AI Attacks Won’t Just Be Fast, They Will Be Unpredictable

It’s easy to envision AI-enabled attacks overwhelming human teams by making human decisions faster than humans. But this leaves us with the notion that humans can prevent an attack if they can shore-up the most likely points of entry or just move that much faster. A much more likely, and uncomfortable, scenario is that we can’t even predict the shape of the attack or location it will target. AI will create attacks beyond a human’s imagination. In fact, this reality has already been exemplified in multiple man versus AI showdowns. To highlight this I will focus on one specific game, actually on one specific move.

It was March 2016 and the second game of the historic Go match between one of the world's top players, Lee Sedol, and Google’s AI-enabled Go player, AlphaGo. In the 37th move, the machine made a decision that confused everyone – from the throngs of reporters and Go experts, to the match commentators, to Lee Sedol himself. "That's a very strange move," said one Go expert. "I thought it was a mistake," said another. 

 Despite human perception this was a mistake, Move 37 changed the path of play, and AlphaGo went on to win the second game. When interviewed after the game, Lee Sedol called it "not a human move," but also called the move "beautiful.”

 The Chinese game of Go was invented more than 3,000 years ago and is believed to be the oldest board game still played today. Humans have had thousands of years to perfect their strategy. Yet in a very short period of time an AI-enabled system not only won but did so with moves no human would have imagined, let alone predicted. 

 Move 37 demonstrates the power of modern artificial intelligence and foreshadows the unpredictability of attacks we will see from offensive AI systems. One additional, and even scarier, thought is that Go is a turn-based game, so the AI is forced to operate at the same pace as a human. An AI driven cyber-attack won’t wait for us to make our move, it will move at machine speed.

 So, should we just give up? Of course not, but with AI attacks on the horizon we do need to change our mindset and the model we use to think about defending our digital enterprises.

My next article will propose a potential model we could apply to a watch the data strategy.