\n<\/p>\n
On November 2,<\/span> 2022, I attended a Google AI event in New York City. One of the themes was responsible AI. As I listened to executives talk about how they aligned their technology with human values, I realized that the malleability of AI models was a double-edged sword. Models could be tweaked to, say, minimize biases, but also to enforce a specific point of view. Governments could demand manipulation to censor unwelcome facts and promote propaganda. I envisioned this as something that an authoritarian regime like China might employ. In the United States, of course, the Constitution would prevent the government from messing with the outputs of AI models created by private companies.<\/p>\n This Wednesday, the Trump administration released its AI manifesto, a far-ranging action plan<\/a> for one of the most vital issues facing the country\u2014and even humanity. The plan generally focuses on besting China in the race for AI supremacy. But one part of it seems more in sync with China\u2019s playbook. In the name of truth, the US government now wants AI models to adhere to Donald Trump\u2019s definition of that word.<\/p>\n You won\u2019t find that intent plainly stated in the 28-page plan. Instead it says, \u201cIt is essential that these systems be built from the ground up with freedom of speech and expression in mind, and that U.S. government policy does not interfere with that objective. We must ensure that free speech flourishes in the era of AI and that AI procured by the Federal government objectively reflects truth rather than social engineering agendas.\u201d<\/p>\n That\u2019s all fine until the last sentence, which raises the question\u2014truth according to whom? And what exactly is a \u201csocial engineering agenda\u201d? We get a clue about this in the very next paragraph, which instructs the Department of Commerce to look at the Biden-era AI rules and \u201celiminate references to misinformation, Diversity, Equity, and Inclusion, and climate change.\u201d (Weird uppercase as written in the published plan.) Acknowledging climate change<\/em> is social engineering? As for truth, in a fact sheet<\/a> about the plan, the White House says, \u201cLLMs shall be truthful and prioritize historical accuracy, scientific inquiry, and objectivity.\u201d Sounds good, but this comes from an administration that limits American history<\/a> to \u201cuplifting\u201d interpretations, denies climate change, and regards Donald Trump\u2019s claims about being America\u2019s greatest president as objective truth. Meanwhile, just this week, Trump’s Truth Social account reposted an AI video of Obama in jail.<\/p>\n In a speech touting the plan in Washington on Wednesday, Trump explained the logic<\/a> behind the directive: \u201cThe American people do not want woke Marxist lunacy in the AI models,\u201d he said. Then he signed an executive order<\/a> entitled \u201cPreventing Woke AI in the Federal Government.\u201d While specifying that the \u201cFederal Government should be hesitant to regulate the functionality of AI models in the private marketplace,\u201d it declares that \u201cin the context of Federal procurement, it has the obligation not to procure models that sacrifice truthfulness and accuracy to ideological agendas.\u201d Since all the big AI companies are courting government contracts, the order appears to be a backdoor effort to ensure that LLMs in general show fealty to the White House\u2019s interpretation of history, sexual identity, and other hot-button issues. In case there\u2019s any doubt about what the government regards as a violation, the order spends several paragraphs demonizing AI that supports diversity, calls out racial bias, or values gender equality. Pogo alert<\/a>\u2014Trump\u2019s executive order banning top-down ideological bias is a blatant exercise in top-down ideological bias.<\/p>\n It\u2019s up to the companies to determine how to handle these demands. I spoke this week to an OpenAI engineer working on model behavior who told me that the company already strives for neutrality. In a technical sense, they said, meeting government standards like being anti-woke shouldn\u2019t be a huge hurdle. But this isn\u2019t a technical dispute: It\u2019s a constitutional one. If companies like Anthropic, OpenAI, or Google decide to try minimizing racial bias in their LLMs, or make a conscious choice to ensure the models\u2019 responses reflect the dangers of climate change, the First Amendment presumably protects those decisions as exercising the \u201cfreedom of speech and expression\u201d touted in the AI Action Plan. A government mandate denying government contracts to companies exercising that right is the essence of interference.<\/p>\n You might think that the companies building AI would fight back, citing their constitutional rights on this issue. But so far no Big Tech company has publicly objected to the Trump administration\u2019s plan. Google celebrated<\/a> the White House\u2019s support of its pet issues, like boosting infrastructure. Anthropic published a positive blog post<\/a> about the plan, though it complained about the White House\u2019s sudden seeming abandonment of strong export controls earlier this month. OpenAI says it is already close to achieving objectivity<\/a>. Nothing about asserting their own freedom of expression.<\/p>\n The reticence is understandable because, overall, the AI Action Plan is a bonanza for AI companies. While the Biden administration mandated scrutiny<\/a> of Big Tech, Trump\u2019s plan is a big fat green light for the industry, which it regards as a partner in the national struggle to beat China. It allows the AI powers to essentially blow past environmental objections when constructing massive data centers. It pledges support for AI research that will flow to the private sector. There\u2019s even a provision that limits some federal funds for states that try to regulate AI on their own. That\u2019s a consolation prize for a failed portion of the recent budget bill that would have banned state regulation for a decade.<\/p>\n For the rest of us, though, the \u201canti-woke\u201d order is not so easily brushed off. AI is increasingly the medium by which we get our news and information. A founding principle of the United States has been the independence of such channels from government interference. We have seen how the current administration has cowed parent companies of media giants like CBS into apparently compromising<\/a> their journalistic principles<\/a> to favor corporate goals. Extending this \u201canti-woke\u201d agenda to AI models, it\u2019s not unreasonable to expect similar accommodations. Senator Edward Markey has written directly<\/a> to the CEOs of Alphabet, Anthropic, OpenAI, Microsoft, and Meta urging them to fight the order. \u201cThe details and implementation plan for this executive order remain unclear,\u201d he writes, \u201cbut it will create significant financial incentives for the Big Tech companies \u2026 to ensure their AI chatbots do not produce speech that would upset the Trump administration.\u201d In a statement to me, he said, \u201cRepublicans want to use the power of the government to make ChatGPT sound like Fox & Friends<\/em>.\u201d<\/p>\n As you might suspect, this view isn\u2019t shared by the White House team working on the AI plan. They believe their goal is true neutrality, and that taxpayers shouldn\u2019t have to pay for AI models that don\u2019t reflect unbiased truth. Indeed, the plan itself points a finger at China as an example of what happens when truth is manipulated. It instructs the government to examine frontier models from the People\u2019s Republic of China to determine \u201calignment with Chinese Communist Party talking points and censorship.\u201d Unless the corporate overlords of AI get some backbone, a future evaluation of American<\/em> frontier models might well reveal lockstep alignment with White House talking points and censorship. But you might not find that out by querying an AI model. Too woke.<\/p>\n This is an edition of<\/em> Steven Levy\u2019s<\/strong><\/em><\/a> Backchannel newsletter<\/strong><\/a>. Read previous coverage from Steven Levy here<\/a>.<\/strong><\/em><\/p>\n<\/div>\nMarx Madness<\/h2>\n
In on the Action<\/h2>\n
\n