Mark Zuckerberg, in a (much) longer version of an essay originally published last month in the opinion pages of the Wall Street Journal:
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?
We propose a philosophy based on individual empowerment as the source of prosperity, invention as the primary purpose of superintelligence, and balance of power as the foundation of safety.
The themes of this essay are reminiscent of those in the “Personal Superintelligence” one Zuckerberg published a year ago. It has the same aspirational messaging about empowering individuals with decentralized superintelligence, the same acknowledgement of the new risks of this technology, and the same total reluctance to define “superintelligence” in any meaningful sense. The term appears 55 times in his latest essay, not one of which explains what it actually is. I suppose it must be synonymous with generative A.I. but the beauty of a proprietary term like this one is in its flexibility.
This is a long — superlong, even — essay and there is much to cover. Others have explored its unsubstantiated predictions, its assumptions, its contradictions, its hypocrisies, and its shallowness.
What I would like to focus on are the specific references Zuckerberg makes to studies, initiatives, or plain anecdotes. Despite the word count, there are few concrete references; much of this is about visionary concepts, not grounded reality. Despite their paucity, Zuckerberg also provides no links or citations, making it particularly difficult to track down the source of claims like this one (emphasis mine):
People fear that automation will outpace individuals’ capability growth, leading to job displacement followed by a difficult period as people learn new jobs. But there is no rule that AI must increase automation faster than it increases individuals’ capabilities or demand for new skills. Recent statistics suggest it may be more likely that individuals’ capability growth could match or outpace automation, in which case people will gain the ability to do many new things before their current jobs change. This would lead to a healthy balance and potentially even job growth.
What this actually refers to I have no idea. There is enough hedging — “recent”, “suggest”, “may be”, “likely” — that it kind of means nothing at all. What I think Zuckerberg is saying in this paragraph is that people can learn new things faster than A.I. can take jobs, but I could not find “recent statistics” backing this up.
There are some relevant observations, though. A June research note (PDF) from the Stanford Digital Economy Lab found, using data from ADP payroll services, that automation-friendly entry-level positions saw employment declines since the November 2022 launch of ChatGPT. Meanwhile, a report published by PWC in June (PDF) suggested increasing wages and employment in jobs that could be automated, even at the entry level. One possible difference is in methodology; while Stanford’s information came from payroll data, PWC’s was from job ads. PWC’s findings were also highly sector-dependent, and indicate entry-level jobs now require more skills to compete with automation. I think PWC’s report is the closest match to what Zuckerberg is arguing, but it is difficult to know for sure. Also, how exactly are young people fresh out of school supposed to develop skills like “team building” and “stakeholder management” before they can be hired?
Here is another anecdote from Zuckerberg’s essay, this time about how Meta is committed to building its physical A.I. infrastructure to support “high-paying local jobs, investment in schools and public services, ensuring energy prices don’t rise, and taking care of the environment”:
For example, in Richland Parish, Louisiana, where Meta is building a large data center, teachers received a $50,000 bonus this year because of the increased tax revenue from our investment. The superintendent told us that teachers are now moving there from across the country and he believes it will become one of the nation’s best school districts.
This is true, but elides the reasons why teachers are somehow getting bonus cheques from a data centre construction project they have nothing to do with. Reporting about this for the New York Times, Jacey Fortin writes that this undeniably exciting windfall is due to a “decades-old ordinance that directs 1 percent of local sales tax revenue to educators”. However:
But many of the workers are likely to leave in the next few years, ending the tax boom. And the Meta payment, which amounted this year to about $22 million, represents only a fraction of the taxes that a company would normally pay to build in the area.
Meta’s tax bill should be over nine times the amount it actually pays, and it will drop precipitously when construction is complete. But Meta made backroom negotiations to get all kinds of concessions and incentives to build its enormous data centre in the region. If the tax breaks had not been implemented and Meta constructed the data centre anyway, it would raise enough money to cover the salaries of all Louisiana public school teachers for more than a year. Eli Tan and Maureen Farrell, also of the Times, noted one resident’s sixfold rent increase thanks to housing demand from construction crews, but quotes the state’s governor saying the tax revenue pre-Meta was “zero”.
Zuckerberg:
We help keep electricity prices low by building our own energy-generating infrastructure wherever we invest. This ensures that not only are we not consuming energy that could have gone to the local communities, but in some cases we even supply a surplus of low-cost energy back to the communities. We think this is an important investment principle for sustainability.
Meta has previously emphasized its sustainability bonafides, but Zuckerberg omits such discussion here, perhaps because it would require considerable caveats that could be a distraction from the essay’s marketing angle. In Louisiana, for example, Meta is paying $2.65 billion to build several power plants with Entergy, but the region still needs more power. So Entergy is looking to acquire a plant in Texas, which could increase customers’ rates. In El Paso, at a different data centre project, Meta is only contracted to cover power plant construction costs for one to five years, after which costs will be transferred to customers. Like in Louisiana, it is also gas-fired.
Closer to home, a gas-fired plant in Alberta will not come online until years after Meta’s to-be-built data centre is up and running, all while electricity rates are predicted to climb.
Zuckerberg:
Developing personal agents requires a new way of thinking about alignment. Most labs today view alignment as a defensive measure for enforcing a centralized set of values. For example, one leading model was aligned to refuse helping draft a letter to prospective parents at a school because it thought standardized testing was unethical.
If this is a real anecdote, I cannot find the source.
And, if you can believe it, those are all the specific and checkable claims Zuckerberg makes in this essay. The rest of it is strikingly like one of those Microsoft concept videos from a decade ago albeit without the After Effects artist expense. One last thing, though, from Zuckerberg:
Privacy is an important foundation for individual empowerment and freedom. […]
Meta continues to make virtually all of its income from advertising targeted by some of the most privacy hostile technologies ever created. While “superintelligence” appears over fifty times in this essay, the words “ads”, “advertising”, and “advertisement” go unmentioned. They appear not even once.