[ad_1]
In a sport where the smallest design elements could be the deciding factor, Team Emirates New Zealand turned to McKinsey to gain a competitive edge in their bid to retain the America’s Cup.
When the first America’s Cup was awarded in 1851, sailors could hardly have imagined that at the 36th edition of the event, its competing crews would be supported by an army of intelligent robots. But when Emirates Team New Zealand sets out to defend the Cup in the waters off Auckland this week, its meticulous preparations will be aided by an advanced AI robot developed in partnership with world-leading consultancy McKinsey & Company.
Without in any way diminishing the exceptional abilities of the highly talented and dedicated sailors, it is fair to say that the current America’s Cup is fundamentally one of innovation in high-tech design. McKinsey Senior Partner Brian Fox, a competitive sailor himself, puts it succinctly; “Every boat in the America’s Cup is designed with a computer simulator. Whichever team has the best simulator and uses it most effectively gains the advantage.”
During the long run-up to the event, Team New Zealand and Cup challenger Luna Rossa Prada Pirelli of Italy came together to establish the boat’s design parameters and stock components for the 36th America’s Cup to ensure a certain level of sporting competitiveness between the two 75ft vessels. while leaving key areas in their designs open to tuning. The release of ‘Class Rule’, almost three years ago now, could essentially be considered the first shot from the starting gun.
Not just another crew member
Shortly after, the New Zealand team drafted in McKinsey to help in the hunt for its latest ultra-human crew member – one “that could sail thousands of ships at once”. With often millions of dollars being spent to reduce performance times by milliseconds—these are, after all, the fastest monohulls in the world—prototyping and testing potentially revolutionary design elements isn’t cheap, nor is the process particularly fast when constrained by budgets and human resources. restrictions. .
One area where the 2021 class rule allowed tinkering was in the ships wings, the mechanism that lifts the entire craft out of the water and allows speeds of up to 100 kilometers per hour. In the words of McKinsey, if this opportunity to adjust is used well, it can translate into a huge advantage during the day. The consultancy’s role would be to maximize Team New Zealand’s hydrofoil testing capacity, thereby enabling greater innovation.
“The simulator was key to the team’s victory in 2017; sailors used it to test new ship designs without actually building them,” notes the firm. “But this simulator required multiple team members to use it simultaneously to function properly. With the sailors’ scheduled training, travel and competitions, it was a logistical challenge.” McKinsey’s solution was to build an AI robot that could independently run design models through a simulator.
The first step was to bring in a team from McKinsey’s QuantumBlack analytics arm, with data analytics and machine learning experts from Australia and the UK, who built the initial infrastructure to migrate and run the Team New Zealand simulator in the cloud. From there, things start to get particularly technical, with the firm then taking an innovative new approach known as “deep reinforcement learning” to essentially teach the AI robot how to become a professional sailor.
“This technique allowed the robot to learn dynamically and gain greater accuracy through continuous feedback,” explains Nic Hohn, QuantumBlack’s chief data scientist for Australia and co-leader of the project. “When you start, the AI agent knows nothing and learns through trial and error using countless variables – and is refined over and over again. Since the robot is constantly experimenting, if you teach it to learn in the right way, it will compress into hours that would take a human years to understand.”
While a single robot could replace the intensive work of a huge number of humans, that robot would ultimately benefit from the support of its own team, with McKinsey creating a parallel network of thousands of robots to share information with each other. learned the ropes of high-octane cruising. This collective level of learning at scale was a critical breakthrough, the firm says, dramatically reducing project time and cost.
Deep reinforcement learning
The end result? Within two weeks, the robot mastered the basics of sailing in a straight line and then moved on to more complex maneuvers. A month or so later, she defeated the human sailors and was then able to begin testing different variants of the hydrofoils on her own – in rapid repetition 24 hours a day – without disturbing their plans. Soon the crew will be learning sailing techniques from a robot and now Team New Zealand’s latest entry has been described as “terrifying”.
Ultimately, Team New Zealand’s design process was accelerated tenfold by working with McKinsey. In addition, the successful application of deep reinforcement learning by consultants excited its enormous potential across industries. To emphasize the importance, QuantumBlack co-founder and chief scientist Jacomo Corbo highlights the concept of the complexity of the game tree and the board game Go, famously conquered by Google’s DeepMind technology.
“This is one of the most comprehensive deployments of deep reinforcement learning in the public cloud,” said Corbo. “One way to think about the difficulty of a problem: the complexity of a game tree roughly corresponds to the size of the space one must move through, taking into account the set of possible game paths or the sequence of decisions to be made. Go, an extremely complex board game, has a game tree complexity of 170 – our sailing problem has a game tree complexity of almost 2900.”
[ad_2]