Utility Week

20 | 27TH APRIL - 3RD MAY 2018 | UTILITY WEEK Operations & Assets Analysis A n explosion of disruptive technologies are propelling humans into a second machine age that will fundamentally alter the way we live and work. This radical transformation – the likes of which hasn't been seen since the industrial revolution – means the playbook for business must be completely rewritten. That was the message when Massachu- setts Institute of Technology (MIT) professor Andrew McAfee took to the stage at Accen- ture's International Energy and Utilities Con- ference in Paris earlier this month. "We have been walking around for a long time now with a basic guideline – a basic rule – about how to divide up all of the work that needs to be done", McAfee told delegates. According to this mantra, computers are good at maths: "Let them do the math. Let them do the arithmetic. Let them keep all this boring routine work which is important but we're not very good at as human beings. "Then all of us human beings are free to do the things that we are good at and, in fact, the things where we cannot be replaced. And that has to do with judgement; our experience; our intuition. In America, we say our gut. "The computers are bad at that. If we gave them the judgement calls, they'd fail terribly". Hippos versus geeks McAfee, a principal research scientist at the MIT Sloan School of Management, said evi- dence of this thinking can be easily found in most book shops. "When CEOs retire they love to write books. And the book that they love to write is about how they made this very tough decision straight from the gut. "No CEO has ever written a book called: 'The spreadsheet that made a tough decision'." Speaking to an audience of top-ranking executives from around the globe, McAfee noted, much to their amusement, that the majority of businesses make decisions on the basis of the "highest paid person's opinion" – hippo for short. "It really is how most companies make most of their decisions, most of the time," he remarked. "What's interesting is that on the team there is very oen a group doing number-crunching; doing analysis. And they present the results of their analysis to the hippo." McAfee said these the number-crunchers – the geeks – are the "mortal enemy" of the hippo: "In this context, the geek is some- body who follows the evidence." "That sounds easy," he added, "but here's the difficult part: if the evidence doesn't agree with their experience – their gut, their instincts – a geek walks away from their own judgement and follows the evidence. "A hippo will do exactly the opposite… If this analysis doesn't agree with it, I don't like the analysis too much". McAfee said there is growing body of evi- dence on how the geeks perform versus the hippos, and the results so far show the geeks in a clear lead. He cited a research paper examining 136 peer-reviewed studies which made a direct head-to-head comparison between the two styles. These studies looked at important, consequential decisions: "What medical condition does this patient have? …Which of these people should we hire into our com- pany? Which of these suppliers would be a good partner for us to work with?" The paper found that in only 6 per cent of cases were hippos found to make bet- ter decisions, while 46 per cent of the time they fared worse. Hippos should become "an endangered species", McAfee concluded. He noted that these studies were all con- ducted in an era of small data. "Everybody in this room knows we do not live in the era of small data anymore," he added. "We live in the era of huge data. "In fact, in many cases it feels like the amount of data out there is absolutely over- whelming and we're lost in it." The birth of artificial intelligence and machine learning provides a solution to this information overload, as McAfee explained: "What this means very simply is that we have a set of tools that are unbelievably good at reading huge amounts of information, see- ing the patterns and getting insights from them. "In other words, these tools are making the gap between hippo and geek even bigger than it has been up until now." To the doubters, he offered a fascinating example of just how far artificial intelligence has come in the past few years. The game of Go was created in China more than 2,000 years ago – perhaps as many as 4,000 – and by McAfee's reckon- ing is yet to be surpassed as a test of "pure strategy". "If you play Go, you think that chess is kind of cute," he said. "When you're ready to play with the adults, then you can come play Go with us". While Deep Blue – a computer developed by IBM – became the world's best chess player in 1997, until fairly recently machines were "laughably bad" at playing Go. A sur- vey of researchers at an artificial intelligence conference in 2015 found the average esti- mate of when a computer would outwit the world's best Go player was 2027. However, this milestone was reached just one year later, when Deep Mind's artificial intelligence Alpha Go went up against Lee Seedol – a player known as a once-in-a-gen- eration talent akin to football's Lionel Messi or basketball's Lebron James. He was expected to be extremely dif- ficult to beat "because his style of play was so creative and insightful and unpredictable and intuitive… He played like a very serious hippo". Alpha Go won in four games out of five, with the match highlight coming on the 37th move in the second game. The machine made a decision that no decent human player would make – a fact which Alpha Go itself recognised. It calcu- lated the likelihood of a human competitor doing the same at just 1 in 10,000. McAfee said it turned out to be a winning decision: "What we think we're learning is that Alpha Go was able to see enough possi- ble ways that the game could unfold over a long period of time to realise that stone could be very valuable by the end of the game." The new machine age Computers crunch the data while humans make the decisions, right? Wrong, declared MIT professor Andrew McAfee at a utilities conference in Paris last week. Tom Grimwood was in the audience.

• Cover