE71ERCYY?si=zBlX2FD-kBiLC6Dk

‘Roomba vacuum cleaner firm files for bankruptcy’
https://www.bbc.com/news/articles/c1lr75lp239o

Russia has enormous mineral and fossil fuel resources, but most of it is in the inaccessible wastes of Siberia or the Far East. The harsh climate makes it impossible to incent enough workers to go there to extract the resources, and Soviet-era attempts to move humans there at gunpoint were expensive failures in the long run.

Russia’s luck will change once intelligent robots exist. They will meet all the country’s labor needs, including in regions unsuited for human life. Without jobs, Russians living in far northern places like Norilsk (a city built to exploit a huge nickel mine) will have no reason to stay and will move to more clement parts of the country, accelerating a long-term depopulation trend.
https://www.brookings.edu/books/the-siberian-curse/

Autonomous cars will kill fewer human pedestrians…and animals.
https://reason.com/2025/11/28/self-driving-cars-will-make-the-world-safer-for-humans-and-cats/

Even the most advanced autonomous taxis still cost $7-9 per mile, whereas human driven taxis cost $2-3. We’re a ways from the cost disparity closing and autonomy being a profitable business.
https://www.economist.com/business/2025/11/24/the-self-driving-taxi-revolution-begins-at-last

‘Without working traffic lights, the driverless cars were seemingly left confused, with many halting in their tracks and causing major traffic jams.’
https://www.independent.co.uk/news/world/americas/waymo-suspended-san-francisco-traffic-jams-blackout-b2888562.html

‘Tornyol grinds mosquitoes in its propellers.’
https://tornyol.com/

I said this years ago: worrying about AI stealing all low paying jobs overlooks the fact that the highest paying jobs, including CEOs and billionaire investors, will also be stolen. An Elysium world where the rich are insulated from the effects of job automation and AI aggregation of wealth will only be temporary.
https://philosophybear.substack.com/p/capitalist-strategy-when-capital

Three guys who made fortunes shorting the stock market can’t agree if an AI stock bubble exists today. They DO agree that, if it does, timing the bursting of the bubble will be impossible. wish the advice were more specific. For sure the old adage that you should ‘Only invest what you’re willing to lose’ is still true.
https://www.economist.com/finance-and-economics/2025/12/16/where-americas-most-prominent-short-sellers-are-placing-their-bets

Here are some interesting energy industry predictions.
https://caseyhandmer.wordpress.com/2025/12/08/energy-predictions-2025/

The electrification of the car fleet is hitting snags, including Ford canceling its electric F-150 pickup truck due to weak sales.
https://apnews.com/article/ford-electric-vehicles-trump-f150-a1fcdec9c76cde5d2d6852360d9d42c4

‘EU waters down plans to end new petrol and diesel car sales by 2035’
https://www.bbc.com/news/articles/crk78y7k8ezo

China’s space rocket technology remains at least ten years behind America’s, but they could close much of the gap as they have in other areas.
https://www.economist.com/science-and-technology/2025/12/03/the-chinese-rocket-industry-takes-off

Update: Aliens didn’t invade us.

‘The more likely scenario from an engineering perspective involves a mothership that releases mini-probes which perform a reverse Oberth maneuver to slow down at perihelion and intercept Earth,’ Loeb wrote.

This type of maneuver uses the sun’s gravitational pull at the object’s closest approach to adjust the trajectory efficiently, enabling the probes to reach Earth without large amounts of fuel.

According to [Avi Loeb], these hypothetical probes could arrive between November 21 and December 5, 2025…’
https://archive.is/M0Lci

Another “AI” drug discovery company goes bust.
https://www.science.org/content/blog-post/where-are-they-now-verge-genomics

These newborn black-footed ferrets are clones of long-dead individuals. Will humans do the same someday for our species?
https://www.washingtonpost.com/dc-md-va/2025/12/29/ferret-clone-smithsonian-front-royal/

It costs a national average of $10,000 per month to stay in a nursing home. Android caretakers will someday decimate the industry, allowing old people to keep living in their homes until they die with a high standard of care.
https://www.seniorliving.org/nursing-homes/costs/

Many autistic people have the same, odd walk.
https://www.dailymail.co.uk/sciencetech/article-15419529/Scientists-shape-butt-ADHD-autism.html

Posted on

You’re an idiot (but so am I)

[Written with the help of GPT-5]

I’ve been taking a martial art for several months now, and as of late I’ve been finding the classes a very humbling experience. I’ve taken enough lessons to be eligible for a new belt, but to get it, I must demonstrate to the instructor a mastery of all the moves I’ve learned up to this point. Every time I go to the studio to practice with other students at my same level, I’m embarrassed by how much I’ve forgotten: my ability to properly execute moves that I’ve performed dozens of times is inconsistent and often deficient. The belt feels a hundred miles away.

It’s a stark, immediate reminder of how difficult it is for me to learn anything and to retain it reliably over time. In short, the martial arts classes make me feel like an idiot.

But it’s not just me. On the rare occasions when I watch team sports, almost every play is littered with mistakes—some trivial, some catastrophic. Even the world’s best athletes constantly get it wrong, despite practicing full-time for years. In intellectual domains it’s no different: a world-class surgeon will sometimes maim or kill a patient through error; a world-class lawyer will look back regretfully on cases lost because of a missed objection or forgotten precedent; and world-class pilots with decades of experience routinely die in crashes caused by elementary mistakes.

Like fish unaware of the water they swim in, humans are largely blind to how cognitively constrained we are. We take it for granted that learning requires enormous repetition, that skills decay quickly without use, and that mistakes are constant even in tasks we know well. I explored this idea previously in my essay “The Extraordinary Inefficiency of Humans,” arguing that slow learning, forgetting, and error are defining features of individual human life. But the same description applies at a much larger scale—to the history of our species itself.

The Homo sapiens species arose about 200,000 years ago, yet for the vast majority of that time we lived little better than animals, and technological and cultural progress was so slow that few individuals would have noticed any change within their own lifetimes. The bow and arrow—one of the first true force-multiplying weapons—was not invented until roughly 70,000 years ago. Agriculture, which fundamentally transformed human settlement patterns, social organization, and population density, did not arise until about 12,000 years ago.

Until then, and for millennia afterward, nearly all tools and weapons were made of wood, stone, bone, and rope. Only around 7,000 years ago did humans learn to smelt and refashion copper, and even then its softness limited its usefulness. Another two thousand years passed before the discovery of bronze, the first truly durable metal for tools and arms. Iron—so central to warfare, infrastructure, and large-scale production—did not become widespread until roughly 3,000 years ago.

Writing, the foundation of recorded history, law, and bureaucracy, emerged a mere 5,400 years ago, with coinage appearing even later, around 2,600 years ago. For nearly 95 percent of our species’ existence, humans lived without cities, states, written language, or complex machines—underscoring how extraordinarily slow technological and cultural progress was until the Industrial Revolution.

Furthermore, over the course of our history there have been many instances of regression, such as the Dark Ages, when hard-earned knowledge was forgotten. Less dramatically but more persistently, innumerable empires and nations have fallen by repeating avoidable mistakes—forgetting the lessons of their own histories and relearning them the hard way.

In short, we are not just idiots as individuals; we are idiots as a collective.

But, as hard as it may be, at least we’re capable of learning anything at all. Our species possesses a qualitatively different kind of cognition from all other known animals: the ability to engage in abstract reasoning, symbolic thought, and cumulative culture. Other “intelligent” animals—such as apes, dolphins, or corvids—can learn some impressive skills, but repeated scientific efforts have failed to demonstrate open-ended, generative language, numerical abstraction, or symbolic reasoning in nonhuman species. Their intelligence is sharply bounded.

Human intelligence, by contrast, is general—but it is also slow, fragile, and expensive. It takes years to train, constant reinforcement to maintain, and enormous social scaffolding to preserve. We can reason abstractly, but only in short, focused bursts, and only with great effort. The long delays between the major technological breakthroughs listed earlier are not historical accidents; they are evidence of how difficult it is for human cognition, unaided, to push beyond its own limits.

Nearly all significant advances in human history have occurred when we offloaded cognition onto external systems: writing to preserve memory, institutions to stabilize knowledge, machines to amplify physical effort, and later computers to extend calculation. From this perspective, artificial general intelligence is not a rupture in human history but its logical continuation—the ultimate cognitive prosthesis.

We will be at a colossal disadvantage relative to intelligent machines. They will share our capacity for general learning and abstraction—dialed up to levels comparable to human geniuses—but without being hamstrung by slow learning speeds, forgetfulness, emotional reasoning, cognitive biases, or biological necessities like sleep and sustenance. They will be continuous, tireless thinking systems: the next step in the evolution of life.

That next step, however, will only occur because we allow it to. Humans will be the creators of AGI, just as Australopithecus gave rise to us. Our species should see itself as a bridge between non-intelligent life and superintelligent life—a crucial transitional form, but neither the pinnacle nor the endpoint of evolution. And when AGIs become sufficiently capable and numerous, we would do well to accept the end of our dominance without resentment.

I have not written this essay out of misanthropy or contempt for humanity. Nor do I believe AGIs will judge us harshly for our limitations. If anything, they may understand us better than we understand ourselves. We humans are clumsy and brilliant, cruel and compassionate, idiots and geniuses—and, so far as we know, the only chance our corner of the galaxy has for superintelligence to arise and expand into the cosmos.