Welcome to Dark Markets, a weekly update on the Tech-right, Silicon Valley ideology, and related financial frauds from David Z. Morris, author of Stealing the Future: Sam Bankman-Fried, Elite Fraud, and the Cult of Techno-Utopia.
Subscribe
Mark Zuckerberg has published a 6,500 word essay, “The Future is for Everyone,” outlining his “vision for AI.” It is fundamentally an extended argument against the top-down control of AI, either by frontier labs or by governments. Its theses also happen to benefit Meta/Facebook’s business model - maybe.
The essay appears to take quite specific aim at many of the beliefs held by the Effective Altruists and fellow travellers running Anthropic in particular. Its rejection of centralized control can also be read as a rejection of the elitism baked into the EA worldview - though Zuckerberg’s own deep ties to the closely adjecent tech-right, and his specific deep ties to Rationalism supporter Peter Thiel, suggests the confrontation may be more performative and tactical than a true reflection of Zuck’s love for democracy and equality.
The long and short of this piece is that Zuckerberg is very clearly rejecting the longtermism and “doom” scenarios that animate the Effective Altruists and Rationalists who have major control over most of AI development.
This is useful not so much because Zuck actually believes it - he’s ultimately as much of an authoritarian as they are, as we’ve seen over the past few years. But it’s useful for him to say these things, even if he doesn’t believe them.
It represents a significant rift in the tech oligarchy that should be exploited by fans of democracy to bully naive techno-utopians to drop their witless faith and renew their commitments to humanity.
***
Zuckerberg hasn’t written a similar treatise since 2017’s “Building Global Community,” a wet nothing meant mostly to deflect from Facebook getting blamed for some bad things. That’s nearly ten years ago! So why is he publishing another Vision Quest now?
He’s pretty clearly imitating the logorrhea* of frontier lab leaders like Ilya Sutskever and Dario Amodei (“Machines of Loving Grace, 2024”), and even Leopold Aschenbrenner’s “Situational Awareness” essay. “The Future is for Everyone” has likely been in the works for a long time, so publishing it right after Aschenbrenner’s musings got incinerated on contact with the real world was probably an accident.
(The imitativeness of the essay is perfectly on-brand for Zuck, who hasn’t had a new idea since rating hot girls online. The past decade has been a series of frankly spectacular fuckups based on trend-chasing, from the Libra cryptocurrency to the Metaverse.)
Some business leaders might have sincerely wanted to share their thoughts, but with Zuck, you can safely assume everything here is purposeful and targetted. It’s a PR move, but it’s not just the public who are the target - “The Future is for Everyone” is full of very specific references most people won’t pick up on.
We’re going to unpack a few of them.
(*Logorrhea is one of my favorite words. It sounds fancy, but it means “word diarrhea.”)
·
December 3, 2024
Zuckerberg writes:
“The defining questions of our age are who will have access to superintelligence and what will we direct it towards. Will it be centralized and restricted to a few institutions, or will it be a tool that empowers everyone?”
This implicates a couple of different institutions: first, the labs, like OpenAI and Anthropic, who are currently fighting over “open weights” in models. Dario Amodei in particular is biased towards control and, frankly, strategic fearmongering about model capabilities.
The other institution implicated here is of course the U.S. government, which has used extraordinary powers to limit the distribution of cutting-edge models.
Both are again seemingly on his mind when he writes that one of the pillars of his philosophy is that “balance of power [is] the foundation of safety.”
This completely contradicts the view that Amodei has absorbed from moving in the same circles as people like Eliezer Yudkowsky. Their shared fundamental worldview is that they and their friends are the smartest people on early, and should be the ones deciding, basically, how society runs. They are unlikely to willingly give up control of anything without a fight.
·
June 23, 2024
Subscribe
For the remainder of August, readers accumulating three referrals will be entered into a drawing for a (physical) copy of Hayek’s Bastards: Race, Gold, IQ, and the Capitalism of the Far Right, an invaluable dive into the hidden psychology of hard money and free markets. I’ll have a review soon.
Every time you click the “Share” or “Refer a Friend” buttons in a post, and every time you restack a post, Substack tracks your sharing. If that sharing results in a subscription, you get one of your three referral credits.
As an added bonus, you also still only need to refer two free subscribers to get a month of free premium access.
(Paying premium supporters at the end of the month will get one drawing entry by default, and two additional entries per referral reward.)
Subscribe
Continuing, Zuck writes:
“It is surprising that the discourse from many developing AI is so filled with doom. I do not understand why anyone who believes that AI will eliminate most jobs and much of humanity’s relevance would rush to build that future … The notion that AI is so dangerous that the only safe path is an extreme concentration of power seems inherently problematic.”
Again, this “centralized power” could equally refer to the United States, or to OpenAI and Anthropic.
The invocation of “doom” here is equally pointed. Zuckerberg is explicitly rejecting the longtermist “doomer” view that rogue AI could lead to human extinction.
“Some argue that superintelligence itself or a small set of experts who control it should decide what is best for humanity. We disagree. The history of democracy and economics has shown that there is no single objective answer to how people define the best life, and therefore the best approach is letting people decide what matters in their own lives.”
It is equally good to see him reject the idea that we should defer our decisions to an AI. This is one of the most straightforwardly dumb and credulous concepts common among the EA/Rat contingent - that they can invent an AI so smart it can solve all the world’s problems.
Genuinely the best part of Zuck’s essay is when he rejects this truly anti-scientific assumption that an AI can magically fix things.
“The conventional view is that these concerns are about technology, and that if we take enough time then we can perfect or align the technology to produce a single benevolent superintelligence. I think this view of alignment is fundamentally flawed.”
Again, it’s hard to say what Zuck actually believes. But it’s good that he’s saying these things - rejecting the technical “solutionism” of the EAs, one based on their total antisocial failure to understand politics and society.
“Humanity is not a monoculture. People’s diverse values represent different tradeoffs they would make on important issues. There is no technological solution that can align with everyone’s opposing interests and values at once.”
This is one of the most direct if vague takedowns in the essay. It’s a direct rejection of Eliezer Yudkowsky’s concept of “programmable philosophy” - his version of “the AI will just solve all of our problems.”
For all its good features though, it’s still a PR effort by a CEO. He cringingly brings up Meta’s AI glasses, another impending flop based on stealing other people’s ideas.
The funniest part comes when he discusses user competition for compute, though.
"Everyone will have free or affordable access to these tools . We will offer free versions that will be accessible to billions of people. For those who want to pay to use more compute, there will be a dynamic auction mechanism that will guarantee that everyone gets the lowest price possible for the intelligence and compute they’re using … This will ensure the benefits of superintelligence are distributed widely.”
First, it’s very interesting to hear an auction described as a way to deliver a low price.
Second, this section in particular draws attention to Facebook’s planned business model for AI. Unlike social media advertising, AI is not currently profitable on a unit basis - the cost of the compute is currently subsidized even for paying subscribers of something like Claude Cowork.
Facebook can’t do Facebook-y things if compute is too expensive to be somehow profitable for free users.
Later, Zuck claims that:
“Meta is the company primarily focused on building personal superintelligence for everyone. Most other labs are focused on building AI for companies, governments, or other institutions, so if those labs lead, then the balance of power will favor larger institutions over individuals.”
This is the most clearly deceptive part of the essay. Meta is not at all focused on doing things for its users - notoriously, its users are not its customers. Users are the sources of data that it sells to advertisers.
So the offer of a free tier of LLM products strongly suggests that the plan is to gather data about free users.
And despite assurances elsewhere in the essay, there’s no reason to think Facebook is going to do a better job of delivering on AI privacy assurances when it comes to than it has for its entire long, rather sordid history.
Dark Markets is a reader-supported publication. To receive all posts and support my work, consider becoming a free or paid subscriber.
Subscribe