Book review: ‘Power and Prediction’

Not many authors start their new book with the frank admission that their previous one was fundamentally in error. “We were wrong,” say Ajay Agrawal, Joshua Gans and Avi Goldfarb in their preface to ‘Power and Prediction’ (Harvard Business Review Press, £22, ISBN 9781647824198), a compelling essay on the disruptive economics of artificial intelligence. But you can’t really blame them for partially missing the mark in 2017’s ‘Prediction Machines’ – some might say they’re being too hard on themselves here – because their starting position was solid enough: technologies will always evolve, while sturdy and reliable economics will obey the same rules it always has done.

Again, to be fair, ‘Prediction Machines’ came out five years ago and a lot has happened in the world of AI since then. While the economic model the authors outlined ‘remains useful’, half a decade later AI has moved on. There’s an emerging and critical new chapter in the technology story to tell, say the authors, especially related to the businesses of fintech, pharmaceuticals, automotive and retail.

Because AI is a transformative technology, there is inevitably a phase between initially realising its potential and witnessing its widespread adoption. This is where we are today, say the authors, who detect an uneven timeline on which some industries integrate larger AI systems quicker than others. These are the winners who gain the ‘power’ of the book’s title in their markets. The downside to this power is the ‘AI bullwhip’ that can create resonating effects in supply chains from seemingly small movements upstream. The ‘prediction’ part is more to do with the concept of ‘ungluing’ an entrenched corporate culture from its pre-AI rule systems. We need to understand that there is uncertainty in the ecosystem, while creating space for the AI to do its job – which is essentially to make predictions based on combined multiple decisions.

The authors contend that by quickly reaching an understanding of how and when to apply AI, younger organisations find themselves more able to gain a competitive edge over older corporations that can’t unglue themselves from their self-imposed inertia.

Where the economics kicks in is when replacing old systems rewards technological innovation. It’s a ‘fortune favours the brave’ moment because AI creates advantage for early adopters, while venture capitalists queue up to invest in agile start-ups that have bought into the AI credo. Those watching from the sidelines, weighed down by the fear that we’re handing over power to the machine, are missing the argument that a well-integrated AI leads to better decision making, while humans remain in control.

When it comes to the all-pervasive fear of AI bias, there was plenty of that going around before we came up with algorithms to help us detect and reduce it. All in all, ‘Power and Prediction’ is a timely and insightful follow up to ‘Prediction Machines’.