Collaboration Forum Series 5: Building a brighter future (Week-9)

We’ve seen it everywhere; a lot of start-ups are turning towards AI now.

There is a company called Carro in Singapore, which sells second-hand cars, for example. It’s very popular and has just raised $360 million. The whole buying experience is completely contactless. The robot will show you the car in real time, check for defects, tell you where it’s parked, and you can open it with a QR code. You can sit in it, drive it, and put it back. And there’s a chatbot throughout the journey answering your questions.

We’ve also worked with a large telecommunications company. It was getting complaints from business customers because they couldn’t resolve issues quickly enough. We suggested that AI could classify complaints into different categories. We looked at the history of how these problems were resolved and which engineers were involved, so they could be reassigned. Once on site, AI could help them based on how something was resolved in the past. The company reduced its labour cost by 30 per cent and ‘time to resolution’ by 40 per cent. Human error, which used to occur, has been virtually eliminated.

But while corporations are galloping ahead, the government is watching, and is setting up guidelines for AI. What I like about Singapore is that technology is viewed in a balanced way. There are many benefits, but we must make sure we’re controlling the downsides. That’s what a smart city supposed to do: use technology to enhance people’s lives and protect their rights.

This has actually happened. Microsoft had a chat bot called Tay, but after a while it was learning racism from social media and discontinued. So, it is the responsibility of companies to check the data that AI is being trained on, to make sure it’s not biased. There are fairness metrics and bias statistics that companies can use to help.

But there might be instances when something is unacceptable, even when it’s not in the data, so that’s about humans deciding what’s right. In Singapore, starting next year, we have a certification in AI ethics for product managers. The product manager is the person who heads up design. They can ask the right questions to make sure that AI is representing the values of a company. I hope, of course, those values align with good morals and society.

Think about the problems to be solved and the vision of the business. Invite a group of people, including AI engineers or advisers, to help identify these challenges. Companies can also look at competitors to see what they are doing. Once these things have been explored, it will be possible to work out how AI can help. The third step is whether a company has the data to achieve its aims, because 99 per cent do not have perfect data.

In the beginning, that means it’s easier to choose something doable. Try and complete it in three months, be agile, and use the cloud. At the same time, think of three other problems that can be solved, using the same data. After three months, start testing it and within four months, it will have worked, or not. If it has, then it will be easier to get the budget for the next phase.

If you don’t have the talent, you can get it from three places: consulting firms, freelancers, or some software-as-a-service platforms. You don’t have to make your own custom AI; have a look at what’s happening on Google, or AWS, for example.