New AI Startup Wants to Take Your Job ‘as Fast as Possible'

Guys, you’re not cool enough to pull this off

---

IMPORTANT REMINDER: The current Three-Year Birthday Offer gets you a yearly subscription at 20% off forever, and runs from May 30th to July 1st. Lock in your annual subscription now for $80/year. Starting July 1st, The Algorithmic Bridge will move to $120/year. Existing paid subs, including those of you who redeem this offer, will retain their rates indefinitely. If you’ve been thinking about upgrading, now is the time.

---

From yesterday’s New York Times:

Mechanize, a San Francisco start-up, is building artificial intelligence tools to automate white-collar jobs “as fast as possible.”

I will ask the question everyone’s thinking: why would these fools dare to say something so outrageous and share it with the New York Times?

They did it 1) because they know this narrative—one tone more aggressive than their competitors’—will get them attention, the most important currency nowadays (the NYT’s article and this one are proof), and 2) because the AI industry is getting cocky; they live in a bubble of superiority and preachiness (read o3’s takedown of Sam Altman’s recent essay “The Gentle Singularity”).

Some of them genuinely believe the imminent loss of a huge portion of white-collar jobs is inevitable, so why not say it out loud? (Isn’t it funny that yesterday I published a Straussian interpretation of Altman’s essay, and today these guys do my job for me?)

Kevin Roose, the journalist who covered the news, starts the piece by recalling that “years ago . . . most tech executives at least had the decency to lie about it.” But I, for one, applaud Mechanize’s straightforwardness.

Except for one thing—they’re wrong.

At least in two important ways:

1. As a bottom-shelf startup, Mechanize has no real shot at taking anyone’s job.

2. Living in a bubble keeps you from seeing the danger outside. They’re blind to how far people are willing to go to sabotage a project so smug in its rhetoric and so caustic in its intent.

Apparently, those Waymo cars that protesters set ablaze in Los Angeles—how stupid, whatever the reasons—haven't lit up that many brains.

Ok, enough ranting, that’s not my style. Let’s take a look at Mechanize’s message and arguments, because as enraging as their slogan is, they have some good ones that require closer inspection.

(I will set aside in this post the question of how they intend to do it, although I agree with Roose’s remark: “I have some doubts about whether Mechanize’s [reinforcement learning] approach will work, especially for nontechnical jobs where success and failure aren’t as easily measured.”)

Tamay Besiroglu, co-founder of Mechanize (who you may know because he also co-founded Epoch AI in 2022, which I’ve featured many times in this blog, especially as creators of the FrontierMath benchmark), said this to the NYT:

We want to get to a fully automated economy, and make that happen as fast as possible.

That’s Mechanize’s pitch: what they want and when. Two possible readings here.

1. They believe that an automated economy will liberate us from our daunting jobs and make everyone more wealthy (”a rising tide lifts all boats”).

2. The benefits won’t be distributed evenly, and thus, rich people, shareholders, and the industry will come out wealthier, but the general public will be relatively poorer and without a job.

I'm sympathetic to the first interpretation (also the second, of course, but that’s more uncontroversial, so I’ll get to it a moment). To soften the force with which I opened this post, I'll explain how Mechanize's terrible presentation might be offset by the potential upside of an automated economy.

In “The Human Toll of Waiting for AI to Take Over,” I wrote:

So much time thinking about AI has changed how I see people. I notice the workers I used to overlook, doing the jobs no one wants to think about. Living in Madrid, that doesn’t mean miners or fishermen, but warehouse staff, delivery drivers, supermarket stockers, and cleaners. Like the worn landscapes of a city you’ve walked a thousand times, they fade into the background, part of the scenery. They blend in like mundane noise we drown out with premium headphones. They’re the subservient cogs in a system that feeds from them without giving back.

They’re everywhere, wheels of the world. They always were. I wasn’t looking. And that’s what unsettles me most: Not the shame of having ignored them blinded by the unspoken privilege of being a writer, but how unsurprising it feels that I did—does anyone ever think about them?

How selfish toward my own identity and ideology would I have to be to not want them automated out of their modern tragedy? They spend their lives languishing in jobs they don’t want for a pay that’s barely enough to keep returning the next day for another eight (or twelve) hours. Automation would liberate them from this perpetual condemnation.

But—and this is where the second interpretation of Mechanize’s story comes in—I finished that essay with these words:

I only ask these aspiring revolutionaries [Mechanize fits perfectly here] to pay attention and deep consideration—not just today, like I do, but always—to the transitory, liminal spaces. And the ones who dwell there.

Automation is a good thing if, and only if, “the ones who dwell” on those soul-crushing jobs are better off after automation has taken hold. The transition must be monitored by people—mediated by incentives—that have as their north star the well-being of the displaced.

But here’s the thing: Can a Silicon Valley AI startup make the benefit of those affected by automation its top priority? The question itself sounds like an oxymoron.

And it doesn’t help Mechanize’s case that, in a rare departure from his trademark AI optimism, Roose had this to say about their vision:

[I] found their pitch strangely devoid of empathy for the people whose jobs they’re trying to replace, and unconcerned with whether society is ready for such profound change. . . . like many A.I. companies working on labor-replacing technology, Mechanize has no novel policy proposals to help smooth the transition to an A.I.-driven economy, no brilliant ideas about expanding the social safety net or retraining workers for new jobs — only a goal of making the current jobs obsolete as quickly as possible.

Here’s the underlying, unfortunate truth as I see it: The story the NYT presents is profoundly enraging because it’s not intended for you and me or NYT readers, but for potential investors.

That’s why Besiroglu doesn’t care about being “devoid of empathy,” the bad publicity from articles like mine, or the irate comments of NYT readers. They care about those investors who will, I have zero doubt, put their financial stake higher in the hierarchy of priorities than anyone’s well-being.

This is the sad state of affairs. As well-intentioned as Mechanize’s founders might be—and to be clear, I’m not doubting their intentions, only the degree of free will they’ll retain once they enter the game—they need money to pursue whatever grand ambition they have. That money they need but don’t have is a rope around their necks. A loose rope at first that will eventually asphyxiate them—and that’s when they start to make concessions.

Besiroglu and his co-founders show a textbook case of overconfidence: their outlook far exceeds their ability to make it happen.

Think of OpenAI—the most powerful startup in Silicon Valley by far. It was “forced” to morph into a capped-profit (OpenAI co-founder Greg Brockman told journalist Karen Hao in 2020 that “in order to stay relevant,” they would have to ask for a lot of money, which was the beginning of the end). And now they want to make it into a benefit corporation (for-profit) despite the temporary abandonment of the conversion.

In Bloomberg’s Matt Levine’s words, “OpenAI Will Get A Bit More Normal,” which means, stripping his take from jargon, that it will get a bit more shareholder-friendly.

Which means, as the rules of the game are today, world-unfriendly.

So, yes, an automated economy as a “state of affairs” is a good thing. Isaac Asimov predicted a utopia of this kind fifty years ago. Kevin Roose recalls that John Maynard Keynes predicted the same thing 40 years before Asimov. And we could probably find earlier sources foreshadowing the same earthly paradise. But how it happens is a completely different matter.

Can a private company be the herald and harbinger of a peaceful humanity-first transition into an AI-automated world? I just don’t think so.

I do understand the libertarian principle that personal self-interest begets collective well-being through competition and other forces directly untamable by well-intended humans. But I hate to break it to you that there are too many ways to hack this otherwise beautiful, spontaneous process. And, right now, the most hackable of the vehicles that can make it happen is the Silicon Valley AI startup.

The NYT article ends with an especially incendiary comment by another Mechanize co-founder, Matthew Barnett, who was responding to Roose’s question on the ethical aspects of labor automation:

If society as a whole becomes much wealthier, then I think that just outweighs the downsides of people losing their jobs.

I’m sorry, Mechanize, but if this is the level of the discourse, which ignores the reality that millions of people outside your bubble experience daily, you won’t be able to make it enticing.