Artificial intelligence in the arms race: Commentary by Avi Ben Ezra
Artificial intelligence is at the epicenter of the arms race, and whoever has superior AI will win.
For most people, the threat of AI has been limited to economic dislocation and the sci-fi robotic apocalypse. Yet, AI advancements are taking place in the private sector, outside governments’ control or scrutiny, and there is speculation that it is quietly being used for defense. Experts believe that a new arms race is developing particularly between United States and China.
The Chinese president realized the power of AI, and its superhuman capacity to think, after AlphaGo defeated the number one Go player. He obviously foresees, like some experts, AI evolving the ability to rewrite its own code in a few years’ time and exploding its IQ as high as 10,000. Humans will be like ants compared to such intelligent giants.
Achieving this artificial superintelligence will require breakthroughs in transformative technology whose circumstances and timing cannot be predicted at present. However, President Xi and other presidents saw the possibilities of AI in the global balance of power when AlphaGo won in Go, a game of strategy.
AI in games
Strategy games come in two types. First, there are games of complete information, such as Tic-Tac-Toe, chess and Go in which players see all the parameters and options of the other players. Such games are generally easily be won with practice. Then there are games of incomplete information, such as Rock, Scissors, Paper, in which players can learn the rules and know the optimal strategy. However, no one is certain how the opponent will play so there is no definite winning strategy and winning is left to chance.
Humans used to win games against computers at one time and there was belief that humans’ ability to think abstractly and to narrow down decision making to a few good choices would always beat the computer. Then in 2016, AlphaGo, an AI system, defeated the Go world champion 4-1. In 2018, AlphaZero, a new AI system with ability to self-learn, defeated AlphaGo 100-0 by accumulating knowledge and inventing unheard-of strategies within 40 days. In 2017, AlphaZero was pitted against Stockfish in 100 games and, within 9 hours of learning the chess game, it won 28 games, drew 72 games and lost none. No chess grandmaster has ever beaten Stockfish yet AI superintelligence beat it.
In 2017, Libratus, another AI system, beat the best poker players in No-limit Texas Hold’em, a poker game. In 2019, Pluribus beat multiple top professional poker players at the rate of 5 big blinds per hour. Poker is a game of incomplete information, uncertainty, and complexity that combines strategy, mathematics, psychology, timing and luck.
By the beginning of 2020, AI had beaten all human players and the best computer programs ever designed.
AI and the arms race
Avi Ben Ezra the CTO of the SnatchBot chatbot platform says “It is normal that most analysts talk about the US and China, but actually, with the military chatbots that we created – you have in excess of 40 countries who tackle a range of issues from information warfare to cybersecurity and fraud detection with clever AI chatbots that are integrated with robotic process automation (RPA)”.
Poker mimics life because of uncertainty. In the US-China rivalry, China’s objective is to replace America as the dominant superpower. It knows America’s defense budget, force development plans and probably its military resources, capabilities and specifications to a certain degree. However, America’s alliances keep shifting, its capabilities and projects are classified and international crises are unpredictable. Therefore, the best that China can do is to invest optimally in order to exploit America’s weaknesses while managing its own risks and weaknesses. The US will do the same. Both countries’ defense planners are compromised in their outcomes by bureaucracy, internal rivalry, politics and vested interests. There’s obviously a lot of uncertainty.
Since AI beats the best humans in poker, its capability is obviously being tested in defense. In a few years, AI systems will be making all military decisions as generals that never tire, have no fear, are never distracted, and always perform at their peak. No human decision makers can compete.
Under those circumstances, the country with slightly worse AI will lose every battle and the winner will control AI. No one knows how it’s going to play out, but it’s certain that AI will lead the arms race as each nation places it at the very core of national achievement.
Bringing AI and RPA together from a military perspective is just like with any other organization: it improves efficiency and drives down cost. Yet the key issue is obviously that maintaining a technological edge, is at the heart of the strategy for several opposing players in the game.