AI Mania: From Tulips to Tokens

Michael Pollan has written about everything from mushrooms to apples and corn, but his coevolutionary story of how tulips and humans shaped each other may have been the most prescient given recent market parallels. GPU compute powering crypto for the last decade has spilled over into a bigger boom in language models. From bulbs, to coins, and now tokens. With animal spirits driving markets higher and CEOs blaming artificial intelligence for layoffs, who is using whom? AI, like the apple and corn plants, is gleefully spreading its seed while we prompt for another sweet hit of dopamine.

“Things are in the saddle and ride mankind.” – Ralph Waldo Emerson

The reductionist worldview that species are simply carriers of DNA hardly explains the complex relationship between humans and plants, let alone our situation with AI. The term itself has become a meme in the original Dawkins sense and started replicating on its own. We know this self-reinforcing system of models and data is powerful, yet can’t even agree on whether agents are truly intelligent or perhaps conscious.

It turns out Claude is not actually a real human. Racknitz's cutaway of the Mechanical Turk, 1789.

I haven’t had the pleasure of reading Pollan’s latest book on the topic of consciousness. My unpopular opinion is that AI has not solved “the hard problem” despite its obvious usefulness and the hype. I think Yann LeCun has it right in claiming that animals exhibit learning abilities far beyond machines. Donald Hoffman goes further and puts math behind a worldview where conscious agents are the base layer rather than something matter works its way up to. Sam Altman says we’ve reached the singularity, and while investors raise capital and profit, the rest of us are in amygdala hijack, worried about the next paycheck. If the labs truly believe these systems are waking up, then owning them and renting them out by the token is a question none of them seem eager to face.

Full self driving, according to Elon Musk

2014: in less than a year you will go from highway on ramp to highway exit without touching any controls

2016: we are less than two years away from complete autonomy

2019: feature complete this year, and a million robotaxis on the road the next

2026: a quarter to half of the United States by year end

A Turing Award and Nobel prize winner predicted in the 1960s that the programming occupation would become extinct, because computers would program themselves. Sixty years on we are still here, though it hasn’t felt that way on the roller coaster of hiring and firing. Grace Hopper was told that machines could only do arithmetic when she proposed writing programs in English, and the chips in every data center now carry her name. Coding was never the bottleneck anyway; relationships and understanding drive value. Tools amplify expertise and we are accountable for impact.

One thing that becomes obvious when attempting to analyze this ecosystem is the lack of transparency. Anyone who has worked on neural networks knows how difficult attribution can be, but it turns out the industry is currently exhibiting an almost fractal opacity. We created the content their systems are built on, yet they won’t even share how much energy each query uses, how much water goes to cooling, or who ends up holding the land and the aquifer when a site goes quiet.

“All these plants, which I’d always regarded as the objects of my desire, were also, I realized, subjects, acting on me, getting me to do things for them they couldn’t do for themselves.” – Michael Pollan

Anecdotally I see agents searching the web not once but 5-10 times because the model is stale, and all that traffic contradicts the SaaSpocalypse hypothesis. Managing them exhausts not only my attention but our limited natural resources. New transmission takes four to eight years and transformers are back ordered, so demand shows up on everyone’s monthly bill and as a hum the neighbors can’t sleep through. Erin Brockovich is mapping the proposed sites and the towns fighting them. Fear of losing to China ends every local objection, which is how the military industrial complex has always worked. While we can do more faster, the quality is debatable, and the externalities aren’t accounted for at any level.

33 operational, 66 under construction, and 8,837 places where people wrote in with concerns. The data center map, July 2026.

Lyla June said at a recent food forests course that humans have the potential to be a keystone and serve the planet rather than dominate it. At PNW Climate Week I went on an agrivoltaics tour called Sun, Soil & Sovereignty, where the same acre grows food and generates solar electricity. Data sovereignty could look like that, with compute sited where the panels and the crops are, and the people who live there holding a stake in the land, the water, and the machines. It is becoming a public utility and should be owned like one. That reframes the race...