ANP
NOS Nieuws•vandaag, 20:33
What is the effect of superior artificial intelligence on human decisions? Those decisions are getting smarter and more creative, a new study shows. At least, when it comes to the board game Go. The successes of the AlphaGo computer program appear to have inspired human Go players to perform better.
For their study, the researchers, including the Dutchman Bas van Opheusden, analyzed 5.8 million moves that occurred in Go games between 1950 and 2021. After 2016, when AlphaGo also started beating the best human players, the quality of human moves clearly went up.
This was partly about moves and strategies that the players had copied from AlphaGo, but they were also increasingly introducing novelties: new, innovative moves that had not previously appeared in the game of Go. Not only were those moves new, they were also of a high level on average.
From Go to poker
“Superhuman artificial intelligence can improve human decision-making through an increase in novelty,” the scientists have broadly captioned their study. This raises the question of whether artificial intelligence can also have this effect in other situations. Does it make us more creative in a broader sense?
“I suspect that our results apply to a large number of board games and games of chance,” says Van Opheusden. “In poker it has led to the introduction of new strategies, such as the so-called overbet.” In such an overbet, a player makes a bet that is larger than the existing pot.
AI has also led to new strategies in poker.
Researcher Bas van Opheusden
Whether the effect also applies to the latest hype in the field of artificial intelligence, ChatGPT, is something he is less sure about. “In a board game, the outcome is unambiguously defined, and AlphaGo’s moves are objectively better. For generative AI like ChatGPT and Dall-E, success is defined in terms of human judgments: how beautiful is this picture, and how helpful is this chatbot? ” How such programs influence human decisions is therefore more difficult to predict and investigate.
Better or ‘better’?
Philosopher of technology Giulio Mecacci of Radboud University thinks it’s an excellent study, but also sees those human judgments as an obstacle. “Many political and ethical decisions are based on human values and preferences. It is difficult to say which decision is ‘better’ in such an area, and whether artificial intelligence can help.”
That is also not the kind of decision-making that the researchers will focus on in the near future. They first want to investigate whether they also find the same effect in other board games, and how exactly this is achieved. “The increasing availability of superhuman AI systems opens up exciting new fronts for the study of human cognition,” they conclude their article.