Machine learning allowed the America’s Cup winning Team NZ to more rapidly design and test new features on their boat, and to do so without having the pull teams members away from their preparations.

According to Tim Fountaine, Senior Partner McKinsey & Company, “The problem that we were trying to solve was to help them test many more boat designs. And the reason that that was important, was because the teams don’t have very long to prepare and they have to design their boats and optimise them as best they can in that time.”

He told Which-50, “The more that they can design and test, the better. The other thing is that the race rules mean there’s a cap on how many parts or components of the parts you can build.”

Technology has always sat at the heart of America’s Cup campaigns. For instance, a number of years ago the Team NZ built a digital simulator to test designs of parts before they were produced.

“The simulator simulators really good, but it requires actual sailors to sail it [the boat],”  he said.

McKinsey, which was the technology partner to the team, used reinforcement learning to create an artificially intelligent digital sailor to sail the yacht in a simulation.

The way that the reinforcement learning works is that we define the goal and the algorithm should try and optimise for it. In this case, it was the speed [at which] the yacht can make it to the next mark. And then in simple terms, we let the algorithm test different ways of doing that. And you can see to start with, it doesn’t do a great job. It capsizes the boat in the simulation or it goes really slowly. It goes the wrong direction, and eventually it learns just like a child learning to walk .”

According to Fountaine, “It got better and better each time. It would remember what worked and what didn’t, and it would then take that forward to the next iteration.

Previous post

Employee revolt sees Apple sack sexist within days of hiring

Next post

Colonial Pipeline Paid a $5 million Ransom to Cybercriminals, but What's the Real Cost?