Macrodoses #6
AI theater, authorship, a cursed ad campaign, and a surreal Soviet synth-pop track
Reboot Macrodoses are an informal dump of what our editorial board is reading, watching, playing, seeing online, thinking about, or otherwise experiencing. It includes mini-essays, reviews, links, pictures, plus perspectives from our beautiful brilliant readers—if you’d like to publish a 250-word Letter to the Editor about our previous essays, you can now do so here.
In today’s issue of macrodoses, we have:
A review of a new play inspired by Sam Altman’s all-too-brief firing from OpenAI
A survey of authorship practices in academic publishing
That Anduril ad campaign
A bop from the Georgian SSR circa 1988
“Doomers”
By Hannah Scott
As a resident of San Francisco, I cringe when out-of-towners confidently present their assessments of the city’s cultural landscape, which often strike me as inaccurate caricatures. The truth of tech culture is often stranger than the fictions about it. For this reason I admit I was skeptical when I learned that Matthew Gasda, playwright behind 2022’s “Dimes Square,” which satirized the infamous Downtown New York scene, had set his sights on people building AI in San Francisco.
A boardroom drama in two acts, “Doomers” departs from Sam Altman’s short-lived firing from OpenAI in 2023. It opens with a crisis management all-hands over cans of Yerba Mate and takeout tacos. Having just been removed from his position as CEO of a fictional AI company, MindMeld, by his board, Seth (Sam Hyrkin) attempts to rally the support of his colleagues between rapid-fire rounds of tweets and texts to friends at the DoD.
The stakes couldn’t be higher. For Alina (Emily Anderson), the Head of AI safety in the room, the total annihilation of humanity is on the line. For Seth and his lackeys, it’s a “really dope future” that hangs in the balance. Untangling these warring motivations behind building or not building AGI is the meat of “Doomers.”
For the accelerationist camp, these motivations don’t appear to reach much deeper than a glib, starry-eyed enchantment with the awesomeness of AGI. Jeff (Tommy Harkness) and Seth seem to be motivated less by a cartoonishly villainous version of capital accumulation or world domination, but by a juvenile obsession with building “special” tools, with spearheading a feat of engineering on par with the moon landing, with building a civilization that glimmers with a scale to rival the Roman Empire.
These expressions of a desire to “win” a boyishly sci-fi future, delivered earnestly by the actors, felt true to conversations one might overhear in Patricia’s Green or at Tartine. The doomers take an opposite view of the future, though their fears about total extinction feel just as improbable.
In the second act, which serves as a sort of funhouse mirror of the first, Harriet (Anna Connelly) and her fellow board members struggle to triage the fall-out from her move to oust Seth. The clash of visions for the future among the board members feels muddier and less poignant, but perhaps that’s because this group, which is ostensibly tasked with keeping its sociopathic CEO in check, is utterly powerless against the velocity of his quest to achieve a shibboleth of a goal.
In brief moments throughout the play, characters snap out of their science fictions and yearn for a different life. Jeff recounts a recent visit to his parents’ home in North Dakota, gazing into the audience and wishing he were playing board games on the porch. Harriet reveals that she was late to the emergency board meeting because she sat motionless in her car, transfixed by a Schubert symphony.
Such cracks in these futurists’ population-saving mission reveal an unshakeable desire for the immediacy of embodied experience. Their sentimental requiems for humanity come off as slightly heavy-handed. There’s just one character who seems to have fully shed this sticky nostalgia — Seth.
See “Doomers” through March 22 at Pallas Gallery in San Francisco and through April 19 at the Brooklyn Center for Theater Research in New York; doomers.fyi.
Giving credit where it’s due
By Hamidah Oderinwale
Recently I've been thinking about the construct of authorship. Authorship best practices are fuzzily defined: with best practices set by your field, your academic school of choice, and perhaps even by your department.
Generative AI makes it difficult to identify a clear primary source. If you’re an LLM-copier and paster, you’re basically left referenceless. An important function of authorship is assigning credit to an intellectual contributor. Who do you praise when their work supports your new discovery? On the flipside, it's also a means for accountability. Who do you reach out to when you spot an error?
I’ve learned that there are many differences and quirks to observe in the wild. For example, the “certified random order” policy has been acknowledged by the American Economics Association. Researchers submitting to an AEA journal with randomly-listed author lists even have to sign consent forms.
Courtesy of Stanford's Secure Computer Systems Group’s PI’s CV, I learned that authors are listed by contribution, then seniority. If necessary, ties are broken in alphabetical (or reverse-alphabetical) order.
As another case-study we can also look to machine learning. Chris Olah, co-founder of both Anthropic and Distill, is committed to being liberal with accreditation. In practice, if you scroll to the references section of a ML paper you might notice that they are often very long. This is partly due to relatively long author lists. If you happen to reference GPT 4’s technical report, you’re citing the 150+ contributors that made it happen.
ML has given into the spirit of openness and the positive-sum ideal. Insofar as there are disciplines that have yet to adopt similar policies, it’s a question of how, why, and if it’s worth it. Tentatively, I hope the answer to all three is yes.
ok but, like, actually don’t work at Anduril
By Jacob Kuppermann
After years of living in the Bay Area, it takes a lot for a tech company’s ad campaign to activate much of anything in me at all. Any traveller through San Francisco, the Peninsula, and surrounding environs cannot help but become inured to the unrelenting deluge of awful billboards and bus stop ads — nebulous LLM wrappers, desperately pivoting survivors of the 2010s boom-and-bust, whatever “PostHog” is.
So in a way I was pleased to witness the utter horrors of Anduril’s new “(Don’t) Work At Anduril” ad campaign. The thesis of the campaign is that America’s corniest and 2nd most important Lord of the Rings-themed defense contractor is not like the other girls; unlike the decadent giants of a prior age, Anduril and its patriotic band of drone designers are working hard every day to relentlessly improve their relentlessly lethal hard tech systems blah blah blah. They’re doing fake graffiti on their own billboards. They’re doing right wing uwu smol bean posting on their corporate account. They’ve got their hawaiian-shirted sicko cofounder doing cameos in mockumentary-style recruitment videos. It’s pathetic.
If the 2010s in bigtech were defined by a perpetual adolescence on the part of both tech labor and tech consumers — a veneer of luxe perks and ZIRP-subsidized millennial lifestyle products covering up widespread labor exploitation and unethical data collection — campaigns like Anduril’s (and the right-wing turn in tech imagery writ large) represent a childish conception of growing up, a leaning-in to the most jingoistic aspects of the American dynamism meme. This is the same set who founded the companies they now rage against — Mark Zuckerberg talking about “masculine energy” and seeming befuddled by trans-inclusive bathroom policy at his own company fits the same mold. This desperate cocktail of disowning the past and tough guy posturing in the present will of course age just as badly as the twee corporatism of the 2010s — for now, though, (actually) don’t work at Anduril.
"О стресс!"
By Kevin Baker
I've found myself unexpectedly obsessed with this song. Released in the Georgian SSR in 1988, "О стресс!" opens with goofy bass synthesizer riffs that seem to promise a mock-Soviet novelty song, after which it mutates into something far more disorienting. Vocalist Manana Todadze rattles off a breathless inventory of intrusions, her accelerating, staccato pace bombarding the listener with her sense of exasperation. Seemingly out of nowhere, Boris Shkhiani's croaking, bagpipe-like chorus interrupts with pleas to a doctor about stress taking an “interest” in him and haunting his dreams armed with a syringe. After that, it’s really all over the place. Nominally a synth-pop song, it hopscotches between genres and constantly subverts the listener’s expectations. The first time I listened to it, I thought it was funny. On the 1000th listen, I’m convinced it is a message sent through time, from one decaying, hypernormalizing society to another.
🌀 microdoses
If you’re in San Francisco, Reboot is hosting our first-ever screening featuring editorial board member Morry Kolman’s “Mr. Beast Saying Increasingly Large Amounts of Money” on March 26! RSVP here.
Check out Jon Rafman’s Proof of Concept, now on view at Sprüth Magers in Los Angeles through April 12, for an endless stream of slop operas set against a backdrop of ecological collapse.
Friend of Reboot Lucas Gelfond recently put out a web-based version of exiftool, which lets you read, write, and edit metadata in the browser. Play around with it here!
Subscribe to Reboot for more critical analysis of tech and culture (and certified bangers):