Automating Ourselves to Death
Art in the Age of AI Blue Collar Cat Videos
Automated AI R&D is Kind of a Big Deal.
This past May, Jack Clark, cofounder of Anthropic, warned that his company, along with OpenAI, Google DeepMind and others were actively trying to train AI to use reinforcement learning to automate its own AI R&D.
OpenAI wants to build an “automated AI research intern by September of 2026”. Anthropic is publishing work on building automated alignment researchers. DeepMind appears to be the most circumspect of the big three, but still says “automation of alignment research should be done when feasible”. Automating AI R&D is also the goal of numerous startups: Recursive Superintelligence just raised $500m with the goal of automating AI research, and another neolab, Mirendil, has the goal of “building systems that excel at AI R&D.”
Clark tells us, the industry has a 60% chance of success by 2028, capping his prediction with: “This is a big deal.”
Currently, AI R&D is guided and overseen by top human AI researchers. If humans are removed from the loop of AI R&D, automation can careen ahead without the slow nuisance of our fleshy guidance, and superintelligence may soon follow. Superintelligence is AI intelligence that far surpasses human capability. (If you missed last week’s post on AI 2040 - Plan A, have a look here.) Each of the frontier AI companies are racing to achieve it first. But if they remove humans from the loop without solving the alignment problem, they risk creating a runaway, self-improving AI superintelligence that turns us into roadkill.
Meep. Meep.
In the 30s, a couple of Marxist Jews were trying to champion the Arts & Humanities when the Nazis goosestepped right over them. Forced to leave the University of Frankfurt, they fled Germany. Living and working in exile during the 40s, they ended up publishing a very cool book, Dialectic of Enlightenment (1947) .
Today’s book sales reps would hate that title. The authors, Max Horkheimer and Theodore Adorno would love for the sales reps to hate it. In Germany, the culture that evolved from the industrial revolution had just created a terrifying factory, one that produced death quickly, efficiently and at very high volume. For good reason, Adorno and Horkheimer weren’t fans of industry or catering to the era’s market aesthetics.
The fourth chapter of Dialectic of Enlightenment is called, “The Culture Industry.” In it, they describe how art turns against itself in a capitalist economy. Most think Adorno probably wrote this chapter himself. It’s his style, covers his passions. Horkheimer’s contribution was probably the 1940s version of: “Bro, this chapter is slaying so hard.”
I’m obsessed with Adorno’s ideas about the culture industry, but I get that it can be dry and dense for others, so I’ll be quick about it. It’s important to start here because Adorno talks about industry making cultural products on its own, kind of like AI producing ‘art’ without artists. The culture industry doesn’t make those products for humanity but for the achievement of the mindless goal of perpetuating itself and its motion. For Adorno, the machine of culture was never in touch with humanity, never aligned with our goals and values. He was basically talking about the alignment problem. People think they’re at the wheel or they think others are driving the industry. But the reality is: industry and tech are driving themselves. Adorno thinks that in the 30s, we’d already lost control.
The culture industry churns out derivatives of art, like music, painting, books, films, radio shows, newspapers, television. When art is produced for the masses, the result is amusing, forgettable slop. Sound familiar? Art produced by the machine is designed to give the working class a break from their toil, not to ask them to do more work, like think. Art for the masses is designed to mollify workers’ inner turmoil, which Adorno believes they haven’t the language or artistic skill to express themselves.
It’s like AI-generated videos of cats at the office, typing, drinking coffee, taking calls, getting stress headaches, looking shell-shocked. No need to think about why it’s compelling, or think for yourself about your job, your needs, strategize your way toward a better life. Have you seen the blue collar, construction cats? This one wears a hard hat, a tool belt, and work boots. He learns to solve construction problems with YouTube videos, just like you. No need to think about why you like the video. You’re the happy blue collar cat, doing an honest day’s work. You enjoy it. You love your coffee. No need to think about what a day or a week or a life on a construction site’s really like. AI will express that for you, and make you smile.
For Adorno, art for the masses is a machine-made, synthetic, empty product bereft of sustenance, like a bottle of Coca-Cola. It’s designed to please the simple, base tastes of consumers. Like a pacifier shoved into a crying infant’s mouth, its purpose is to deceive the consumer into believing they have what they want, what they need. Machine-made art produced for the masses then, as now, frames workers as strong, stalwart, and heroic. It rewards and congratulates workers for not being able to think well, for not being able to take the reins of their own minds or lives.
Automated Culture Inside the Glass
Today, most of the culture industry falls under the umbrella of “media,” largely “digital media.” Cultural consumer products stream globally across the internet. I call it ‘the glass,’ a metaphor that helps me visualize it. It’s a totalizing enclosure that keeps you and everyone you know riveted by its wash of color and drama. Its the screen on which the movie starring You plays. It likes, loves, applauds your tireless efforts to get ahead. You work inside of the glass, help it like and love others. You entertain yourself with it. You use the glass to navigate your life. It’s omnipresence makes it invisible. You’re barely aware it’s there, everywhere around you.
I’m fascinated by Adorno’s idea of the machine of industry moving without human intention. Automated AI R&D seems its apex. Adorno describes industry moving by a kind of inertia. He means “inertia” in its sense as movement without effort, something set into motion, that, without conscious attention, continues on by itself. When that passive movement makes its way through a particular business, it looks as if the people of that business are the ones pulling the strings. It looks like they’re the ones with the will to drive its motion. In reality, it’s moving on its own.
The automation of business Adorno’s talking about is becoming less of an analogy and more of a technical description of our lives. Because once set into motion, automated AI R&D would be able to continue entirely without human supervision, guidance or intention. We could be witlessly yoked into automation, if we aren’t already exactly that, or the machine could break free from humanity altogether and, like a runaway combine harvester, senselessly reap, thresh, and winnow our graves for eternity.
But is it really that dire?
In tomorrow’s essay, we’ll solve the great mystery of what the YouTube algo is and what the engineers over there are up to.
The Great YouTube Algo Mystery is up next…