Four Battlegrounds: Power in the Age of Artificial Intelligence

Like the author's previous book 'Army of None' the author has the ability to explain complex technology and the related issues to laypersons. Every policy maker and interested citizen should read this book. And to paraphrase Trotsky, "You might not be interested in Artificial Intelligence, but Artificial Intelligence is interested in you". Business people should take a look too.The section on US government procurement was interesting too. In a multilateral organizations these barriers to adoption are multiplied x 10.For multilateral peacekeeping organizations (UN, AU, EU, NATO, etc) the main benefits I can see coming from artificial intelligence are in logistics, fleet maintenance and early warning intelligence. The multilateral organizations are not well funded (the whole of UN peacekeeping is funded less than the Sanitation Department of New York City). So they will be dependent on goodwill of the richer member states to supply AI products and expertise. These concepts are already in a embryonic form in things like FRIM (Field Remote Infrastructure Monitoring) run out of UNGSC which is implementing 'smart city' type technologies to create energy efficient 'smart camps' for peacekeepers.Right wing 'Black Helicopter' fantasies in the US notwithstanding, the UN is NOT going to be using lethal drones or other forms of lethal targeting anytime soon. NATO is probably the exception here but that would be down to its member states to provide.Hopefully the P5 and other technologically advanced countries will see it in their self interest to fund this to improve the effectiveness of multilateral peacekeeping and to reduce costs.

If you keep up with the geopolitical issues of contemporary East Asia, you find that at some point, many of the individual issues crossflow through each other. The same goes for books and research where authors initially inspect topics like semiconductors, Artificial Intelligence (AI), and human rights and then, taking cues from geopolitical interests, generate more general, multi-discipline works. Paul Scharre’s most recent “Four Battlegrounds: Power in the Age of Artificial Intelligence” is one such example.Initially having researched the theory, moral responsibility, future tech and military application of AI (2018), Scharre widened his latest book’s content to cover how AI is already being used across a spectrum of economic, political, surveillance, security and military applications. The book jumps on the new cold war research bandwagon comparing and contrasting how China and America’s respective government models influence AI advances.Scharre hasn’t changed his overall opinion of the risks in future AI weapons applications. He is really just bridging the gap between theory and application. There is still a scary future in AI autonomous weapons if not pursued prudently. In all the hype of future “what ifs”, where Scifi entertainment never quite matches with eventuality, Scharre brings his readers back to a contemporary reality in nascent AI application. Scharre’s thesis on how AI will continue development depends on four factors that governments and organizations must consider: data, computing, talent and institutions.The US and China have been co-dependent, in a way, on AI research. Less than one third of AI talent in the US, according to Scharre, completed their undergrad in the US. In terms of the four battlegrounds, China has surpassed the US in most indicators. On the one hand, China’s AI-enabled surveillance tech is not hampered by human rights considerations, allowing China’s wide network of institutions to research, mobilize, test and weaponize data sets that would be considered invasive and illegal elsewhere. This applies directly to facial recognition data sets that global Chinese companies obtain from as far away as Africa and as close as Xinjiang. The acquisition of the data sets that some hail as a simple competitive advantage over other AI firms, is condemned by others, like Scharre, who emphasize the use of the data sets for repressive surveillance.Scharre’s work is commendable for bringing the AI discussion down from apocalypse-sphere and to the everyday reality and practical application of current AI tech, highlighting still the vast limitations to AI and the practical importance of maintaining humans in the loop. For instance, while AI has made strides in object recognition, self-driving cars and strategy game dominance, the first is only achieved after human-assisted inputs of thousands of examples of a single object. China is achieving significant advances in self-driving cars but the achievements in that field are based on literally millions of hours of driving data based on specific rules-of-the-road parameters. Reportedly, while an AI model can now easily beat any human Go champion, all it would take to destroy the AI opponent would be to amend the board dimension. The millions of hours of data in current AI models, still mostly engineered to complete a single task, can be defeated via a simple false line of code or a last-minute adjustment to the rules.And in war, rules are in constant flux. Author’s like Scharre warn that given the promise of AI, the biggest global threat is a race to the bottom; the temptation to put AI on the battlefield where the fog of war would immediately fluster the current level of AI advancement and lead to devastatingly mortal effects for all sides. Humans are able to perceive what is out of place, illogical or missing. New data or a “distributional shift” is lethal to both the AI tool and possibly everything in its vicinity, were it found on the battlefield.All research works in advanced technology, like this one, risk being quickly contradicted due to the rapid and unexpected advances in the science lab and field. Scharre predicted that China’s authoritarian use of AI-surveillance would make mass social protests impossible. However, thousands of Chinese citizens still managed a protest against the CCP’s COVID-19 policies across a number of cities in November 2022, which likely had a significant contribution to Xi Jinping relaxing the zero-COVID policies. That doesn’t mean, however, that the surveillance state was not able to make a large number of arrests after the fact, using the aforementioned surveillance technology.Perhaps the more dangerous advancement in AI is the weaponization of information. AI has both positive and negative inputs to social media use, messaging, journalism, new analysis and strategic communications. To make it worse, Scharre states that “Our brains are hard wired for disinformation.” When we think about the psychology of AI research and application, there are significant policy challenges. There is a fantastic similarity between assessing AI advances and assessing a rival’s capabilities. Competitors often err towards overestimating their rival’s potential. In many ways, AI is contributing, ironically, to an inability to think rationally, analyze information wisely and make prudent decisions.A few critical remarks. Much of Scharre’s data seems to come from pre-COVID era research, possibly even data collected during the research of his earlier book on AI (2018), with only a few geopolitical references to 2021 and 2022 events despite the 2023 publication date. Both books are good reads; not contradictory works but instead complimentary.Despite the impressive advances in programs like ChatGPT, facial recognition or self-driving cars, with the use of AI, humanity may become dumber before we become smarter. AI still lacks in the highest quality areas of political, social and cultural analysis. As Scharre has argued previously, AI is unable to fathom commander’s intent on the battlefield. A war using AI before AI is ready, would be a disaster. And while AI will replace many jobs across disciplines, it is already creating, as expected, more jobs in STEM but will make positions that require quality analysis and diplomatic human interaction in the social sciences even more crucial.

Much is written and more is talked about on this topic of artificial intelligence. Everyone has an idea of what this is, some are right and some are a bit too dramatic. This book sorts all of that out for you. It lifts this veil of mystery surrounding the topic. If you read this book you will get a great understanding of what is going on in this field.This book follows what is occurring by various nations, primarily the US and China. It does not say to go into code tricks or anything quite that elaborate. Through this book you will find Skynet isn't here yet but it is close.

The chief complaint in the one star reviews is that the graphs are blank. I reached out to the author via Twitter, who is aware of the publisher's error. He was incredibly responsive. He even sent me the errata directly (see pictures). I received my copy yesterday with the graphs printed.That said, this is a phenomenal book. Bostrom's Superintelligence shaped the way I think about AI in philosophical terms, while Scharre's Four Battlegrounds laid out the current state of affairs in the race for AI. It profiles competition for data, computing power, talent, and institutions and provides sound policy recommendations to ensure competitiveness in AI. It is a phenomenal follow on to his previous book Army of None.

four key elements; data, computing power, talent, and institutions. Data is a vital resource like coal or oil, Advanced computer chips are the essence of computing power, Talent is about people, and institutions that effectively incorporate AI into their economy, society, and especially their military.

Paul typically puts out great content, that's highly relevant to those in the field. When I look to stay up-to-date on current and emerging tech like AI and autonomous systems, I turn to authors like him. This book has high quality content that's current and informative. Would suggest to anyone looking to get smart in those fields. The book DID have missing graphs in my hardcopy, which was disappointing, but I didn't want to pin that on the author. That aside, great read, 5/5 content.