Sooraj Shah reports from Gitex Technology Week 2018 in Dubai on the lessons that some of technology’s biggest names have learned from implementing artificial intelligence.
Internet of Business says
Representatives from Google, LinkedIn and Microsoft took part in a panel discussion at GITEX 2018, the biggest technology conference in the United Arab Emirates, this week – and it was clear from the discussion that they are all betting big on AI.
LinkedIn, the social network that was purchased by Microsoft in 2016 for $26.2 billion, may not be the first name that springs to mind when it comes to artificial intelligence. However, according to Igor Perisic, chief data officer (CDO) at the network, “AI is like oxygen for [its] product.
“Without it, we wouldn’t be where we are. It took us a long time to get there,” he explained.
The company had to take three different transformational steps to get where it wanted to be with AI – and none of them were primarily to do with the technology itself.
First, LinkedIn had to rethink the way it used data, and then completely change the way it thought about R&D.
“Normally, if something needs to be changed to red instead of blue then you can raise a ticket and it fixes it,” said Perisic. “But AI doesn’t work like that, so we had to convince and evangelise executives to change the development cycle.”
Third, the team had to rethink the LinkedIn mobile app. “The interface is completely different to how it was before, and it uses data to see how it’s performing. You want to be able to optimise and personalise it to make sure it works well for the user,” Perisic explained.
Microsoft’s Ali Dalloul, general manager of strategy and commercialisation for AI, perception, and MR cloud, told delegates that there were many different things that the technology giant had to consider when it came to AI.
For example, which industries would be impacted the most, how organisations would need to realign their business strategies around the technology, how to “think big, but start small”, and how to have a data governance model that enables businesses to make use of their data.
Microsoft also had to consider the accountability of using such a technology, and the transparency and ethics of how data is used, he said.
When it came to considering how AI could help its customers, Microsoft decided to test the power of the technology on its own products.
“At Microsoft we started with our own apps – so we infused AI into all of that, whether it was Office 365 or the Xbox. Secondly, we made it available to all of our developers as building blocks to unleash their productivity and amplify human ingenuity – by giving them all of these building blocks, to rebuild and customise AI models,” Dalloul stated.
Now, the company uses AI for a variety of purposes in house, including to forecast its own revenues, he said.
Tips for others
All of the executives on the GITEX 2018 stage offered advice for organisations that want to use or incorporate AI in their own business strategies.
Cassie Kozyrkov, chief decision scientist at Google, advised organisations to ensure that they know which part of AI they want to be involved in.
“There are actually two disciplines of AI. The first is building from scratch and the other is applied AI, or what we call decision intelligence,” she said.
She added that the two could be described as the difference between how microwaves are built, and how they are used to cook food. By that analogy, decision intelligence would cover how to make new recipes and do new things in the kitchen.
Kozyrkov warned that any businesses that don’t know which type of AI they are after could fail. “You need to know what business issue you have before you go to the ingredients, and you have to look at the ‘microwaves’ that are already available and try them out and see if they’re working for you. And then you hire a team of researchers to build you something that you need,” she said.
“You don’t start with hiring 20 PhD nerds who don’t know anything about cooking, or the business problems, and they’re surprised when it flops.”
LinkedIn’s Perisic added that companies should learn to “crawl before they run”.
“AI assistance can be very complex and then things can go wrong,” he said. “If you start with a complex algorithm and start running with it, it will fail. Secondly, make sure you are creating value for your users, and thirdly even if you haven’t started yet, it’s never too late,” he said.