I talk to Paul Clarke about education, machine learning, and the art of the mid-life crisis
Fewer people know what we are than you might think,’ says Paul Clarke, the CTO of Ocado. He mentions that they are ‘the world’s largest online-only grocery retailer’ but quickly gets to the prevailing point: ‘also, we are a technology business.’
I like Paul. He talks, I think, like an exercised academic: articulate, brisk, and, at notes, just shy of exasperated. ‘We’re much more of a technology business than people really appreciate. We build everything ourselves. We’re much more like a Google or an Amazon, than a Sainsbury’s or a Tesco.’
They’re also a “platform” business, in the sense that they’re now selling the technology they originally developed for internal use, he explains. ‘Then you start getting into the deep technology side. We’re doing robotics, public and private cloud, machine learning, and artificial intelligence. There’s a wider kind of disruption agenda, if you like, based on the technology.’
The Orchestra
Perhaps he could talk me through Ocado’s most exciting project, to help explain? ‘I suppose the swarm robotics technology that went live in our Andover warehouse at the end of 2016 and which will soon be going live in its big sister warehouse in Erith, is the most exciting project.’
‘It brings together robotics, motion control, swarm-based orchestration, all sorts of different applications of AI and machine learning, sensing and simulation. It’s the culmination of a load of competencies we have acquired over the years.’
‘You want all of these instruments playing in harmony, where now the instruments represent different applications and services.’ AI acts as the conductor. (Switching metaphors, Paul tells me that AI is ‘the one to rule them all’; a reference to The Lord of the Rings.)
‘For our swarms of robots whizzing around, assembling customer orders, there are three levels of machine learning’, Paul explains. First, ‘there’s what’s embedded into the robots themselves. Then there’s the real-time control and orchestration of the robots and the optimisation of the storage underneath them.’
‘Finally, there’s what I call the “robot healthcare” which is where you need to keep each of the robots healthy. For example if a robot’s battery pack is degrading, then it is brought in for a service. That’s done by streaming all the robot data to the cloud where machine learning based analytics keep the swarm in tip-top shape.’
‘I think [AI] is this extraordinary glue that lets you do the exciting things with other technologies...unleashing the full potential of those technologies. As well as the fact that you can use one generation of AI to train the next. It’s going to be an extraordinary enabler or “turbo charger” of all manner of things. And “we ain’t seen nothing yet.”’
Why not?
‘Consumers are still at a very early stage of addiction. Some are kind of intrigued and some have negative sentiments. I think most are intrigued – when an application does something that is smart and unexpected, like work out where you live even when you’ve never told it that, just by looking at your GPS data.’
‘As this wave of smartness washes over us, I think consumers are going to get properly addicted to smartness. I think they’re going to start demanding it all over the place. And I think it’s going to be a source of differentiation across different products. People are going to expect those applications and services to play nicely and smartly with one another.’
The Time for Talking
It’s an interesting premise, and one Ocado has bet on. Their strategy is to be a technology company that does groceries; not a grocery store. It seems like an effortful strategy. But if it’s right, they’re a long way ahead. The gap between companies’ digital competence levels is probably larger than ever.
‘I have a filter on my email for “digital transformation” because I will not read emails about it,’ says Paul, (which, I realise, is why we had previously struggled to get through to him.) ‘Seriously, if you’re not already doing it at scale and you’re still standing around the swimming pool working out how to learn to swim, it’s too late now. It’s going to come so fast and like a tsunami, it will just keep on coming.’
I decide to push back. What takes years for a pioneer takes hours for a settler, I argue. Tech makes tech easier: building a website is easy where once it was hard. You don’t even have to be very clever to build a cryptocurrency any more. It simply wouldn’t be intelligent or economical for most companies to take a route like Ocado’s where they can buy things off the shelf. If we’re using swimming metaphors, it’s possible to buy a boat.
‘Well, maybe,’ answers Paul, who sounds deeply unconvinced. ‘You can make the case that a Hayes modem is still a means of communication. But it isn’t going to compete very well against broadband.’
‘The great thing about cloud-based services…[they] will democratise access to AI. And they are, now. And it’s great, because you can pick and mix. You can take a sentiment analysis service from this cloud vendor, and a speech to text recognition service from this vendor, and an image processing from this one…and you can stitch them together and build a smart application.’
‘It’s going to move the bar for smartness by making it more accessible and ubiquitous. I think lots of services are suddenly going to become much smarter, much more quickly. That’s a really important part of the picture. ’
But as the barriers to entry for AI go down, the level of standard AI integration will skyrocket. ‘Where we have to “roll our own”, we do…and it takes time to acquire these competencies. It’s not easy. For some applications, you can [buy them off the shelf], but if everyone else is using the same services, and it’s an important part of your service, you’ve lost your competitive advantage.’
‘Or maybe you find very innovative ways to combine them that other people hadn’t thought of. As always, it’s going to come back to your creativity and your capability for innovation.’
‘These are just new tools in the toolbox. And everyone is going to have access to those tools. I don’t think it’s at all certain that they’re going to inoculate you against this future. People need to be getting their feet wet, now. They needed to be getting their feet wet a while back. Which is why I say, slightly provocatively, the time for talking is over. Because I believe it is.’
Talking Education
In Ocado’s Andover factory, a robot with a soft rubber grip for handling fruit stands and learns next to the workers it could one day replace. AI watches humans, it learns, it simulates. Alarmists, based on Osborne and Frey or some other piece of research, look at Ocado and see an impressive blueprint for the future. AI both heightens the basic level of engineering compute available for Ocado (‘You couldn’t [maintain healthy bots] with human beings looking at a swarm of 3,500 robots in one warehouse let alone warehouses all around the world’ says Paul); and saves them labour costs.
Our own research suggests there is a greater emphasis on automating processes in 2018 than there has been in previous years. Whether we will reach the longed-for productivity boom; the renaissance of “more fulfilling” creative and technical jobs; or an unskilled unemployment crisis, remains to be seen. But it is happening soon.
‘You know what? I don’t know who is right,’ said Paul. ‘I’m firmly in the optimistic camp. I believe AI can encapsulate the best of us but clearly, it’s a technology that can and will be weaponised in different ways.’
‘But for me, it’s all about education. And when I say education, I do not just mean skills; refreshing; retraining. I mean education in its most holistic and fundamental sense. I believe we have to fundamentally reboot and recreate our education system from the ground up, to cope with what’s coming towards us.’
‘I say this not just as an employer but as a father: for what lies ahead, we have to do something significant, in terms of big thinking, disruptive thinking, when it comes to education. I think we also have to think long term. We have to go all the way back to the primary school where this digital skill pipe starts. And that pipe carries on through to tertiary education and into business, too, in terms of lifelong learning.’
‘We have to teach children the sorts of metaskills that people talk about, such as creativity, collaboration and entrepreneurship, which will not be devalued by these new technologies in the same way that many of the other things we teach them now will. But we also have to teach them the art of self-reinvention; we have to make the midlife crisis into an art form.’
‘People are going to have to fuzz this boundary between education and work. People are going need to go back and have multiple bites on the educational cherry. We’re going to have to get universities and colleges playing more collaboratively with organisations like ours. And we’re already having those conversations with educational institutions. It’s not going to be a, “go to school, go to university, go to work, do not repeat.” It’s going to be something much more circular.’
‘Long-term thinking is required here, which is a real challenge in [a country with] a five-year democratic term. Which is why personally, I would argue we need to do something a bit like what Finland have done and take the educational football off the political playing field.’
Which seems unlikely. It does seem likely that, if you believe in the imagination of the middle-aged population, the “midlife crisis” could indeed become an art form. Life begins at - what’s the number again? But it will require vocal campaigning and sustained effort to see the kinds of changes our education system might need.