The Generative AI Moment

Historians will look back upon 2023 as a marker of the beginning of the AI Era. Indeed, the launch of the GenAI model ChatGPT in November 2022¹ marked a major turning point in both the rate of technological progress in the larger AI field and the mass adoption and diffusion of generative AI models. The scale and speed of GenAI’s diffusion is nothing short of astounding. Consider the following: the IBM PC first shipped in 1981, and the iconic Apple Mac followed in 1984. Yet it would take until the early 1990s, nearly a decade, before there were 100 million PCs in use. Just six months after its release, there were already 100 million ChatGPT users.² GenAI is on pace to achieve the speed of diffusion in one year which the Internet took seven years to realize.³ 

The speed and breadth of GenAI’s technological diffusion – driven as much by commercial incentives as technical progress – is more than in any previous period of AI development. Already, GenAI tools demonstrate the ability to solve university-level math questions,⁴ score at a human equivalent on university entrance exams,⁵ and compose realistic visual representations of complex and sometimes novel phenomena. Multimodal GenAI tools create images,⁶ video,⁷ and music⁸ from text inputs, turn natural language into code,⁹ and run tests to catch bugs.¹⁰ Models fine-tuned on specific sets of data are being developed and used for domain-specific applications in finance,¹¹ science,¹² marketing,¹³ research,¹⁴ and more. 

This Is Only the Beginning

What began as a trickle of progress in GenAI has grown exponentially over the past year into what is no less than a technological revolution – one that is ongoing, and that is still in its very early stages. GenAI has propelled an enormous wave of AI capabilities that will impact all aspects of our lives. 

The United States must be prepared. GenAI is only one segment of a much larger and rapidly growing field of AI-accelerated technologies. Extremely powerful and strategic capabilities could be developed within the next five years – if not sooner. A technological revolution is underway that will impact all aspects of our society. GenAI has the power to drive significant advancements in science, medicine, and technology, enabling breakthroughs in areas like drug discovery, personalized healthcare, and climate modeling. It could help the world find cures for diseases that have long eluded us. It has the potential to enhance productivity, efficiency, and innovation across industries, leading to significant economic growth and improved quality of life. 

We stand at the precipice of a new era of intelligent machines that heralds unprecedented opportunities and challenges for society at large. As AI tools continue to evolve and mature, the choices made today in terms of regulations, ethical guidelines, and governance frameworks will shape the trajectory of AI’s impact on individuals, communities, and nations. This AI moment represents a critical juncture in human history. It is an era that holds immense promise, but also great peril. 

Understanding the Moment We Are In

The GenAI ecosystem that has begun to emerge consists of a handful of mostly (though not exclusively) American companies with massive foundation models;¹⁵ many more companies building an application layer of specialized products on top of these tools; and other actors releasing increasingly capable open-source general and specialized GenAI models.¹⁶ In the coming months and years, significant capabilities will come from a wide set of smaller entities or open-source users. Further specialized GenAI systems will be refined for different applications, such as medicine or law. Some models will likely be built from scratch on domain- or application-specific data, while others will be fine-tuned from larger frontier models through licensing or other commercial use agreements. Additionally, we will likely see the emergence of clusters of AI systems that interact with one another in pursuit of specific goals. 

In what will likely result in an even more fundamental change than the third industrial revolution,¹⁷ GenAI is already transforming the way we work and make decisions. With GenAI, every person in the world can access an AI-powered tutor on their personal phone. Software engineers are able to code up to twice as fast using GenAI tools,¹⁸ businesses are using GenAI to improve their products,¹⁹ and experiments have shown that GenAI raises productivity for writing-intensive jobs.²⁰ GenAI, at once both a tool for and a subject of scientific discovery, is revolutionizing the scientific process as scientists are able to conduct experiments faster, cheaper, and at a greater scale with AI-powered machines.²¹  

GenAI has the potential to accelerate U.S. productivity growth, which has been sluggish (with brief exceptions) for the past few decades.²² While it is difficult to measure the exact impact AI will have on productivity, what is certain is that every major leap in economic growth can be traced back to the role of innovation and technological change. As economic historian Robert Gordon reminds us, “Innovation is the ultimate source of all growth in output per worker-hour.”²³ History demonstrates that technological change raises output directly and that capital investment follows innovation – already, we see enormous funds flowing into GenAI. 

Period of Debate, Uncertainty, and Hope

As with any fast-moving, transformational technology, there is the potential for peril if GenAI is not smartly, responsibly, and effectively handled. These tools have sparked intense debates and a wave of uncertainty regarding their impact across society, the workforce, and even the relative military balance between nations. With their ability to generate human-like text, images, and even deepfake videos, these models raise concerns about the authenticity and trustworthiness of information in the digital age. The potential for misuse and manipulation has ignited discussions about the ethical implications, privacy concerns, and the need for robust regulations to safeguard against malicious uses of GenAI.

Furthermore, the rapid advancement of GenAI models has fueled apprehension about job displacement across various industries. As machines become increasingly adept at performing cognitive tasks, there is a growing realization that certain professions may become obsolete or significantly transformed. This has spurred discussions about the future of work, the need for reskilling and upskilling programs, and the potential impact on economic inequality. The uncertainties surrounding the extent and pace of job displacement have raised questions about the adequacy of social safety nets and the necessity for policy interventions to mitigate any adverse effects.

Moreover, GenAI models have triggered concerns about shifts in military and national power dynamics. In addition to the fact that greater economic power translates directly into military power, increasingly powerful AI models will also have an outsized impact on relative military balances and hard power equations. As nations compete to gain an edge in AI development and deployment, there is a potential for strategic imbalances and vulnerabilities in critical areas such as cyber warfare, autonomous weapons systems, and surveillance capabilities. The rapid integration of AI into military applications has prompted debates about arms control, international norms, and the need for global cooperation to ensure the responsible and accountable use of these technologies.

In this rapidly evolving landscape, the debates and uncertainties surrounding GenAI models highlight the need for thoughtful and comprehensive discussions to address the risks, mitigate potential harms, and establish ethical guidelines and regulatory frameworks that address the complex intersection of technology, society, and power dynamics.

The PRC and the Generative AI Moment

Technology lies at the heart of the strategic competition between the United States and the PRC. To date, the PRC’s publicly released LLMs have lagged behind those of American companies – such as OpenAI, Google, and Anthropic – but access to open-source platforms and cloud infrastructure could enable PRC firms to accelerate the development of their domestic models.²⁴ That said, the PRC still faces a number of systemic challenges in developing and deploying its own LLMs. One challenge the PRC faces is a scarcity of data available for training LLMs in Mandarin Chinese, as less than two percent of the Internet is in Chinese compared to nearly 60 percent in English (although PRC entities are developing English-based LLMs as well).²⁵ The language issue is further exacerbated by the Chinese Communist Party (CCP)’s comprehensive system of information control and censorship that exacerbates this challenge. The PRC’s domestic Internet domain is firewalled away from outside influence, minimizing the exchange of digital information with the rest of the world. At the same time, China maintains access to foreign data via products like TikTok, potentially an advantage for its AI industry.²⁶ The PRC’s authoritarian tendencies and risk aversion to releasing potentially difficult-to-censor LLMs could impede progress, despite the ability to undertake economic reforms for technological advances. 

Additionally, Beijing currently relies on U.S.-designed AI chips despite actively pursuing self-sufficiency for years, creating another potential choke point for China’s AI development. Training and inference for LLMs and AI applications currently requires large-scale supercomputers which are powered, in turn, by clusters of thousands of specialized AI chips called Graphical Processing Units (GPUs). American export controls have aimed to limit PRC access to these advanced microelectronics.  According to a recent assessment of the PRC’s AI capabilities, GPUs designed by U.S. companies “remain the most popular GPUs for training Chinese large-scale models.”²⁷ Market estimates (as of 2022) indicated that U.S. firm Nvidia accounted for as much as 95 percent of China’s GPU market.²⁸ Furthermore, the PRC’s domestic chip manufacturing capabilities at advanced nodes (e.g. 7nm) are presently operating at low yields, though Beijing remains determined to catch up.²⁹

However, these challenges are not necessarily insurmountable and should certainly not lead to complacency on the part of the United States. Leadership in GenAI should not be taken for granted: the nation has never confronted a full-spectrum competitor and technological power at the scale of the PRC. In the 1990s, amidst the advent of the Internet, many Western commentators speculated that the CCP’s centrally-controlled, top-down governance system would struggle to control the free-flow of information and democratic values embedded within the architecture of the World Wide Web.³⁰ Instead, the seemingly rigid PRC system adopted, adapted, and even excelled at remaking a domestic version of the Internet in its own image.³¹ 

Of late, the PRC has run headlong into multiple obstacles that have stalled the country’s economic growth. These strains are exacerbated by high levels of national debt, unfavorable demographics, lessening Western demand for Chinese products, and a series of poor policy decisions by the PRC leadership.³² The United States must be clear-eyed about the challenge and China’s ability to evolve in the face of emerging technologies. The PRC has demonstrated advantages in other strategic sectors such as drones, advanced batteries, and advanced network hardware. China is a world leader in building modern infrastructure, such as high-speed rail.³³ The PRC has stepped up production of advanced fighter aircraft³⁴ while its shipbuilding industry is turning out warships at an impressive rate.³⁵ China has also made impressive strides in space. In addition to remote lunar rovers and a newly built space station,³⁶ it is quickly catching up to the U.S. in space launches.³⁷  

The Other Revolution – The Changing Character of War

There is another extremely consequential, technology-driven revolution underway – the ongoing revolution in warfare. Much like the telegraph, steam engine, and railroad, which were driven by commercial interests but ultimately changed the way wars were fought, AI is similarly impacting warfare, driving some of the most fundamental changes we have observed in decades. Nowhere are the battlefield realities of this revolution more visible than in Ukraine, where AI is used for imagery analysis and AI-equipped drones are employed for target detection and tracking. Militaries around the world are closely watching the fighting there in search of new sources of military advantage, taking stock of what works and what does not on Ukraine’s highly dynamic battlefield where opposing sides adapt their methods and systems on a daily basis, and traditional military equipment is paired with the newest technologies in innovative ways.³⁸

Of great concern in this period of great power competition, the PRC’s military has recognized that a technologically-driven revolution in warfare is underway and is organizing itself for future warfare, including developing new warfighting concepts and supporting capabilities.³⁹ In that vein, the People’s Liberation Army (PLA) has set an objective to intelligentize its armed forces by 2027.⁴⁰ That is, to put them in a position to leverage big data and AI algorithms to identify rivals’ vulnerabilities, use influence operations to prevent opposing military leaders from understanding their environment, and employ multiple attacks to overwhelm military defenses. 

That the PRC seeks to erode U.S. military advantages should come as no surprise, as serious competitors do not simply cede military advantage to their rivals. Yet the Pentagon’s lagging response to PRC’s military modernization has led to a decisive and potentially dangerous shift in the relative military balance in the Asia-Pacific. While the United States seeks to avoid a destabilizing AI arms race with the PRC, the U.S. military is not moving with the sense of urgency required to keep pace with this fast moving technology and adapt to the changing character of warfare. Generating military advantage in this new age of innovation requires more closely linking operational concepts with new technologies, specifically AI. For that reason, the United States needs “Offset-X,” an effort to identify and develop the next disruptive competitive advantage to bolster conventional deterrence.⁴¹ Like its antecedents, Offset-X makes explicit choices and prioritizations among available technical pathways and solutions in order to generate response options to reverse an eroding military balance. 

The Need for a New Strategic Approach

Because of the outsized impact it is poised to have on both economic growth and relative military advantage, maintaining the United States’ leadership in GenAI is of paramount importance to our national security interests. America’s first-mover advantage in GenAI has given the United States a leg up on potential challengers, provided the nation with unique opportunities to drive innovation in this field, and – if action is taken now – will allow the nation to shape the trajectory of this ongoing technological revolution. Exceptional talent in AI research and development, a thriving startup ecosystem, and unmatched excellence in enterprise software development built a U.S. lead in creating cutting-edge technologies and solutions, providing a critical advantage in anticipating and countering emerging threats by strengthening America’s defense capabilities and enhancing its intelligence gathering and analysis.

Furthermore, GenAI has the potential to revolutionize areas like critical infrastructure, advanced networks, biotech, and cybersecurity. By maintaining leadership in these battleground sectors, the United States can secure its technological sovereignty and ensure the resilience of its national infrastructure. This is particularly crucial in the face of growing competition and attempts by adversaries to gain technological dominance. However, the United States must also bring its allies and partners along on this journey of discovery and innovation as they remain an unprecedented force multiplier, one that challengers lack. 

The United States is faced with a historic opportunity to provide a vision that shapes the trajectory of this technology, establishes effective international norms for global security, and promotes democratic values and interests. This is the moment to lead the way, setting the standards for responsible and ethical AI development and deployment. By collaborating with allies and partners, the United States can establish a framework that safeguards national security while upholding human rights, privacy, and democratic principles. Embracing this moment is not only a matter of strategic advantage but also a testament to the United States’ commitment to innovation, progress, and the prosperity and security of not only its citizens, but its allies and partners as well. The United States must invest in research, foster talent, and create an enabling environment that promotes entrepreneurship and technological leadership. This is the moment to shape the future.

The United States has been in the lead for strategic technologies before and lost advantages due to a lack of urgency and focused effort.⁴² Learning from these missteps, the United States must pursue a new strategic approach for this transformative period that protects its advantages in GenAI. Because GenAI is changing far too quickly for traditional policy making and regulatory processes, the United States needs a more nimble approach that provides adequate safeguards and regulation while at the same time does not stifle innovation, close off areas of opportunity, or allow strategic competitors to get ahead. 

This effort will require public-private partnerships at a scale for which the United States is not currently organized. A new model is needed – one that regularly brings together leaders in the public sector with world-leading academics, the private sector, and philanthropists to coordinate closely and address this rapidly evolving sector. This new public-private structure could provide advice to governments, conduct its own research, and work with other leading AI centers around the world.⁴³ 

A Warning – Preparing for the Next Wave

GenAI heralds an era of great promise and potential peril. As to the latter, it is important to highlight potential developments in the larger AI field spurred by the development of GenAI. As more powerful AI models, specifically AGI, loom on the horizon,⁴⁴ the nation must be prepared. AI has emerged as a transformative force capable of reshaping every facet of our existence. With unparalleled computational power and the ability to process massive amounts of data, AI algorithms have the potential to solve complex problems, revolutionize industries, and augment human capabilities. The significance of AI lies not only in its ability to augment our cognitive ability but also in its potential to revolutionize our understanding of intelligence itself.

Throughout history, certain moments stand out as turning points. These pivotal junctures serve as signposts, illuminating the transformative power of ideas and the extraordinary minds that shape the course of human events. Albert Einstein’s letter to President Franklin D. Roosevelt in 1939 was one such epochal event.⁴⁵ The letter, written by a visionary mind, recognized the immense power locked within the atom. It ignited a series of scientific discoveries and technological advancements that reshaped the world order, giving birth to the atomic bomb, nuclear energy, and the delicate balance of nuclear deterrence. Its significance transcended the realms of science and reached into the realms of politics, diplomacy, and global security.

Today, we are standing at the precipice of another transformative era – the Age of AI. The advent of GenAI heralds a new frontier, one in which the boundaries of human potential are challenged and redefined. As we imbue machines with human-like reasoning and decision-making capabilities, we confront profound philosophical and ethical questions about the nature of consciousness, morality, and the very essence of what it means to be human. In a similar way that the Atomic Age brought forth a new era of enormous potential but also fraught with grave risks, the Age of AI carries its own set of challenges that demand careful consideration.

---

Endnotes

1. Introducing ChatGPT, OpenAI (2023).
2. AI and the Automation of Work, Benedict Evans (2023). 
3. Abraham K. Song, The Digital Entrepreneurial Ecosystem—A Critique and Reconfiguration, Small Business Economics (2019). 
4. Adam Zewe, New Algorithm Aces University Math Course Questions, MIT News (2022).
5. Benj Edwards, OpenAI’s GPT-4 Exhibits “Human-Level Performance” on Professional Benchmarks, Ars Technica (2023).
6. DALL-E 2, OpenAI (last accessed 2023).
7. Make-A-Video, Meta (last accessed 2023).
8. Andrea Agostinelli, et al., MusicLM, Google Research (2023).
9. Andrew Tarantola, Natural Language Programming AIs are Taking the Drudgery Out of Coding, Engadget (2023); Tanya Tsui, Coding with ChatGPT, Medium (2023).
10. Codium AI (last accessed 2023).
11. Introducing BloombergGPT, Bloomberg Professional Services (2023).
12. Steph Batalis, et al., Large Language Models in Biology, Center for Security and Emerging Technology (2023).
13. See Raghu Ravinutala, The Power Of Domain-Specific LLMs In Generative AI For Enterprises, Forbes (2023); Copy.ai (last accessed 2023). 
14. See GPT Researcher x LangChain, LangChain blog (2023); Elicit (last accessed 2023).
15. Key players include OpenAI (GPT-3.5, GPT-4) with Microsoft, Google Deepmind (PaLM2, Gemini), Anthropic (Claude, Claude2), and Inflection (Pi). Large Chinese entities developing generative AI models include Baidu (ERNIE), Alibaba, Huawei (Pangu), the Beijing Academy of Artificial Intelligence (WuDao 2.0), and Tsinghua University, however their capabilities as publicly demonstrated appear at least a few months behind the state of the art in the U.S. ecosystem.
16. Key players developing open-source GenAI models include the UAE’s Technology Innovation Institute’s (Falcon), Meta (LLaMa, LLaMA-2), and Mosaic (MPT). See Cameron Wolfe, The History of Open-Source LLMs: Better Base Models (Part Two), Deep (Learning) Focus (2023). 
17. he digital age of information and communications technology is commonly identified as the 3rd industrial revolution. While it changed the way humans interact with computers and new platforms certainly sped up countless processes, it did not extend across the entirety of human life, as AI will increasingly do. Nor did we not see breakthroughs that significantly augmented human intelligence, which remains the most promising feature of the new wave of AI tools. On the Internet as the 3rd Industrial Revolution, see Jeremy Greenwood, The Third Industrial Revolution, The AEI Press (1997).
18. Sida Peng, et al., The Impact of AI on Developer Productivity: Evidence from GitHub Copilot, arXiv (2023).
19. Paul Smith-Goodson, The Extraordinary Ubiquity Of Generative AI And How Major Companies Are Using It, Forbes (2023).
20. Shakked Noy & Whitney Zhang, Experimental Evidence on the Productivity Effects of Generative Artificial Intelligence, Science (2023).
21. Eric Schmidt, This Is How AI Will Transform the Way Science Gets Done, MIT Technology Review (2023).
22. “From World War II until the early 1970s, labor productivity grew at over three percent a year. In the early 1970s productivity growth slowed dramatically, rebounding in the 1990s, only to slow again since the early 2000s.” See Martin Neil Bailey, et al., Machines of Mind: The Case for an AI-Powered Productivity Boom, Brookings (2023).
23. Robert J. Gordon, The Rise and Fall of American Growth, Princeton University Press at 569 (2016).
24. Yuka Hayashi & John D. McKinnon, U.S. Looks to Restrict China’s Access to Cloud Computing to Protect Advanced Technology, Wall Street Journal (2023); Josh Ye, Alibaba Rolls Out Open-Sourced AI Model to Take on Meta’s Llama 2, Reuters (2023). 
25. Languages Most Frequently Used for Web Content as of January 2023, By Share of Websites, Statista (2023).
26. Beyond software and data, the PRC faces a number of other constraints and limitations relative to the United States, which are discussed later in  the section of this report titled: Comparative Analysis: U.S. & PRC Generative AI Landscape.
27. Jeffrey Ding & Jenny W. Xiao, Recent Trends in China’s Large Language Model Landscape, Center for the Governance of AI at 7 (2023). 
28. Che Pan, Tech War: China Chip Veteran Says Nvidia Is Hard to Replace in Artificial Intelligence, Urges Start-Ups to Catch Up, South China Morning Post (2022); Kif Leswing, Meet the $10,000 Nvidia Chip Powering the Race for A.I., CNBC (2023). Huawei’s Kunpeng 920, a leading indigenous chip developed in China, was reportedly only 18 percent as efficient as Nvidia’s A100 chip as of July 2023. Jeffrey Ding, The Wudaokou Origins of China’s Large Models, ChinAI (2023).
29. Anton Shilov, Huawei’s Breakthrough 7nm Chips Projected at 50 Percent Yield, Tom’s Hardware (2023).
30. Clinton’s Words on China: Trade is the Smart Thing, New York Times (2000).
31. Bethany Allen-Ebrahimian, The Man Who Nailed Jello to the Wall, Foreign Policy (2016).
32. Lingling Wei & Stella Yifan Xie, China’s 40-Year Boom is Over. What Comes Next?, The Wall Street Journal (2023). 
33. Dan Wang, China’s Hidden Tech Revolution: How Beijing Threatens U.S. Dominance, Foreign Affairs (2023). 
34. Douglas Barrie, et al., China’s Air Force Modernisation: Gaining Pace, International Institute for Strategic Studies (2023).
35. Joseph Trevithick, Alarming Naval Intel Slide Warns of China’s 200 Times Greater Shipbuilding Capacity, TheDrive (2023).
36. William Harwoood, China Launches Fresh Crew To Tiangong Space Station, Maintaining A Permanent Residence In Space, CBS News (2023).
37. Svetla Ben-Itzhak, Is the US in a Space Race Against China?, Phys.org (2023).
38. Shashank Joshi, The War in Ukraine Shows How Technology is Changing the Battlefield, The Economist (2023). 
39. Robert O. Work & Greg Grant, Beating the Americans at Their Own Game: An Offset Strategy with Chinese Characteristics, Center for a New American Security (2019).
40. China’s PLA Aims to Leverage Advanced Technology for Use of Unmanned Weapons, Artificial Intelligence, Says Report, Economic Times (2023).
41. The Future of Conflict and the New Requirements of Defense, Special Competitive Studies Project (2022); Offset-X: Closing the Deterrence Gap and Building the Future Joint Force, Special Competitive Studies Project (2023).
42. Examples include hypersonic weapons in the defense space and 5G technology in the commercial.
43. We have done this before. During the Cold War, the U.S. government relied on strong public-private partnerships with federally funded laboratories and industry research centers that turned out impressive breakthroughs, such as the U-2 and SR-71 reconnaissance aircraft by Lockheed Martin Skunk Works’ partnership with the Air Force and Central Intelligence Agency. See Thomas Mahnken & Tai Ming Cheung, The Decisive Decade: United States Competition in Defense Innovation and Defense Industrial Policy in and Beyond the 2020s, Center for Strategic and Budgetary Assessments (2023). 
44. Sébastien Bubek, et al., Sparks of Artificial General Intelligence: Early Experiments with GPT-4, arXiv (2023).
45. Letter From Albert Einstein to President Franklin D. Roosevelt, Franklin D. Roosevelt Presidential Library and Museum (1939).