A human player has recently beaten KataGo, a top-ranked Go AI computer programme.

Kellin Pelrine, a research scientist intern at FAR AI, is one level below the top amateur ranking, but he was able to beat his machine opponent, an open-source computer Go program, in 14 out of 15 games, by using suggestions from another computer program (The Centaur Strategy) designed by his company.

FAR AI has made a program by playing over a million games to find a blind spot that even intermediate players could exploit.

Our adversaries do not win by learning to play Go better than KataGo — in fact, our adversaries are easily beaten by a human amateur. Instead, our adversaries win by tricking KataGo into making serious blunders. Our results demonstrate that even superhuman AI systems may harbor surprising failure modes.

The FAR AI strategy is not too different to the (implicit) strategy used by Alpha GO to defeat the then World Champion Ke Jie, who declared:

After humanity spent thousands of years improving our tactics, computers tell us that humans are completely wrong. I would go as far as to say not a single human has touched the edge of the truth of Go.

It is the strategy used by Paris to defeat Achilles, finding the weak spot, the heel of your adversary.

The CEO of FAR AI, told The Financial Times that it’s common to find flaws in AI systems using adversarial attacks. Yet, big AI systems continue to be released without adequate protection. This vulnerability has already appeared in some of the most popular AI tools right now, such as ChatGPT, which would break when prompted with anomalous tokens.

By chatting and playing against AI programmes we are in a new war stage, a mutual adversarial refinement of our own skills and capabilites.

Who will win? Like in most wars, very likely the answer won’t be so clear cut.