Ocado CTO Paul Clarke on the Explosion of Applications for AI
As technological planets align, artificial intelligence flourishes. . .
The first planet relates to the significant recent advances in AI component technologies such as natural language processing, image recognition, convolutional neural nets, deep reinforcement learning and others.
Then there has been a significant reduction in the availability, cost, power usage and physical footprint of the silicon chips required to accelerate embedded applications of AI, combined with an increase in their performance.
Building AI into your applications used to require hiring specialist engineers with PhDs and some applications of AI still do. However for other applications, the advent of cloud-based AI services has completely disrupted accessibility to AI, made it a commodity, and in so doing has redefined the baseline for smartness. Now for a few cents you can call a cloud API, pass in some data and get back a smart prediction or insight. As well as lowering the cost of smartness, cloud is also enabling companies to store and process at lower cost all the data required to train their AI solutions.
Where the Internet of Things (IoT), Big Data, Cloud Computing, Robotics and AI collide, we are seeing the birth of smart systems that can move, are able to communicate with one another and interact with the world around them via IoT. This is the cocktail of technologies that power Ocado’s end-to-end e-commerce, fulfilment and logistics platform for online retail.
However when it comes to disruptive technologies, I firmly believe that AI really is the “one to rule them all”.
Firstly, to do really exciting things with other disruptive technologies you often need to apply AI. For example, an autonomous robot without the AI to process its vision and sensor data, is just a dumb pile of metal; AI turbocharges other disruptive technologies.
Next, there is the fact that AI is a recursive technology because you can use one generation of AI to help train the next. And finally, AI can generate knowledge or at least unearth knowledge that was previously hidden to human eyes.
For me, these characteristics put AI in a disruptive league of its own.
And in the current armageddon vs salvation debate around AI and robotics, I am definitely in the optimistic camp. I believe we need to balance the constant stream of negative stories about these technologies with some positive ones.
For example, I am really excited about:
- How we might use these technologies to augment us human beings, to achieve things we currently cannot do, don’t want to do or where we add little value
- How we might use them to make smarter use of our scarce resources whether that be time, energy, water, land, road capacity etc
- How they can help improve our physical and cyber security, including how smart agents can help us safely navigate a world awash with sensors and data on the back of IoT
- How we might use them to offset or hopefully reverse the impact of major problems we are facing such as climate change, pollution, famine and poverty
- How we might use them to generate insights, manage complexities and make discoveries that are beyond our human minds. For example, analysing medical scans or data from wearables to identify the signatures of medical conditions in their early stages
- How we might use them to tackle major societal challenges such as remote healthcare and providing care (and even companionship) to our growing elderly demographic
At the same time, the transformational impact of these technologies means we can’t ignore the major ethical, philosophical and societal challenges that lie ahead. That’s why as an employer, a citizen and a father, I am particularly concerned about what we are doing (or rather are not doing) to prepare the next generation for the smart automated world they will inhabit.
To prepare for this future, I believe our education system needs a fundamental reboot!
To start with, we need to be weaving digital literacy throughout the curriculum. And true digital literacy is much more than just teaching our kids to code – not that we are even accomplishing that task terribly well.
For starters, it’s also about data literacy – learning how to manipulate, visualise and gain insights from data, modelling with data, the dangers of bias and so on. It’s also about understanding the amazing possibilities, current limitations and potential dangers around technologies such as AI and robotics, how to tame and harness these technologies, the important ethical and philosophical questions they present, what it will mean to be human in a world awash with smart machines and so on.
Along this journey, we also need to change the perception of technology by girls at a young age in order to get more women excited about technology, and help remedy the significant shortage of women with STEM qualifications.
Coding and data literacy should be mandated just like English and Maths are. They are essential transformative skills not just for those who may go on to become computer scientists but for everyone. They are also stepping stones towards literacy in AI. A side effect of mandating these subjects would be to help reduce the gender gap in STEM and the subsequent earnings disparity within the labour market. Then we need to be exploring using AI to help teach subjects such as computer science where we lack sufficient skilled teachers.
Unfortunately many of the skills and techniques we are currently teaching our children will be as devalued in the years to come, as the encyclopedia has been by the world wide web. That’s why I believe we need to focus on teaching those meta skills that will endure such as collaboration, creative thinking, intersectional thinking, entrepreneurship and so on.
As a company we decided to tackle this problem at source and, as a result, we developed the Code for Life initiative (https://www.codeforlife.education/), developed entirely in-house by our Ocado Technology team and rolled out as part of our corporate responsibility programme.
The idea was for our engineers to volunteer their time to build online resources to help teach children digital literacy skills. The first example of this was an application called Rapid Router – a game to introduce primary school (KS1/ KS2) children to coding, which was launched as part of the new primary school curriculum in September 2014. Rapid Router is now being used by over 100,000 users and nearly 2,000 schools across more than 50 countries. The next development is a multi-player game aimed at GCSE level students – a sort of virtual robot wars – which will launch later this year. After that, our focus will shift towards data literacy and ultimately AI literacy.
Data are the food of AI and we need to make it easier for businesses and organisations across all sectors to share their data. We need data marts (along the lines of that created as part of the MK:Smart initiative) to facilitate the exchange of data on both a free of charge and chargeable basis. We need standards not only to facilitate the exchange and aggregation of data but also to enable different datasets to be mashed together to create new datasets. We need data passports to hold the metadata associated with datasets and to control what purposes these datasets can/ cannot be used for and by whom.
The UK has given birth to a number of important AI startups and innovations. Indeed we have all the raw ingredients to make us the world’s leading AI powerhouse; all we have to do is not miss this incredible opportunity!
Paul Clarke is Chief Technology Officer at Ocado, the world’s largest online-only grocery retailer. He spoke recently at the D/SRUPTION event, Disruption Summit Europe, take a look at what he had to say on the art of scalable and sustainable disruption: