The challenge is a lot bigger than they think

Last week, the White House released its long-awaited AI Action Plan. The product of six months of work by some of the brightest minds in the Trump administration, the plan is an impressive document, built around three pillars—accelerating innovation, building American AI infrastructure, and leading in international diplomacy and security. It has already been accompanied by three executive orders relating to each of these pillars, with several more soon to come, representing a significant shift of direction from the caution that characterized the Biden administration’s approach to AI. Describing itself as “America’s roadmap to win the [AI] race,” the plan may soon be seen as inaugurating an effort comparable to a new Manhattan Project.

The advent of artificial intelligence already represents a technological breakthrough at least on par with the harnessing of nuclear energy nearly a century ago. Like nuclear energy, it is a technology clearly capable of doing extraordinary good for humanity or extraordinary harm. And like nuclear energy, its breakneck development is happening in the midst of tense international competition between superpowers—although since we are thankfully not in open conflict with China, the current AI race more closely resembles the post-1945 nuclear race with the Soviet Union than the Manhattan Project’s race against Nazi Germany.

The AI Action Plan recognizes the high stakes and high risks of this competition, seeking to roll back overly burdensome regulations that would stifle innovation, but without dismissing the real risks of AI. For instance, the plan highlights our current woeful ignorance when it comes to understanding the inner workings of major large language models and calls for DARPA research to better understand and control AI. And it warns that “The most powerful AI systems may pose novel national security risks in the near future in areas such as cyberattacks and the development of chemical, biological, radiological, nuclear, or explosives (CBRNE) weapons.” With AI as with nuclear, our government seems to be attending to the maxim “with great power comes great responsibility.”

That said, there are at least three significant differences between our situation today with AI and with nuclear science eight decades ago, which together suggest the need for this administration to expand and deepen its AI Action Plan—if it is to secure our American future.

First, nuclear technology was, at the outset at least, almost entirely a military and industrial technology. It was housed in powerful reactors requiring enormous infrastructure, not in your living room or your pocket. While AI systems require enormous investments in data centers and research labs (one of the key priorities of the Action Plan), they also have innumerable consumer applications that already saturate the market. This consumer-facing AI poses a whole slew of additional questions and challenges largely unaddressed by the Action Plan: in a world where 75% of teens have already experimented with AI companions, how can we combat the retreat from reality and the warping of humanity that the mass diffusion of these technologies is likely to engender? How are we to address the rampant cheating that is leading to a breakdown of education, or the atrophy of human skill and knowledge that comes from overreliance on easy (and often misleading) AI answers? Such questions are just as urgent as “how are we going to beat China?” since “What does it profit a man if he gains the whole world, but loses his soul?”

Second, any attempt at sensible AI regulation today has to reckon with the immense power of entrenched industry players. This was not the case with the Manhattan Project, which, although drawing on the industrial might of corporations like DuPont, was in many ways creating a new industry from scratch with the federal government clearly in the driver’s seat. Today, the government is playing catch-up. NVIDIA, which has established a near monopoly in advanced AI chips, currently enjoys a market capitalization of over $4 trillion (likely the largest in history even adjusting for inflation), with AI-powered titans Microsoft, Alphabet, and Meta not far behind. Such immense market power has already enabled them to sway critical Trump administration AI policy in their favor.

Finally, today we are living in an increasingly post-religious and post-truth world, which poses a problem for one of the AI Action Plan’s putative goals: promoting “human flourishing.” To promote human flourishing requires a commitment to human nature, something few in our tech companies seem to have much handle on or concern for. And whereas the Action Plan (and one of the accompanying executive orders) lays great stress on the need for AI models to “pursue objective truth rather than social engineering agendas,” as Pontius Pilate famously said, “What is truth?” One cannot simply demand that AI models be truth-seeking without a commitment to order society itself around objective truth—as revealed in nature and Scripture. Is that something that this White House is truly prepared to do?

The challenge before us, in short, makes the nuclear race look like a walk in the park. Our leaders will need exceptional wisdom, courage, and determination if we are to win the AI race without losing our souls.