AI doomsters and Dial F for Frankenstein

Regular readers know we've been discussing the deeply dysfunctional and just plain weird culture of the techno-optimists for going on 20 years now.

Probably the first thing you need to keep in mind when dealing with the movement is that the optimism here is about the speed and magnitude of upcoming technological changes, not necessarily about whether their impact will be wonderful or catastrophic as long as it's big and soon. The AI doomsters are an outgrowth of the movement, and their ideas still find their greatest audience there. The conflict between the abundance crowd and the doomsters is essentially a sectarian disagreement between people who believe that large language models are on the verge of achieving god-like powers, which they will use to usher in a golden age free of want, and people who believe that large language models are on the verge of achieving god-like powers, which they will use to exterminate humanity.

This leads us to arguably the second most important thing to remember about techno-optimists: this is not a movement based on deep thinkers or profound insights into the nature of technology. Instead, the primary intellectual foundation comes from postwar science fiction, mostly filtered through movies, TV shows, and comic books. (Other influences include a fascist-friendly form of libertarianism, the Ithuvania effect, and constant and shameless grifting.)

Techno-optimists tend to have a notably weak grasp of the thing they claim to be obsessed with. Just listen to Elon Musk going off script talking about space travel or Marc Andreessen explaining how he simply tells chatbots not to hallucinate. Conversations that pass themselves off as explorations of the future of technology largely consist of sci-fi tropes being passed around. Ideas such as bipedal humanoid robots are judged not based on the soundness of their engineering, but on how well they would fit in with old Galaxy magazine covers or Mystery in Space comics.

 

 

 

 

One of the most common of these tropes dates back at least to Dial F for Frankenstein (Playboy, January 1965), a decidedly minor Arthur C. Clarke work, a lead-up to a pun that is weak even given the low standards of the genre. (If you want lighthearted Clarke, stick with Tales from the White Hart.) It is, however, one of the earliest examples of a popular sci-fi trope which has blossomed into something close to a religious movement in the past few years.

The idea is that if you construct a sufficiently complex system, superintelligence will almost immediately and spontaneously emerge, along with other human traits such as vengefulness, the need to dominate and control, and, most importantly, an overpowering survival instinct.

 

It is a trope born of equal parts anthropomorphism and narrative necessity. That sort of thing is fine for literary purposes, but we now have rich and powerful people who actually believe this, and that's not fine at all.

Patrick Boyle on the anniversary of Brexit

I could pull pretty much any random passage from a Boyle video and strike gold, but this one in particular caught my attention.

The first story was told by the Leave campaign, who drove a large red bus around the UK with the words, "We send the UK £350 million a week. Let's fund our NHS instead."

The pitch was that Britain was hemorrhaging money to Brussels, money that was being wasted, and that leaving would let the country reclaim its borders, slash bureaucracy, revitalize the health service, and, most importantly, bring back the bendy banana, which the EU had supposedly outlawed.

This was promoted heavily by right-wing populists and, more surprisingly, by a smaller group of left-wing Euroskeptics who wanted to leave the EU not to deregulate the economy, but to regulate it far more aggressively than Brussels would ever allow.

The two groups agreed on the destination and on absolutely nothing else. 

 It's remarkable how often factions of the left ends up implicitly and often explicitly supporting the far right, particularly given how these stands almost never turn out well. 

Peter Thiel's moments of honesty are largely limited to him saying bitchy things about Elon Musk

 A little bit of background: back in 1999, I was working for a large financial services company (you've heard of them). One of the biggest topics on everyone's minds was how to crack the online market. Every major bank was devoting serious money and resources to the question.

This was also going on in the startup world. Two of the leading companies in this space were Confinity and X.com, which would merge about a year later into the company known as PayPal.

You often hear about Elon Musk being the co-founder of PayPal, but that somewhat misrepresents what actually happened. These were two different companies with very different approaches and proposed products, and they brought very different things to the table. Confinity was focused on the area of payments and had already launched the PayPal product and had identified eBay as being the most important early market. Elon Musk's X.com was a more general, less well-thought-out idea for an all-services bank, something that would not be practical online for years.

X.com had more money and resources, particularly teams of engineers, and was technically the acquirer rather than the acquiree, while the primary product and basic approach that made the combined company viable enough to sell to eBay came from Confinity.

For an example of the quality of ideas that Musk brought to the table, here's an anecdote from Peter Thiel:

I was thinking of doing a book on PayPal … and [the chapter on Elon Musk] was going to be titled, “The man who knew nothing about risk.” … We had decided to give a credit card to anybody who wanted them. You got up to $10,000 credit limit. Elon had told the woman who was rolling the service out that he wanted 1 million people to be using the new credit card by the end of the year. Fortunately, it was about two levels down from the front page, and so not that many people were able to discover this. Some people did. They wrote us back and said, “You know this is fantastic! I haven’t had credit in years. I can’t believe you’re offering me credit. I haven’t even had a checking account in 10 years!” … We ended up with something like a 50% charge back rate—the worst subprime companies are like 4%-6%. And then, happily, we sort of rolled that product back very quickly. 

The 2018 Quartz article I pulled this quote from goes to embarrassing lengths to depict Musk as a bold visionary rather than someone who is really bad at business (the earlier Quartz article it links to is even worse), but these are not isolated cases. If you follow the man closely, you will find countless examples of Musk failing to grasp basic and sometimes self-evident concepts about the bond market, manufacturing, manned spaceflight, you name it, all delivered with an air of self-confidence that can only be called delusional.

After PayPal, Musk's business career has consisted of running a moderately profitable niche car company and having founded or acquired a string of companies that either lost money or were only occasionally profitable. His unprecedented fortune in no way reflects the money his companies have made. Instead, it is the product of his ability to convince people to give him money, be they investors, stockholders, or taxpayers.

It is a disconnect that sets Musk apart from the robber barons of the Gilded Age and every other multibillionaire (in 2026 dollars) I can think of before the year 2000.

“It really seems like anyone with some renders and a white paper written by someone being gassed up by an overly agreeable AI can get VC funding these days,”

Futurism points us to what happens happens when tech bro "engineers" run into actual engineers:  

A scathing takedown video — a collaboration between Irish aeronautical engineer Brian McManus from the YouTube channel Real Engineering and tech publication IEEE Spectrum — discusses the many pain points of Starcloud, a Y combinator company that raised $170 million earlier this year to develop data center spacecraft with the help of SpaceX, and how the white paper outlining the project feels far more like it was dreamed up by an AI than a team of human experts.

“It really seems like anyone with some renders and a white paper written by someone being gassed up by an overly agreeable AI can get VC funding these days,” McManus argued. “Billionaires will attempt to pull the rug over your eyes and convince you that this technology makes total sense, but reality is, this technology is dumb.”

 

My favorite part occurs around the 7-minute, 40-second mark when the narrator points out that, while it makes sense to have solar panels perpendicular to the sun, having the radiators catch as much sunlight as possible is the opposite of what you want.

It's a small point, but it really brings home just how little thought you have to put into something before you get the Y Combinator stamp of approval and almost $200 million in funding.

 

For a deeper IEEE dive.

Precarious Times

This has gotten lots of attention in tech circles, but not that many people have been connecting the dots. 

From the Information: 

Coinbase has cut its AI spending “nearly in half” even as it increases the number of tokens it uses, by using various measures to control costs. These measures include defaulting to open-weight models from Chinese firms, Coinbase CEO Brian Armstrong said in an X post on Friday night.

Armstrong’s post highlights both the growing usage among U.S. firms of open-weight models coming out of China to control AI spending even as the cost of cutting edge U.S. services rises. Armstrong said Coinbase had experimented with GLM 5.2, a just-released model from Z.ai, and Kimi 2.7, from Beijing-based Moonshot.

Pretty much everybody reading this has a good idea how massive the AI boom is, with current and proposed capital expenditures in real terms exceeding those of the railroad bubble, the dot-com bubble, and the early-2000s real estate bubble combined. And anyone who seriously follows the markets knows the narrative that drives the bubble, one of fantastic profits and unprecedented gains in productivity. But there's one aspect of that narrative that doesn't get nearly the attention it deserves: in order to justify the valuations of these companies and the capex being spent, you also have to assume that this will be a high-margin industry primarily based on one or more of the big, proprietary frontier models running on enormous centralized data centers and owned by either OpenAI, Anthropic, Meta, Google, or xAI.

If a substantial part of the industry follows the lead of Coinbase, the narrative is suddenly very different, with serious, even catastrophic, ramifications for some of the biggest, most powerful companies in the world.

Anthropic, to a degree, is protected by a moat here. Its standing as the leading model for coders protects it somewhat, but for OpenAI, which is already struggling with multibillion-dollar losses, a pushed-back IPO, and hundreds of billions of dollars of commitments coming up in the next three or four years, increased competition and substantially lower profit margins could present an existential threat.

Some of the biggest companies in the world have a high degree of exposure to OpenAI. The debt-ridden Oracle would be in serious trouble if anything happened to the company, as would SoftBank. Chip makers, including the most highly valued corporation in the world, Nvidia, would easily weather the collapse of OpenAI, but they would take a substantial hit in terms of revenue and almost inevitably stock price.

We won't even get into the possible ripple effect that might hit the private credit market.

I'm not saying any of this is likely to happen, but it seems worth thinking about, particularly if you've got money in the tech sector. 

If only we had effective, efficient technology to deal with these crises

I realize this puts me at risk of being one of those people who uses tragedies as an excuse to flog a pet cause, but it's a topic that deserves more attention. 

 

 Europe's heatwave 'linked to 1,300 deaths' as more records broken by Neil Murphy

Europe's unprecedented early summer heatwave may be responsible for hundreds of excess deaths, according to the head of the World Health Organization (WHO).

Temperature records were broken across the continent again on Sunday – including in Germany, Poland and the Czech Republic – as the extreme heat continued to move east.

In a post on X, WHO chief Tedros Adhanom Ghebreyesus said more than 1,300 excess deaths had been recorded since 21 June "linked to high temperatures in Europe".

 

If a drugged out Tech Messiah announces that he can make global warming go away with some untested and probably impractical geoengineering project, editors across the country rush to clear out space for the story, but when actual experts recommend a mature and highly efficient technology to address the problem, the silence is deafening.

Thursday, November 12, 2020

Ground source heat pumps are "the most energy-efficient, environmentally clean, and cost-effective space conditioning systems available," but maybe we can get journalists to talk about them anyway.

I'm joking but I'm not kidding. 

If Elon Musk or some other Silicon Valley visionary proposed some laughable plan based on non-existent technology, reporters would be scheduling interviews within the hour, but a solution supported by experts based on mature, tested systems will get little to no coverage.

One of the biggest crises facing California is a failing electrical grid, particularly during summer heat waves which are going to continue becoming more frequent and severe as the planet warms. Ground source heat pumps and similar technology could greatly alleviate pressure on the grid, especially when coupled with roof top solar. On top of that, its efficiency reduces demand for fossil fuels.

If we're going solve our problems, we can't go on being disinterested in solutions. 



From Wikipedia:

The US Environmental Protection Agency (EPA) has called ground source heat pumps the most energy-efficient, environmentally clean, and cost-effective space conditioning systems available. Heat pumps offer significant emission reductions potential, particularly where they are used for both heating and cooling and where the electricity is produced from renewable resources.
...
Ground source heat pumps are characterized by high capital costs and low operational costs compared to other HVAC systems. Their overall economic benefit depends primarily on the relative costs of electricity and fuels, which are highly variable over time and across the world. Based on recent prices, ground-source heat pumps currently have lower operational costs than any other conventional heating source almost everywhere in the world. Natural gas is the only fuel with competitive operational costs, and only in a handful of countries where it is exceptionally cheap, or where electricity is exceptionally expensive. In general, a homeowner may save anywhere from 20% to 60% annually on utilities by switching from an ordinary system to a ground-source system. However, many family size installations are reported to use much more electricity than their owners had expected from advertisements. This is often partly due to bad design or installation: Heat exchange capacity with groundwater is often too small, heating pipes in house floors are often too thin and too few, or heated floors are covered with wooden panels or carpets.
...
Capital costs may be offset by government subsidies, for example, Ontario offered $7000 for residential systems installed in the 2009 fiscal year. Some electric companies offer special rates to customers who install a ground-source heat pump for heating or cooling their building. Where electrical plants have larger loads during summer months and idle capacity in the winter, this increases electrical sales during the winter months. Heat pumps also lower the load peak during the summer due to the increased efficiency of heat pumps, thereby avoiding costly construction of new power plants. For the same reasons, other utility companies have started to pay for the installation of ground-source heat pumps at customer residences. They lease the systems to their customers for a monthly fee, at a net overall savings to the customer.

 

 

Another one of those stories that we really need Patrick Boyle for.

June 12 (Reuters) - Anthropic said on Friday it will "abruptly ​disable" its most advanced AI models for all users after the U.S. government ordered it to suspend access to the models for foreign nationals, citing national security ‌concerns.

The company received the export control directive to suspend access to Fable 5 and Mythos 5 for all foreign nationals, without being given specific details of its national security concern, Anthropic said in a statement. 

This is a bizarre story with all sorts of strange details and unexpected implications. Boyle's video overview is characteristically informative and thoughtful, too much so to summarize in a blog post, but there are a couple of points I should single out.

As previously mentioned, claiming that their models can have horrifying, even cataclysmic effects has long been a key branding component of companies like Anthropic and OpenAI. Therefore, it's not surprising that the CEOs saw this announcement as a marketing opportunity.

But while President Macron and the other middle-power leaders were busy panicking about their tech subservience and drafting a G7 communiqué pledging to discuss the risks, the tech executives themselves were busy doing something entirely different. They were using the panic to explain how incredibly dangerous—and therefore how incredibly valuable—their technology actually is.

Dario Amodei urged the room to resist the temptation to splinter, while Sam Altman and Demis Hassabis warned the world leaders about the imminent perils of AI-powered bioterrorism and cyberwarfare.

There's a distinct style of marketing in Silicon Valley where the best way to convince investors that your company is worth a trillion dollars is to tell the president of the United States that your product is so terrifying that it might accidentally end all human life.

If you tell Wall Street that you've made a very efficient spreadsheet tool, you get a tech multiple. If you tell them that you've built a digital nuke, you get an intergalactic multiple.

Perhaps the most important part of Boyle's video essay comes near the end when he draws a line through some dots that have been obvious for a while now but which most analysts have failed to connect.

If AI is not a natural monopoly and foreign competitors are offering models that are good enough, immune to US kill switches, and 60 times cheaper, then we end up with a highly competitive market.

And in a highly competitive market, profit margins collapse. The economic benefits of all this AI productivity would instead flow to the people using the models rather than the companies spending billions of dollars to build them.

This might be great news for businesses looking to automate their coding. It's terrible news for anyone buying shares at a $965 billion valuation.

So where does all of this leave the frontier AI industry?

Well, the lesson is simpler than Wall Street would like. For years, the theory was that building artificial intelligence required Silicon Valley genius, billions in venture capital, and an exclusive relationship with a cloud computing giant—ideally one whose CEO won't report you to the federal government as a hobby.

It turns out that there is no monopoly on math. The open-source models are good enough. They're 60 times cheaper, and nobody in Washington can switch them off on a Friday afternoon.

The productivity gains everyone was promised will still happen. They'll just flow to the businesses running the code on their own machines rather than to the investors who paid almost a trillion dollars for a global monopoly on math.

If the industry spends the money it has promised to over the next four years, something it has to do in order to keep the plates spinning, the capital expenditure in real terms will exceed that associated with the railroad boom of the 19th century, the dot-com boom, and the real estate boom of 2007 and 2008 combined. A collapse in the sector would leave an unprecedented hole in the markets.

In order to avoid disaster, things have to play out in a precise way. Large language models will need to become unimaginably profitable by the end of the decade, and they will need to support the spectacular profit margins that almost always require monopolies. If the current generation of models and infrastructure can't start bringing in unprecedented cash very soon, or if we see highly competitive markets based largely on smaller, cheaper models, things are going to get really ugly really fast.

NBC Blue was known for playing the uncensored version of Fibber McGee and Molly*

As I probably mentioned before, if you have any interest in the film and television industry, KCRW's The Business is essential listening. Kim Masters was one of the key figures breaking the Me Too story and was arguably the dean of Hollywood journalists even before that.

Recently, she interviewed former FTC Commissioner Álvaro Bedoya about the potential merger of Paramount/Skydance and Warner Bros. Discovery. While the Ellisons taking over Warner Bros. and particularly CNN has been treated as a done deal in the press for a while now, there are still hurdles to clear, and this interview makes a pretty good case that they are not trivial.

Listening to this got me to thinking about how much attitudes toward antitrust and media have changed over the decades, particularly during the 21st century.

The idea that studios and networks should not be allowed to gain too much power or engage in anti-competitive practices was one of, if not the major force shaping the industry in the 20th century. It effectively ended the studio system and thus had a great deal to do with the age of the Young Turk auteur directors like Scorsese and Coppola. It governed the way TV shows and movies could be sold and distributed. It was even directly responsible for the formation of ABC.

 From Wikipedia:

In the 1930s, radio in the United States was dominated by three companies: the Columbia Broadcasting System (CBS), the Mutual Broadcasting System, and the National Broadcasting Company (NBC). The last was owned by electronics manufacturer Radio Corporation of America (RCA), which owned two radio networks that each ran different varieties of programming, NBC Blue and NBC Red. The NBC Blue Network was created in 1927 for the primary purpose of testing new programs on markets of lesser importance than those served by NBC Red, which served the major cities,^[1] and to test drama series.^[2]

In 1934, Mutual filed a complaint with the Federal Communications Commission (FCC) regarding its difficulties in establishing new stations in a radio market that was already being saturated by NBC and CBS.^[2] In 1938, the FCC began a series of investigations into the practices of radio networks^[2] and published its Report on Chain Broadcasting in 1941. This report recommended that RCA give up control of either NBC Red or NBC Blue.^[1] At that time, the NBC Red Network was the principal radio network in the United States and, according to the FCC, RCA was using NBC Blue to eliminate any hint of competition. Having no power over the networks themselves, the FCC established a regulation forbidding licenses to be issued for radio stations if they were affiliated with a network which already owned multiple networks^[1] that provided content of public interest.^[2]

Once Mutual's appeals against the FCC were rejected, RCA decided to sell NBC Blue in 1941, and gave the mandate to do so to Mark Woods.^[3] RCA converted the NBC Blue Network into an independent subsidiary, formally divorcing the operations of NBC Red and NBC Blue on January 8, 1942,^[2][4] with the Blue Network being referred to on-air as either "Blue" or "Blue Network".^[5] The newly separated NBC Red and NBC Blue divided their respective corporate assets. Between 1942 and 1943, Woods offered to sell the entire NBC Blue Network,^[6][7] a package that included leases on landlines, three pending television licenses (WJZ-TV in New York City, KGO-TV in San Francisco, and WENR-TV in Chicago), 60 affiliates, four operations facilities (in New York City, Chicago, Los Angeles, and Washington, D.C.), contracts with actors, and the brand associated with the Blue Network. Investment firm Dillon, Read & Co. offered $7.5 million to purchase the network, but the offer was rejected by Woods and RCA president David Sarnoff.^[6]

Edward J. Noble, owner of Life Savers candy, drugstore chain Rexall, and New York City radio station WMCA, purchased the network for $8 million.^[1][2] According to FCC ownership rules, the transaction, which was to include the purchase of three RCA stations by Noble, would require him to resell WMCA with the FCC's approval.^[6] The Commission authorized the transaction on October 12, 1943.^[3][8] Soon afterward, the Blue Network was purchased by the new company Noble founded, the American Broadcasting System.^[2] Noble subsequently acquired the rights to the American Broadcasting Company name from George B. Storer in 1944. Additional broadcasting rights to the American name, including from the moribund American Network, were obtained at the cost of $10,000.^[9] The parent company adopted the corporate name American Broadcasting Companies, Inc.^[2][6] Woods retained his position as president and CEO of ABC until December 1949, and was subsequently promoted to vice chairman of the board before leaving ABC altogether on June 30, 1951.^[10]  

 You have to wonder what the regulators of mid-20th-century America would have said about Disney acquiring ABC and Fox Studios.

  

* I do not know of any inappropriate material associated with FM&M. Red Skelton was known for dress rehearsals that consisted of nothing but him telling one dirty joke after another.  

 

I've been waiting years for a good feral pig story so I would have an excuse to use these two quotes.

The first comes from one of the great American novels, Dashiell Hammett's Red Harvest.

"I haven't laughed so much over anything since the hogs ate my kid brother."  

This isn't original to Hammett—I remember hearing it once or twice growing up in Arkansas—but he still deserves credit, like all great writers, for knowing a good line when he hears one.

The second isn't exactly a quote (I'm working from memory here), but it's close. Years ago, on, I believe, NPR, I heard a USDA researcher explaining the difficulties of trapping these animals. As he put it, in terms of intelligence, when you take into account the smartest pig and the dumbest farmer, there is quite a bit of overlap.

Here's John Oliver on a legitimately serious environmental and agricultural crisis.  

 

Newport is also the guy who suggested that Anthropic deserved the Nobel Prize in marketing for the Anthropic launch

Something you won't hear me say often (because, as previously mentioned, I despise The New York Times), but this recent op-ed is essential reading. It's essential because computer scientist Cal Newport is one of the most thoughtful and clear-eyed voices in the AI debate. 

Anthropic recently dropped a classic of the form: a scary-sounding report titled “When A.I. Builds Itself” that claims A.I. could be moving closer to the capability of “autonomously designing and developing its own successor.” The company hopes that this recursive self-improvement will bring “enormous good” to the world, but also openly worries it might lead to humans “losing control” of these systems.

The public reaction to this report focused on a section that seemed to call for a worldwide pause on A.I. development. But if you read more carefully, it becomes clear that a pause isn’t actually what Anthropic proposes. Its report says that “if it were possible” to slow down the technology, then we should, but so long as the “least cautious” actors were advancing full speed, it suggests that Anthropic will have no choice but to do the same.

...

Let’s call this strategy “doom trolling.” It’s one of the defining and most arresting properties of our current A.I. moment, and I’ve come to believe that it’s morally indefensible.

There are really only two options for the intentions of A.I. companies when they engage in doom trolling. The first is that they actually believe that the systems they’re building have a nontrivial chance of producing hugely disruptive events — from destroying the economy in the best case to wiping out our species in the worst. If this were true, every reasonable ethical system would argue that there is only one acceptable response: to immediately stop working on any product that might accelerate such a future, and lobby with all of your resources to help force other A.I. companies to do the same. From a moral perspective, any other reaction would be monstrous.

The second option is that these A.I. companies aren’t really concerned about these risks, and that they’re injecting these doses of unresolvable doom for other reasons. They might want to amplify the perceived power of their technology at a time when they’re setting up their initial public offerings. Or they hope their performative reports and somber interviews will help them compete for top engineering talent coming from a Silicon Valley culture that’s steeped in this type of doomerism. The venture capitalist and A.I. adviser David Sacks recently suggested that Anthropic was using fear-mongering tactics as a method of “regulatory capture,” which can impede upstart competitors. Any of these reasons would mean that these companies are laundering the anxiety of millions to improve the financial fortunes of a vanishingly small number of major stockholders. This cynicism would be equally monstrous. 

When it comes to A.I., we’ve become so used to this tone of helpless, stress-inducing prognostication that we’ve lost sight of its strangeness. Imagine if Ford put out a report saying that it feared its popular F-150 trucks might soon start bursting into flames, but that there was nothing the company could do about it because automotive technology was too inevitable and important to slow down. You’re probably struggling to picture this scenario because no reasonable consumer product company would ever act like this.

The A.I. companies could start behaving the same way. To do so would require that they stop treating A.I. like some inevitable force that they’re struggling to steward. It’s not. It’s a collection of specific tools that these companies are choosing to design and sell according to specific business plans. Accordingly, they need to talk about their offerings like any other consumer product. This means explaining clearly whom these products are for, justifying their benefits and, critically, taking full responsibility for any harm they might cause. Just because A.I. currently enjoys a high-tech sheen doesn’t make it exceptional with respect to common-sense safety standards.

If these A.I. companies insist on continuing to pretend that they’re merely stoic observers of an unavoidable dystopian future, then perhaps it’s time to force the issue. As consumers, we can refuse to play the doom-trolling game. Next time Anthropic releases a dire report, or Sam Altman’s voice cracks as he imagines the disruption that OpenAI is unleashing, we can pivot back to the pragmatic: “OK, but what benefits am I getting by spending $1,000 a month on tokens?” If they continue to ratchet up the doom, then perhaps it’s time to transform dread into ridicule: The earnest pseudoscience of Anthropic’s white papers already borders on satire. The aura surrounding A.I. encourages a fretful submission to these tech leaders, but this could rapidly change.

 

 

A few days after the NYT op-ed appeared, Newport made an appearance on the podcast Better Offline, where he explored the same topics (he has an especially good time putting on his professorial hat and tearing apart Anthropic pseudo-academic white papers) and put those ideas in the context of the profoundly dysfunctional culture of Silicon Valley. We will be coming back to this one.

 

The original Cushing oil boom was started in 1912 by a man actually named Tom Slick. Not relevant to the post but I had to mention it.

In  most big stories, there are one or two pieces of absolutely essential context that get reported in major outlets but which no one seems to pay much attention to.

David Goldman reporting for CNN:

Today, neighboring Cushing is the hub of America’s energy market. It literally provides the oil plumbing for the United States. It’s where America’s benchmark West Texas Intermediate oil is priced and warehoused. From there, it’s piped to refineries around the country.

In normal times, Cushing stores around 40 million barrels of oil with capacity of up to 75 million.

These are not normal times.

Cushing’s current inventory is 21.6 million barrels, according to the US Energy Information Administration.

That’s dangerously close to operational stress levels, the tipping point at which Cushing struggles to supply all of its customers with the oil they demand.

When Cushing’s reserves get below 20 million, they effectively hit empty, scraping the bottom of the barrel of what is largely unusable sludge. 

...

Despite the largest crude supply shock the world has ever seen, US oil and gas prices still haven’t hit record highs during the three-plus months of war with Iran. And they’ve been sliding – sometimes sharply – in recent weeks.

That’s largely because the world was historically oversupplied with oil going into the crisis. Those stockpiles have acted as a global shock absorber.

But commercial oil inventories are rapidly dwindling. Oil stockpiles in the world’s wealthiest nations are falling by 6.3 million barrels per day and are sitting at just 2.6 billion barrels, according to the US Energy Information Administration. That’s just 100 million barrels above operational stress levels, said David Oxley, chief climate and commodities economist at Capital Economics.

The oil market can’t run down to the last drop, like your car can. Below a certain threshold, pipelines can’t maintain pressure and refineries can’t deliver all the various fuel grades their customers demand.

“Like blood pressure in the human body, the issue is circulation,” said Natasha Kaneva, head of commodities strategy at JPMorgan. “The system does not fail because oil disappears, it fails because the circulation network no longer has enough working volume.”

At this rate, the world’s oil market could enter the danger zone within a month.

That CNN story ran on 2026/06/12.

 

Here's another piece of context from Bloomberg.

 

 

 

As Nicholas Grossman‬ commented:

Graph of Strait of Hormuz crossings since June 2025. See that little spike all the way on the right? The one that's much lower than the leftward 2/3 of the graph, and immediately crashes back down?

That's what the US government and quite a bit of US media tell us is fully open, with normal traffic. 

Parting shots -- following up on some recent threads

On the AI bubble...

From Paul Kedrosky:

Hyperscaler and data-center investment, negligible through 2022, is now tracking toward $1.5 trillion annually by 2030. At that level, it will surpass, in real terms, the residential construction boom of 2005-06, the 1990s fiber build peak year, and the 1882 railroad peak combined. And it has reached funding escape velocity, with external debt and equity soon financing the majority of the buildout, with that share approaching 90% by 2030. AI infrastructure cycle is dependent on capital markets in a way that no prior technology cycle ever was.

 On the Colbert finale:

 On his last day hosting CBS’s The Late Show, Stephen Colbert played one of the most iconic songs from the Peanuts soundtrack.

“Oh no! I hope this doesn’t cost CBS any money!” he joked, as Louis Cato and the Great Big Joy Machine performed the song, titled Linus and Lucy, on air. It was a final dig at the network, which many fans believe cancelled the show due to Colbert’s criticisms of the Trump administration.

It did, indeed, cost CBS. Lee Mendelson Film Productions, the California company that controls jazz pianist Vince Guaraldi’s famed Peanuts catalog and had become increasingly litigious over unauthorized uses of the music, announced it had reached a licensing agreement with CBS for an undisclosed amount. The production company said it would donate all proceeds to World Central Kitchen, the disaster-relief food non-profit founded by chef José Andrés.

Adding well-deserved insult to injury, WCK is one of Colbert's favorite charities. 

 

On  the SpaceX IPO, Patrick Boyle:




And on the topic of humanoid robots,  Donald Duck:

1999 AD

Some works actually improve because they've aged badly—badly in the sense that they feel dated or that they now work in a very different way than the creators intended.

Though I'll probably get some pushback on this one, I'd argue this is behind much of the appeal of the films of Georges Méliès. Beyond the admittedly still impressive effects work, the charm comes largely from the antiquated costumes and art design, things that probably looked fairly standard at the time. For modern viewers, the combination of fantastic imagery, clever camera tricks, and a tawdry late-19th-century theatrical aesthetic creates something unique.

*1999* is a bit like that. Though no masterpiece, it's a great deal of fun for retrofuture buffs, and it tells us something interesting about the way people viewed technology and progress almost 60 years ago.

If you've ever taken a deep dive into the techno-optimism of the postwar era (very different from that of today), pretty much every detail will resonate. The botanical experiments reflect the excitement over the Green Revolution. The mention of the Mars colony reminds us that 1967 was the height of the Space Race. Talk of casually flying to the East Coast or Mexico City for a round of golf is hardly surprising at a time when the Concorde had just entered its prototype phase. From the school lessons to the automated kitchen to the disposable clothes and dinnerware are topics you will run across when reading the futurists of the 1960s.

For me, the most striking thing is how, once you get past the size of the computers and the crudeness of the displays, the world that people in 1967 expected to see 30 years in the future still seems ambitious in 2026.

We've discussed before how the late 19th and very early 20th centuries (particularly 1875 to 1910) and, to a lesser extent, the postwar era represented huge and unprecedented periods of ubiquitous, explosive technological change, especially in the United States and countries such as Britain, France, and Germany.

Whether you're talking about serious scholars or popular culture, speculation about the future in the decades after these huge spikes consistently tended to overestimate the coming rate of progress. There were exceptions, but where predictions were inaccurate, they were far more likely to be overambitious than conservative.

This bias is all the more noticeable given that one of the axiomatic beliefs of people like Arthur C. Clarke was that the near future would almost certainly exceed our expectations in practically every way.

    Trying to predict the future is a discouraging and hazardous occupation because the prophet invariably falls between two stools. If his predictions sound at all reasonable, you can be quite sure that within 20 or, at most, 50 years, the progress of science and technology has made him seem ridiculously conservative. On the other hand, if by some miracle a prophet could describe the future exactly as it was going to take place, his predictions would sound so absurd, so far-fetched, that everybody would laugh him to scorn. This has proved to be true in the past, and it will inevitably be true, even more so, of the century to come.

    The only thing we can be sure of about the future is that it will be absolutely fantastic.

    So, if what I say to you now seems to be very reasonable, then I'll have failed completely. Only if what I tell you appears absolutely unbelievable, have we any chance of visualizing the future as it really will happen. 

Speaking of camera tricks, take a good look at the car around the four minute mark.
 

Question for the MBAs in the audience, is losiing $21 billion a year bad?

 Ed Zirton has a scoop:

Today, I can exclusively report, based on audited financial documents viewed by this publication that have been independently verified by the Financial Times, that OpenAI lost around $38.5 billion in 2025, as well as other crucial details about the financial condition of the company. 

 ...

2025 — OpenAI Had $13.07 Billion In Revenue, $34 Billion In Costs and Expenses, and $20.92 Billion In Losses, with a net loss attributable to the company of $38.53 Billion

• Revenue: $13.07 billion
• Cost of Revenue: $7.5 billion
• Research and Development: $19.18 billion
• Sales and Marketing: $5.73 billion
• General and Administrative: $1.57 Billion
• Total Costs and Expenses: $34 billion
• Loss from Operations: $20.92 billion

Please note that 2025 was the year that OpenAI converted from a non-profit to a for-profit entity, leading to a $41.55 billion loss due to changes in fair value of convertible interests and warrant liability. 

Taking into account other minor factors like interest income and interest expense, OpenAI is left with a net loss of $60.35 billion, which it lowered to $38.53 billion by removing $17.87 billion in costs via that “net loss attributable to noncontrolling members capital” and another $3.95 billion via a “net loss attributable to redeemable noncontrolling interests.” 

Ultimately, the net loss attributable to OpenAI in 2025 was $38.5 billion. 

[Tje conversion from non-profit to for-profit is a one time thing. Focus on the Loss from Operation.] 

I'm not a finance guy—most of you can probably read these numbers better than I can—but I have been following the AI bubble closely (arguably too closely for my mental health), and I can tell you that not only is it bad when a company loses over $30 billion in a year, but the specifics of how OpenAI lost this money are even worse, directly contradicting some of the main assumptions supporting a bubble many times larger than the dot-com bubble.

It's useful to think of these costs in terms of training and inference, and possibly to break down training into training and post-training. Training is where the models are built. Inference is where the models are used to answer questions. Though corporate accounting can be murky, you can generally think of training as falling under research and development and inference as falling under cost of revenue.

One of the key assertions of the AI bulls is that while these companies do lose money (even the most enthusiastic and simple-minded booster will concede that point), they are making money on inference, and since training costs the same regardless of how often the model is used, while inference scales on a per-user basis, the economics should continue to look better as demand for AI increases.

But even if you completely eliminated R&D spending, OpenAI would still have lost money in 2025. Now it's true that the shift to token-based billing will help—it might even tip the training-excluded numbers into the black—but OpenAI is committed to spending more than a trillion dollars over the next four years, and in order to honor those commitments, it has to become not just marginally but enormously profitable in the very near future. Looking at these numbers, it's difficult to see where you get that kind of money.

Of course, the really eye-popping number is the more than $19 billion spent on R&D, which mostly translates to training and post-training. Any way you look at it, these are deeply disturbing numbers for anyone invested in the viability of OpenAI as a business, but things get much more disturbing when you consider the context.

GPT-4, the model that revolutionized the field and permanently changed the way all of us think about natural language processing and the way we interact with computers, cost a fraction of what any of the 5-series models cost. Hyperscalers and AI boosters are extraordinarily adept at coming up with self-serving benchmarks that put each new model in the best possible light, but even there, it's next to impossible to argue that the graph of capability to investment is not looking more and more like an S-curve.

This is hardly surprising given the state of the training data and the increased reliance on post-training. Large language models have already revolutionized quite a few fields and have provided us with powerful tools for a number of applications. This can hardly be called a failed technology at this point (though it is very possible that something better will push it into obsolescence in the near future).

But for the foundational narrative of the AI bubble, this is cataclysmic. The tens of trillions of dollars that OpenAI, Anthropic, Meta, Google, and particularly xAi/SpaceX are promising to pump into the economy in the next decade are based on visions of exponential curves and recursive self-improvement. If the future of AI over the next few years is one of flattening growth curves and minimum-wage workers in developing countries labeling data, things are going to get very ugly very quickly.

 

If you're going to do a post on index funds...

If you follow the financial press at all, be it the Financial Times, The Wall Street Journal, Business Insider, CNBC, CNN Business, Patrick Boyle, or Morningstar, there are two things you've heard over and over about the SpaceX IPO: the current valuation is absolutely insane based on the fundamentals, and a huge number of passive investors are about to be forced to buy the stock.

The potential undermining of index funds, without question the most important innovation of the past 50 years for democratizing investments, has gotten so bad that Burton Malkiel, the author of the 1973 classic A Random Walk Down Wall Street, has, at the age of 93, taken to the pages of The New York Times to argue that, as bad as SpaceX is as an investment and as reprehensible as Elon Musk is as a person, the concept of index funds is still worth saving. 

Given that so many millions of Americans are suddenly having SpaceX shares foisted upon them, I understand why some financial experts are criticizing the practice of index investing itself. Right now, just a handful of A.I.-related stocks represent almost half the value of the total stock market index. If A.I. stocks collapse, so will the worth of your index fund.

There are few individuals more responsible for the popularity of index funds than I am. I wrote a best-selling book about them and have argued for this investment strategy for over 50 years. In this fraught moment, I want to explain what you should do to protect your wealth. And, for the hard-core anti-Musk faction among us, I offer solutions to salve your conscience.

Let’s return to the argument against index investing. Many investment professionals consider SpaceX the poster child of the A.I. bubble. Investor fervor pushed the valuation of the company to approximately $2.1 trillion by the close of trading Friday — far more than even the most generous traditional valuation standards. The more generous the valuation, the bigger share of the stock market it captures, so SpaceX would immediately become one of the biggest holdings in the market.

Unlike prior initial public offerings, SpaceX shares are already so expensive there isn’t a lot of upside potential left. When Facebook, now Meta, went public in 2012 with a valuation of $100 billion, shareholders were able to benefit financially from its growth to a $1.5 trillion giant. Amazon went public with a generous (for 1997) valuation of $440 million, and shareholders profited as it grew into a $2.5 trillion behemoth. Not so SpaceX. Because it was owned by private investors for so long, much of the gain will immediately be handed off to its venture and private equity backers rather than preserved for new investors.

Moreover, unlike other public companies, SpaceX is employing a dual class share structure that gives Elon Musk essentially complete control with no independent oversight. Public shareholders will, comparatively, have no voice in corporate decisions. Mr. Musk controls multiple related corporate enterprises, raising the possibility of conflicted transactions within the Musk ecosystem. Many investors will be uncomfortable giving him so much power and holding an index fund in which he has so large a share.

These are all legitimate reasons to worry. But in my view, it would be a mistake to abandon an indexing strategy. Timing the market is impossible. Yes, the stock market is unusually concentrated today, and it is likely to get even more so over the next period with Anthropic and OpenAI looking to go public soon.

But the market has always been concentrated. If you had invested broadly in the stock market of the late 19th century, a lot of your money would have been tied up in railroad stocks. In the late 1970s, a bunch of it would have been in oil stocks. In the late 1990s, a disproportionate share would have been in internet stocks. The overall market’s generous 10 percent investment returns over the past 100 years have been generated by less than 4 percent of all stocks. The rest have not earned returns any greater than the yield on short-term Treasury bills.

Even when the market is down, index funds outperform. From mid-1999, near the top of the internet bubble, to mid-2000, when the market bottomed out, index funds did better than actively managed funds that try to pick and choose the best stocks to hold. S&P published recent results for its index in March. Last year about 80 percent of actively managed equity funds produced returns inferior to the S&P 500 index. And the few funds that outperform in one year are not the same as those that win the next year. When the results are compounded over five-, 10- and 20-year periods, over 90 percent of active equity funds underperform the stock market index. Indexing has proved to be the best investment strategy for building wealth.

I recognize, however, that investing is an emotional decision as well as a financial one. And some investments, even the financially sound ones, may make people morally uncomfortable, especially if they dislike some of Mr. Musk’s business and public policy activities.

So what can you do if you believe that owning SpaceX is morally indefensible? While there is no perfect replacement for an index fund, there are some alternatives to consider.

 Obviously, I'm reluctant to disagree with the father of index funds (as previously mentioned, probably the best and most important innovation  in investing of the past fifty years) on this, but I am far less sanguine about the near future of these financial institutions in a time of unprecedented market scams and irrationality.

These institutions work in large part because they are supported by public trust, uncompromised regulators, and reasonably high standards. We spent most of the 20th century building these things, and we've spent much of the 21st tearing them down.

While there is blame to go around on both sides, the culprits fall disproportionately into certain groups: the Republican Party, following trends that started with Gingrich and kicked into high gear with Trump; whiny, entitled billionaires, most of whom lucked into their huge fortunes; Silicon Valley visionaries and tech messiahs whose constant flood of self-serving fantasies were passed on by credulous journalists.