What do the Federal Reserve, the Patriot Act, Net Zero, COVID lockdown policies, foreign disinformation, and the proposed, globally-coordinated AI slowdown have in common?
Every item on the list was preceded by a crisis, either real or imagined, planting an existential fear in the minds of the public. This created the justification for a massive, globally-coordinated government response.
In many cases, the swift government action was initially greeted with a sense of public relief. But, sooner or later, in all of those cases, the negative byproducts of the power-grabs by the state became clear.
The FISA courts that were expanded in the name of fighting Islamic terrorism were later used to enable deep state spying on a domestic presidential campaign. COVID lockdowns, established in response to a frightening virus we knew little about, persisted long after cases like Sweden, Florida, and Texas, proved their superfluousness. Net Zero policies, created in response to the never-ending climate change emergency, have left Europe economically crippled and dependent on foreign energy.
And in all of these cases, the government’s response resulted in a massive expansion of the bureaucratic blob – new personnel and government departments. New titles and responsibilities, like “Director of the Disinformation Governance Board.” A global sprawl of ideologically captured NGOs, researchers and “civil society” organizations, lining up to drink from the newly-opened spigots of American taxpayer dollars allocated to address the new “crisis.”
Some might say we need to pace the frontier of government expansion, to safeguard the public against these negative byproducts.
And the current frontier is AI regulation. In the space of two weeks, we’ve seen the two leading AI companies – OpenAI and Anthropic – demand the government step in to collectively slow the industry down. This followed the resignation of an Anthropic researcher on the grounds that superintelligent AI could “kill us all,” a statement which attracted over a hundred million views on social media and was aired on primetime channels across the world.
Almost immediately, the global governance machine sprang to life. 70 UK Parliamentarians demanded a ban on superintelligent AI. Former president Barack Obama and former Vice President Kamala Harris joined in. Another British politician urged the United Nations to intervene.
And just a few days before the resignation of the Anthropic researcher, Sen. Bernie Sanders (I-VT) introduced legislation that would ban AI superintelligence and establish a new agency at the federal level dedicated to AI safety. Anyone who violated the bill provisions by developing frontier AI models without the blessing of the government would face up to twenty years in prison.
Most suspiciously of all, Netflix decided to drop an AI apocalypse documentary, just two days ago, featuring interviews with the very people demanding government intervention to prevent the AI doom scenario: Dario Amodei, Sam Altman, Eliezer Yudkowsky, and more. Netflix documentaries, of course, take quite a bit of time to make, which has fed allegations of coordination.
A vaguely-defined existential fear, a unified media response (complete with a ready-to-go Netflix documentary), followed by interventions from the likes of Obama and Harris, and proposals to establish a new government agency.
Rather like the financial crisis of 1907, which saw the creation of the Federal Reserve. Or 9/11, which saw the establishment of DHS and the expansion of FISA courts – both of which were later turned on the American public. Or the largely imagined “Russian disinformation” panic of 2017, which saw the massive expansion of government-funded “counter-disinformation” programs that resulted in the mass censorship of Americans (including the current President, as well as the current HHS secretary). A crisis, real or imagined, followed by inexorable pressure for government expansion.
Before we rush headlong into what’s looking like exactly the same playbook, perhaps we should pause, take a deep breath, and ask some basic questions. Pace the frontier, if you will.
1) The Whom
A key element of Dario Amodei’s call to “pace the frontier” is for all AI companies to grant access to “embedded evaluators who have employee-like access to verify safety practices and report incidents.”
We’ve seen this movie before. It started with external “trust and safety” councils who had access to content moderation systems at platforms like Twitter, whose members included LGBT and feminist activist groups. It ended with a vast, globe-spanning blob of trust and safety, brand safety, and counter-disinformation orgs, often funded to the tune of millions of dollars by U.S. government agencies, which resulted in the mass censorship of social media.
We have every reason to believe that the “AI evaluators” that Amodei wants embedded in every company are the exact same types of people. His preferred watchdog, Model Evaluation and Threat Research (METR) is run by a CEO who – standing in front of a power point presentation featuring migrants on a raft – said that future generations would be “embarrassed” that nations “barred people from a location” based on “where they happen to be born – their nationality.”
That’s the CEO of the company that Amodei wants to be embedded in every single AI company – embarrassed by the existence of national borders.
To make matters worse, we don’t know who’s funding METR. The nonprofit has raised a staggering $71 million over the course of just six months, but researchers scouring the money trail have accounted for just $350,000 of it. The rest, for now, appears to be dark.
If AI safetyists really believe the future of humanity is at stake, and on that basis want to achieve a bipartisan consensus, it is their responsibility to explain how the measures they want won’t end up handing power over the most consequential technology of modern times to people who want to impose fringe left-wing viewpoints on the rest of humanity.
Given that Amodei has come out in favor of empowering these exact people, it’s fair to say that they have a lot of explaining to do.
2) The What
Everyone who goes on TV to predict the coming AI apocalypse is impossibly vague in their predictions. We don’t know what a superintelligent AI would do, they insist – just that it would be more intelligent than us.
Putting aside the fact that there are strong arguments against the notion that LLMs like ChatGPT or Claude will ever be “superintelligent” in any meaningful sense, let’s assume that this is correct. How would a superintelligent AI go about attacking humanity?
The most plausible scenario is AI agents taking control of a biolab and using it to manufacture deadly pathogens. Well, if that’s the main concern then that’s the limiting principle – there would be no need to empower regulators who have concerns other than that one (like making sure future generations are embarrassed by national borders, for example).
The AI doomers insist on vagueness – we must insist on specifics. Vagueness is what justifies the expansion of government without end. If that isn’t their intention, let’s hear the specifics.
3) The How
Following on from the last question – once we have established what we are trying to prevent, not just the vague idea of apocalypse, we can start to address the how.
Is it securing nuclear facilities against cyberattacks? Well wouldn’t you know it – there are already a plethora of government agencies working on that exact problem. The International Atomic Energy Agency has been around for a long time. As has the Pentagon.
Is it about securing biolabs? Have you heard about the International Experts Group of Biosafety and Biosecurity Regulators? Or the International Biosecurity and Biosafety Initiative for Science? Or the Association for Biosafety and Biosecurity?
How about that – the global regulatory bodies for the two areas that people are most concerned about in relation to AI attacks already exist.
Isn’t it funny how, once you start getting into specifics, the case for handing control of our technological future to a bunch of open borders supporters and Effective Altruist weirdos disappears?
Maybe that’s why they haven’t been getting down to specifics. But we should. It’s the only way to pace the frontier.
Allum Bokhari is managing director at the Foundation for Freedom Online, a digital censorship watchdog.