OPINION: Beyond a Race to AGI: Reimagining the Manhattan Project

The U.S.-China Economic and Security Review Commission's 2024 Annual Report to Congress recommended that "Congress establish and fund a Manhattan Project-like program dedicated to racing to and acquiring an Artificial General Intelligence (AGI) capability." To achieve this, the Commission calls for priority funding for the U.S. to reach AGI before adversaries such as China. That said, without a well-considered approach, such a project could have disastrous consequences for the U.S. and the world.

Recent global developments in AI have only intensified the stakes of this technological competition. In February, Chinese startup DeepSeek stunned the tech world and Washington alike with its advanced processing capabilities and shockingly small training cost compared to U.S. models. In response, leading U.S. AI labs announced accelerated model release timelines, likely to comfort investors who panicked in response to DeepSeek's breakthrough. Moreover, President Trump stated that DeepSeek "should be a wake-up call" for U.S. tech companies. This heightened competitive atmosphere has made the Commission's recommendations all that more appealing—and realistic. As Washington weighs its strategy for reaching AGI, it's critical to understand exactly what's at stake.

AGI is commonly defined as an AI system with human or above-human level intelligence capable of learning, improving, and evolving on its own without additional training data. Should AGI be achieved, it would be a massive moment for humanity. Such an advancement would likely lead to exponential growth in technological and scientific progress and an astronomical shift in possibilities. Both good and bad. 

It’s easy to understand the appeal of such a breakthrough, especially to decision-makers in Washington. Such an advancement could shift the balance of technological power, giving whoever achieves it first a competitive edge in everything from military applications to economic innovation. In our current era of intense US-China technological rivalry, the Commission's suggestion of a "Manhattan Project-like program" makes sense. After all, the original project worked the first time, right?

The Manhattan Project was started during the Second World War to create the world's first atomic bomb. This endeavor combined the efforts of world-renowned scientists, industry leaders, military experts, and thousands of people across the U.S. to create this history-defining weapon of mass destruction. 

The Manhattan Project fundamentally shaped the current world order, helping the U.S. secure technological and military superiority that dominated the global balance of power for decades. Similar to the atomic bomb, AGI has the potential to upend the current global system, and because of that, it must be approached with the same degree of national attention, investment, and collective mobilization as the Manhattan Project. However, that's where the similarities should end.  

The Manhattan Project was created with the singular goal of producing a final product, the nuclear bomb, and to do so with extreme urgency. The world was at war, and tangible results were required quickly. Unlike the clear and immediate threat of fascism during World War II, the current U.S.-China technological competition, despite its importance, does not warrant the same intensity nor rushed delivery of AGI. 

Furthermore, the Manhattan Project focused entirely on innovation, leaving regulation, safety, and international norm-building for a later point in time. The consequences of this approach have been severe, including the catastrophic human toll in Hiroshima and Nagasaki, the myriad of disastrous effects on the community around Los Alamos national laboratory, and decades of nuclear proliferation and global security threats. A new national AGI project cannot be allowed to take the same approach because the world is not ready for AGI—at least not yet. 

The U.S. and the rest of the world are severely lacking in safety measures such as independent oversight mechanisms, that are vital for creating and eventually deploying such a technology. AGI developed without proper safeguards could lead to any number of issues, ranging from mass economic disruption to more severe national security risks if the AI system's goals don't match human values and intentions.

Without sufficient preparation and governance, achieving AGI could trigger a domino effect across our planet. Economies would be thrown into turmoil as automation replaces workers throughout countless industries. Our digital infrastructure could face new vulnerabilities from AGI-powered cyber attacks that humans would be incapable of reacting to or preventing. Drone swarms and lethal autonomous systems could revolutionize warfare with capabilities beyond what humans can defend against. AGI-driven surveillance systems could empower systemic oppression and invasions of privacy. Deepfakes and other generative content could become indistinguishable from reality, leading to extreme erosion in societal trust and further proliferation of misinformation. And perhaps the most terrifying is the possibility of rogue AGI systems whose goals are misaligned with humanity's best interest. That may sound like science fiction, but if we wish these situations to remain a thing of fiction, we ought to prepare accordingly.  

AI companies, think tanks, and academia have emphasized the necessity for effective AI regulation and safety. A new Manhattan Project-like program could, and should, be adapted to address these challenges directly and comprehensively.

Others have argued against an AI Manhattan Project for a variety of reasons, ranging from geopolitical destabilization to the risk of losing control of these advanced systems. While these are reasonable concerns that should be acknowledged and addressed, an adapted federally-funded Manhattan Project-like initiative would be best suited to address these concerns rather than a decentralized private approach like we have now. 

This new project should take a more holistic approach to AGI rather than simply racing to achieve it. Like the original Manhattan Project, the AGI project should include leading computer scientists, industry experts, and military officials, but it should go further and include a much wider range of experts, such as economists, anthropologists, historians, and policymakers. These multidisciplinary perspectives will be essential for developing the safety measures that should accompany this technology. AGI will impact everything,  which is why everyone should be represented in the process of preparing for its future.  

This reimagined initiative should not only pursue the creation of AGI, as the Commission recommends but should also aim to ensure the safe and secure development and deployment of AGI. Congress should assemble an interdisciplinary team that should be tasked with four main goals: 

• developing technical alignment solutions that ensure AGI systems reliably pursue human-compatible goals

• creating adaptive regulatory frameworks that evolve with the technology overseen by an independent committee

• instituting a transparent standardized risk and impact assessment report for public release concerning social, environmental, economic threats posed by AGI systems

• building international cooperation platforms and communication channels for AGI-related issues. 

This approach would transform the Manhattan Project concept from a single-track innovation sprint into a balanced initiative that addresses AGI's technological pursuit and its long-term implications.

The path to AGI does not need to mirror the Manhattan Project's path to developing the atomic bomb. By establishing independent oversight bodies with real authority and creating transparent reporting mechanisms, we can ensure a responsible, strategic, and socially conscious approach toward AGI. Safety should not be sacrificed in favor of speedy innovation. This collaborative, multidisciplinary approach would balance national security concerns with protecting the everyday citizen from AGI’s potential harms to employment, privacy, and social stability. We should focus on mitigating the potentially cataclysmic consequences of unchecked AGI development and ensuring that this technology benefits humanity rather than endangers it. 

The opinions and views expressed in this article are my own and not representative of any company or organization I work for.