dmitting one doesn’t know something is practically heresy in Washington, D.C., especially if one is an elected official. This norm made it all the more impressive—though it should not have—when it emerged that Rep. Don Beyer was pursuing a degree in machine learning from George Mason University. The Democrat from Virginia hopes to learn more about the concept to better inform his legislative priorities and, ultimately, deliver better results for his constituents. Regardless of where one stands on Beyer’s politics, his efforts to learn more about what is and will be the most transformative technology of the 21st century is laudable and is worthy of emulation.
Beyer’s return to school couldn’t have come at a more opportune time. In November 2022, OpenAI released ChatGPT, an artificial intelligence chatbot that took the internet by storm, offering both promise and peril, depending on how one interpreted the tool. The tool scored early notoriety for passing law and business exams, becoming a tool for rapid marketing content generation, and even declaring its love for one user. The tool, trained on unfathomably large data sets, is not foolproof. When asked to do rocket science it, in the words of NPR, “crashed and burned.” It is, of course, early days for ChatGPT. Advocates extol the virtues of how it will revolutionize the way we live, work, and love; critics contend it is an overgrown parlor trick, merely regurgitating (and very well) what data it ingests.
It is often said, and rightly so, that technology outpaces the ability of policymakers to come to grips with both it and its implications. Beyer’s pursuit of a graduate degree and the arrival of ChatGPT vividly illustrate this delta. Indeed, while AI is not at the level of C3PO or R2D2 (yet), another resident of the Star Wars universe may well opine that “begun, the AI wars have.” This is a central tenet of Paul Scharre’s absolutely superb book on artificial intelligence, “Four Battlegrounds.” Artificial intelligence is already turbocharging the challenges policymakers face in managing geopolitics, security, and military affairs—it is a state of affairs that will only get worse. Perhaps most alarmingly, the consequences of getting AI policy wrong today could well determine the balance of power in the years to come.
Scharre smartly deconstructs the complexities of artificial intelligence, choosing not to focus on the ins and outs of the algorithms, but on the four titular battlegrounds: data, computing power (or “compute”), talent, and institutions. He argues, quite rightly, that competition in each of these will define the leaders of the artificial intelligence race, and quite possibly the winners of strategic competition. It is an inspired way of illustrating and addressing the panoply of challenges AI touches upon. Data deals with privacy, among other things. Computing power directly affects semiconductor supply chain security and resource consumption. Talent vividly shows the challenges of immigration and the balance between openness and security. Institutions, above all, touche on acquisition, ethics, and much more.
By distilling artificial intelligence into component challenges, Scharre demystifies what is a widely recognized but little understood concept. He weaves in enough of the technical details to illustrate the complexities of artificial intelligence, but stays out of the rabbit hole of technological futurism and the minutiae of deep learning models and neural networks. Interesting though these are, they are not relevant for the people that need to read this book—policymakers, legislators, and the general public.
Indeed, Scharre and his publicists would do well to purchase copies for every member of the House and Senate, one for every governor, and one for the president, vice president, and the cabinet. Reading “Four Battlegrounds” won’t make any of these figures an expert in artificial intelligence, but it will help inform these policymakers and shape the debates that are already happening, and that must happen in the months to come—yes, months.
Scharre writes with a sense of urgency that making the right decisions today and tomorrow are of paramount importance. Investments and decisions in this Congress will have outsized effects tomorrow for both the country, but also America’s strategic competition with China. Indeed, Scharre’s four titular battlegrounds are the areas where policymakers can have a demonstrable, positive effect. As Scharre writes, this is not like the space race, GPS, or the internet. The development of this technology is largely, if not wholly, outside the hands of the government. Yet, government will have a critical role to play to ensure that America’s leadership in AI is maintained.
Woven throughout this narrative are vignettes as to how China and the United States are applying artificial intelligence in theory and in practice, and where the strengths of each lie. In this, Scharre echoes a key theme that has emerged in much of the discourse about strategic competition—that it is as much about systems than countries, a contest of democracy versus authoritarian capitalism. That systemic challenge has definitive impacts in the approach to, and application of, artificial intelligence. China’s authoritarian system is applying AI to control its population, stifle dissent, and erase the Uyghurs. It does constrain innovation, but perhaps not to the degree that many in the West would expect. America, by contrast, encourages innovation with its openness, the free movement of capital, and the incentivization of market capitalism. The greatest obstacle in the introduction of AI into America’s defense and security architecture isn’t a lack of will or interest, but the bête noire of so many in Washington: the acquisition system. While reforms are underway, they still have a long way to go.
The AI ecosystem bridges both systems and both countries and is arguably far more interconnected than many in Washington appreciate. All four of Scharre’s battlegrounds overlap, something policymakers are working to unpack, particularly in semiconductors with the recently passed CHIPS Act. The efficacy of efforts to constrain China’s access to semiconductors and net generation fabrication technology remains to be seen. Scharre quite astutely observes, “Like a drunk in a barfight, the United States threw the first punch in a global chip war with no plan for how to finish the fight.”
On China’s use and application of AI, Scharre is far more measured than other commentators, and his book is stronger for this restraint. While he rightly shows the appalling extent of China’s use of technology to suppress the minority populations, he notes that Beijing’s surveillance is—at the time of writing—far less of a panopticon than many believe. China’s databases are scattered and disconnected. There is not one social credit system. Rather there are many disconnected systems monitoring the behavior of China’s citizens. Could the Chinese Communist Party achieve the all-seeing-eye of Sauron? Certainly, and that is the direction it appears to be heading.
The United States, for its part, struggles with the principles and practices of AI and the private sector’s relationship with the government. Scharre recounts the controversy over Project Maven, a military project that was ultimately undone because of protests by engineers within the participating companies over working with the Department of Defense. Their protests, while perhaps valid, were undone by the same company’s decision to expand cooperation in China. The United States is leading the effort on developing ethical AI systems and behaviors, but there is a risk that China or Russia fields systems without those same scruples or oversight mechanisms.
The pace of Scharre’s book manages to accelerate, which is an impressive feat. Having spent the preceding sections exploring the components of AI and the current practical implications of the tool, in its fifth and final section, “Four Battlegrounds” dives deeply into the theoretical implications of AI in warfare and in security. Artificial Intelligence is not the be-all, end-all tool some acolytes proclaim. As Scharre writes, in its current form, “AI has many powerful uses, but it comes with a price: the brittleness, narrowness, possibility of surprises, and the lack of explainability that characterize many AI systems…”
Defense planners may take issue with suggestions that AI will change the nature of war—its character, almost certainly, but war will almost certainly be the use of force (digital or otherwise) to compel an adversary to one’s will. While Scharre generally does a good job tamping down breathless exhortations about AI’s utility, some do slip through: One AI designer suggests that AI could develop “safe” or “non-exploitable” strategies, ones that do not have flaws adversaries could expose. This is techno-hubris at its finest. One would do well to keep in mind Mark Twain’s observation in “A Connecticut Yankee in King Arthur’s Court:”
The best swordsman in the world doesn’t need to fear the second best swordsman in the world; no, the person for him to be afraid of is some ignorant antagonist who has never had a sword in his hand before; he doesn’t do the thing he ought to do, and so the expert isn’t prepared for him; he does the thing he ought not to do: and often it catches the expert out and ends him on the spot.
As Scharre amusingly writes, it only took some Marines hiding (and giggling) in boxes (a la Solid Snake in the video game “Metal Gear Solid”) or somersaulting across a street to bypass an AI-powered camera. Even “poisoning” AI systems by introducing false data or confusing its algorithms could undermine the tool’s efficacy. In many ways, AI is its own battlespace. For as much as it will augment existing operations, it will be both an actor and a target. It is getting better to be sure, but it is not yet an epochal game-changer. The real impact AI will have on the battlefield is its marrying with legacy platforms and use in novel ways. Here, Washington must be wary of rushing AI and AI-enabled systems to the front in a “race to the bottom,” as Scharre notes. Equally, coming to rely on these systems introduces its own vulnerabilities, both anticipated and unanticipated. AI is a tool, not a panacea.
Scharre’s book is easily a contender for a best book of the year award. He jettisons the fixation on the technology (and takes overused metaphors to task) to focus on the implications of the technology, and in so doing provides readers with a much clearer understanding of what artificial intelligence means both in theory and practice. This is not an Elon Musk-ian end-of-the-world prognostication. It is a grounded, measured, and well-argued deconstruction of artificial intelligence in the real world. This doesn’t make it any less concerning or alarming, but it provides a rationale for righteous concern about what getting artificial intelligence policy wrong means for America. Most importantly, he provides a framework for policymakers to understand AI, which elevates “Four Battlegrounds” beyond merely an interesting book to one that must be read.
a global affairs media network
Begun, the AI Wars Have
Photo by Joshua Sortino on Unsplash
March 4, 2023
AI will be at the center of geopolitical competition moving forward-we are entering an era of "AI Wars." In his review of Paul Scharre's "Four Battlegrounds," Joshua Huminski suggests government leaders read books such as this to learn what they must so we can survive and thrive in this new era.
A
dmitting one doesn’t know something is practically heresy in Washington, D.C., especially if one is an elected official. This norm made it all the more impressive—though it should not have—when it emerged that Rep. Don Beyer was pursuing a degree in machine learning from George Mason University. The Democrat from Virginia hopes to learn more about the concept to better inform his legislative priorities and, ultimately, deliver better results for his constituents. Regardless of where one stands on Beyer’s politics, his efforts to learn more about what is and will be the most transformative technology of the 21st century is laudable and is worthy of emulation.
Beyer’s return to school couldn’t have come at a more opportune time. In November 2022, OpenAI released ChatGPT, an artificial intelligence chatbot that took the internet by storm, offering both promise and peril, depending on how one interpreted the tool. The tool scored early notoriety for passing law and business exams, becoming a tool for rapid marketing content generation, and even declaring its love for one user. The tool, trained on unfathomably large data sets, is not foolproof. When asked to do rocket science it, in the words of NPR, “crashed and burned.” It is, of course, early days for ChatGPT. Advocates extol the virtues of how it will revolutionize the way we live, work, and love; critics contend it is an overgrown parlor trick, merely regurgitating (and very well) what data it ingests.
It is often said, and rightly so, that technology outpaces the ability of policymakers to come to grips with both it and its implications. Beyer’s pursuit of a graduate degree and the arrival of ChatGPT vividly illustrate this delta. Indeed, while AI is not at the level of C3PO or R2D2 (yet), another resident of the Star Wars universe may well opine that “begun, the AI wars have.” This is a central tenet of Paul Scharre’s absolutely superb book on artificial intelligence, “Four Battlegrounds.” Artificial intelligence is already turbocharging the challenges policymakers face in managing geopolitics, security, and military affairs—it is a state of affairs that will only get worse. Perhaps most alarmingly, the consequences of getting AI policy wrong today could well determine the balance of power in the years to come.
Scharre smartly deconstructs the complexities of artificial intelligence, choosing not to focus on the ins and outs of the algorithms, but on the four titular battlegrounds: data, computing power (or “compute”), talent, and institutions. He argues, quite rightly, that competition in each of these will define the leaders of the artificial intelligence race, and quite possibly the winners of strategic competition. It is an inspired way of illustrating and addressing the panoply of challenges AI touches upon. Data deals with privacy, among other things. Computing power directly affects semiconductor supply chain security and resource consumption. Talent vividly shows the challenges of immigration and the balance between openness and security. Institutions, above all, touche on acquisition, ethics, and much more.
By distilling artificial intelligence into component challenges, Scharre demystifies what is a widely recognized but little understood concept. He weaves in enough of the technical details to illustrate the complexities of artificial intelligence, but stays out of the rabbit hole of technological futurism and the minutiae of deep learning models and neural networks. Interesting though these are, they are not relevant for the people that need to read this book—policymakers, legislators, and the general public.
Indeed, Scharre and his publicists would do well to purchase copies for every member of the House and Senate, one for every governor, and one for the president, vice president, and the cabinet. Reading “Four Battlegrounds” won’t make any of these figures an expert in artificial intelligence, but it will help inform these policymakers and shape the debates that are already happening, and that must happen in the months to come—yes, months.
Scharre writes with a sense of urgency that making the right decisions today and tomorrow are of paramount importance. Investments and decisions in this Congress will have outsized effects tomorrow for both the country, but also America’s strategic competition with China. Indeed, Scharre’s four titular battlegrounds are the areas where policymakers can have a demonstrable, positive effect. As Scharre writes, this is not like the space race, GPS, or the internet. The development of this technology is largely, if not wholly, outside the hands of the government. Yet, government will have a critical role to play to ensure that America’s leadership in AI is maintained.
Woven throughout this narrative are vignettes as to how China and the United States are applying artificial intelligence in theory and in practice, and where the strengths of each lie. In this, Scharre echoes a key theme that has emerged in much of the discourse about strategic competition—that it is as much about systems than countries, a contest of democracy versus authoritarian capitalism. That systemic challenge has definitive impacts in the approach to, and application of, artificial intelligence. China’s authoritarian system is applying AI to control its population, stifle dissent, and erase the Uyghurs. It does constrain innovation, but perhaps not to the degree that many in the West would expect. America, by contrast, encourages innovation with its openness, the free movement of capital, and the incentivization of market capitalism. The greatest obstacle in the introduction of AI into America’s defense and security architecture isn’t a lack of will or interest, but the bête noire of so many in Washington: the acquisition system. While reforms are underway, they still have a long way to go.
The AI ecosystem bridges both systems and both countries and is arguably far more interconnected than many in Washington appreciate. All four of Scharre’s battlegrounds overlap, something policymakers are working to unpack, particularly in semiconductors with the recently passed CHIPS Act. The efficacy of efforts to constrain China’s access to semiconductors and net generation fabrication technology remains to be seen. Scharre quite astutely observes, “Like a drunk in a barfight, the United States threw the first punch in a global chip war with no plan for how to finish the fight.”
On China’s use and application of AI, Scharre is far more measured than other commentators, and his book is stronger for this restraint. While he rightly shows the appalling extent of China’s use of technology to suppress the minority populations, he notes that Beijing’s surveillance is—at the time of writing—far less of a panopticon than many believe. China’s databases are scattered and disconnected. There is not one social credit system. Rather there are many disconnected systems monitoring the behavior of China’s citizens. Could the Chinese Communist Party achieve the all-seeing-eye of Sauron? Certainly, and that is the direction it appears to be heading.
The United States, for its part, struggles with the principles and practices of AI and the private sector’s relationship with the government. Scharre recounts the controversy over Project Maven, a military project that was ultimately undone because of protests by engineers within the participating companies over working with the Department of Defense. Their protests, while perhaps valid, were undone by the same company’s decision to expand cooperation in China. The United States is leading the effort on developing ethical AI systems and behaviors, but there is a risk that China or Russia fields systems without those same scruples or oversight mechanisms.
The pace of Scharre’s book manages to accelerate, which is an impressive feat. Having spent the preceding sections exploring the components of AI and the current practical implications of the tool, in its fifth and final section, “Four Battlegrounds” dives deeply into the theoretical implications of AI in warfare and in security. Artificial Intelligence is not the be-all, end-all tool some acolytes proclaim. As Scharre writes, in its current form, “AI has many powerful uses, but it comes with a price: the brittleness, narrowness, possibility of surprises, and the lack of explainability that characterize many AI systems…”
Defense planners may take issue with suggestions that AI will change the nature of war—its character, almost certainly, but war will almost certainly be the use of force (digital or otherwise) to compel an adversary to one’s will. While Scharre generally does a good job tamping down breathless exhortations about AI’s utility, some do slip through: One AI designer suggests that AI could develop “safe” or “non-exploitable” strategies, ones that do not have flaws adversaries could expose. This is techno-hubris at its finest. One would do well to keep in mind Mark Twain’s observation in “A Connecticut Yankee in King Arthur’s Court:”
The best swordsman in the world doesn’t need to fear the second best swordsman in the world; no, the person for him to be afraid of is some ignorant antagonist who has never had a sword in his hand before; he doesn’t do the thing he ought to do, and so the expert isn’t prepared for him; he does the thing he ought not to do: and often it catches the expert out and ends him on the spot.
As Scharre amusingly writes, it only took some Marines hiding (and giggling) in boxes (a la Solid Snake in the video game “Metal Gear Solid”) or somersaulting across a street to bypass an AI-powered camera. Even “poisoning” AI systems by introducing false data or confusing its algorithms could undermine the tool’s efficacy. In many ways, AI is its own battlespace. For as much as it will augment existing operations, it will be both an actor and a target. It is getting better to be sure, but it is not yet an epochal game-changer. The real impact AI will have on the battlefield is its marrying with legacy platforms and use in novel ways. Here, Washington must be wary of rushing AI and AI-enabled systems to the front in a “race to the bottom,” as Scharre notes. Equally, coming to rely on these systems introduces its own vulnerabilities, both anticipated and unanticipated. AI is a tool, not a panacea.
Scharre’s book is easily a contender for a best book of the year award. He jettisons the fixation on the technology (and takes overused metaphors to task) to focus on the implications of the technology, and in so doing provides readers with a much clearer understanding of what artificial intelligence means both in theory and practice. This is not an Elon Musk-ian end-of-the-world prognostication. It is a grounded, measured, and well-argued deconstruction of artificial intelligence in the real world. This doesn’t make it any less concerning or alarming, but it provides a rationale for righteous concern about what getting artificial intelligence policy wrong means for America. Most importantly, he provides a framework for policymakers to understand AI, which elevates “Four Battlegrounds” beyond merely an interesting book to one that must be read.
