Sainsbury's lead architect: 'The next time somebody mentions AI ask them what they're really talking about'

'AI is a meaningless term, it doesn't say anything about anything'

Retailers are investing heavily in AI, from warehouses run by robots, to automated delivery vehicles, to solar-powered mobile stores, to systems in China and South Korea that allow people to ‘pay with their faces' through the use of facial recognition software.

AI is frequently spoken about in terms of revolution, but says Udai Chilamkurthy, lead architect retail and logistics at Sainsbury's, most of its real use cases are evolutionary. General-purpose AI doesn't yet exist, and current applications are extremely task-specific. So it really a case of existing processes being tweaked to add a little more intelligence.

"AI in some senses is a meaningless term, it doesn't say anything about anything," he said during a talk at the Connected World Summit in London last week.

"The next time somebody mentions AI ask them what they're really talking about, and by the time you've reached the third question you'll see they literally have no idea, or they have an idea that might make sense for a specific use case."

Many AI innovations are being developed by small companies and startups. Whereas something like a chatbot is reasonably generic and can be integrated into retailers systems and start adding value relatively quickly, much of the opportunity talked about AI is to do with fundamental change and this is much more difficult to achieve. There is no way that a startup could fully understand the intricacies of the retail sector in which thousands of processes have been tweaked over the years to ensure they make money in a sector where profit margins are low.

"Let's say someone wants to use AI to optimise the supply chain. That doesn't even mean anything. You have to think about what the exact problem is in the supply chain, what is the intelligence you want to gather that your team is not able to gather at this point, how are you going to do it and what will be the wider consequences?"

So AI is a technology that is not readily 'productised' and it makes little sense to have teams dedicated to it, chasing the new and the shiny.

AI is not just another business unit, it's not just another product

"AI is not just another business unit, it's not just another product that you're buying to plug into your organisation as a separate entity," Chilamkurthy said.

Indeed, adding bits of automation here and there is unlikely to achieve the hoped-for benefits in terms of efficiency and productivity and may even unbalance existing systems that work perfectly well as part of the finely tuned machine. To truly live up to its promise, AI must be built into the foundations of the organisation and be allowed to develop from there, he argued.

"How do we make it much more fundamental from the ground up in a more sustainable way so we can make improvements on an ongoing basis? Because that's what AI is: it improves and develops which means you get continuous feedback about what's happening and what it is you should be doing."

Such fundamental interventions raise questions about long-term funding, skills, and the likely effects of turning over aspects of decision-making to machines.

"Using some of the AI tools means that you're shifting away from something precise to something that is probabilistic," said Chilamkurthy, raising the spectre of a retailer marketing cigarettes and alcohol to a 16-year-old that its AI systems had identified as being 21. How many organisations are willing to change their operations from one based on binary decisions made by people to statistically-based ones made by machines?

Chilamkurthy did not go into what Sainsbury's is doing in AI except to say the supermarket is investing significantly in a number of areas. Retailers cannot afford to sit still for long and are always looking over their shoulder at what players like Amazon are doing.

But, he said, the umbrella term 'AI' hinders more than it helps. Really, it is just using new techniques to solve age-old problems, and sometimes the terminology gets in the way.

Artificial Intelligence and Machine Learning Live 2018 - Monday 19 November 2018

What can artificial intelligence and machine learning do for you and your organisation? If you don't know yet, or want to make sure that you're not missing out, Computing's first AI & Machine Learning Live event is for you. To find out more, check out the Computing AI & Machine Learning Live website. Attendance is FREE to qualifying IT leaders and senior IT pros, but places are going fast