At TechInnovation 2019, experts and industry leaders discussed the issues of capability, trust and business value in the use of artificial intelligence.
It was the year 1956, at Dartmouth College, US, that the term artificial intelligence (AI) was first coined by computer scientist Professor John McCarthy. Spawned from the work of famed British mathematician Alan Turing, the field of AI appears to be finally hitting its stride nearly a century after its inception, aided by the wealth of data and superior computing power available today.
A report from PwC projects that AI could contribute US$15.7 trillion to the global economy by 2030. This transformative potential of AI is not lost on the world superpowers US and China—each is aggressively developing AI technology for a wide range of industrial and consumer applications.
The impact of AI was a key topic of conversation during IPI’s TechInnovation 2019. A panel of experts and industry leaders from both the UK and Singapore spoke about how AI is reshaping business and society, also shedding light on the emerging use cases of the technology.
While AI is often discussed as a standalone technology, its full potential is realised when combined with other innovations, said Paul Clarke, Chief Technology Officer of British online grocery retailer Ocado.
Clarke explained that Ocado creates a digital twin of itself—a replica business that exists virtually. “When you combine digital twin technology with AI, you can optimise the virtual twin faster than you could otherwise do in the physical world,” he said. By feeding data to the virtual twin, the parameters of the virtual business can be optimised and then implemented for the physical twin.
Optimisation was also a core priority for Gaurav Bajaj, Director of Logistics Business Development at PROWLER.io, whose company has developed a decision-making AI platform called VUKUTM. “At the core of our technology is a focus on Gaussian processes—an approach where you try something, you learn from the experiment, and then you improve your decision making as you go along,” he said.
But rather than refine the decision-making process with huge datasets, and over a large number of cycles of trial-and-error, VUKUTM has found a shortcut. “We define the metrics to optimise for first, so that we don’t need millions of data points to train the [AI] model,” he said, giving an example of how his team was able to use only 22 data points to develop and deploy an AI programme that helped an asset-pooling company reduce its rate of collection failure by one-third.
Trust in tech
Functionality and business value aside, other issues like transparency and trust in the use of AI were raised during the panel discussion. A problem often cited in AI circles is the notion of a black box—an AI system so complex that the decisions it makes are not comprehensible. “It is very important as an industry leader to make sure that AI is explainable so that there is trust when deploying the technology,” said Bajaj.
But how can we embed trustworthiness into AI? Dame Wendy Hall, Regius Professor of Computer Science at the University of Southampton, UK, explained that focusing on education and diversity is key. In particular, teaching people how to spot biases, and having teams that are diverse enough to eliminate systemic biases, is important in solving the problem of transparency and AI explainability, she said.
Indeed, Laurence Liew, Director for AI Industry Innovation at AI Singapore, noted that diversity in his team has helped it navigate thorny issues surrounding AI in its ’100 Experiments’ programme. “At AI Singapore, 40 percent [of our staff] are computer scientists, 40 percent are engineers and the last 20 percent come from other domains like biology, physics, economics, finance, banking… you name it! Of the 40 projects that we have approved so far, I'm glad to say that none of them has got into any AI ethical issues at this point in time.”
The success surrounding AI has piqued the interest of many companies, and many are looking into using the technology to improve their productivity. Implementing AI, however, may be an overwhelming task for start-ups that lack the right talent or data infrastructure, which can be costly to build.
Asked if start-ups should strive to build AI capabilities or focus on their business goals first, Bajaj suggested that they give priority to defining the metrics of business success. “AI is just another tool to achieve the goals of an organisation. At the end of the day, what technology lies underneath is irrelevant, but it has to solve something,” he said.
Dame Hall, however, had an alternative view. “I think it can go both ways,” she quipped. “There are some very bright kids out there developing some wonderful AI but have no idea what their business goals are going to be.”
She encouraged large enterprises to harness the solutions developed by AI start-ups. In doing so, a win-win partnership could arise, helping companies at both ends of the size spectrum achieve their objectives.