Korngiebel: I wanted to talk a little bit about how you define AI and the benefits that you’re getting out of it today from a practical perspective.
Henderson: I think of it from the perspective of how you apply algorithms to try to solve complex problems, solving things that are commercially interesting but too difficult for our employees to figure out alone. We’ve tried a lot, and failed a bit, and I think that’s the important thing with AI.
For example, we capture every single word that’s sung during a song or in an ad break and all those words are transcoded, categorised in an engine and put through sentiment analysis. That’s a lot of data which on its own isn’t that interesting. We can overlay that with demographic, weather and major event data and combine this with behavioural activity data of millions of people listening on our apps such as which songs they like or when they stop listening. We can then make inferences about why people are listening —not just to songs but to advertisements as well – and change the mix of content and then see the impact of those changes to further improve the model.
Korngiebel: You can’t have a conversation around AI, machine learning, and automation without addressing the idea that it will threaten the employment of many people. What are your views on that?
Smith: Do I believe it will fundamentally change the workforce? Yes, but not in terms of numbers, as I think people will simply end up working differently. There are two very stark views that I see out in the marketplace. There’s the one that says, in the last industrial revolution, more jobs were created than were destroyed. The counterargument to that is to say that situation occurred only because, in the past somebody had a job making something, and now they switched jobs to think about something. But, AI stops us thinking about something. So, what’s next for the workforce? I think the answer to that is probably areas where judgment, interaction, and human instinct are crucial.
Korngiebel: In technology, it’s is easy to talk about the buzz and hype, but when you see the real-world examples you really feel it. What examples are you seeing in the usage of AI and machine learning in your business?
Henderson: People look at the big examples, but it’s the smaller day-to-day stuff that can add real value. You decide that you want to go on holiday, so you book your holiday and then, all the robot does in the background is it automatically takes the information, populates your calendar, sends it to your manager, and marks you out of office. It’s very small things like this that cumulatively starts to create lots of value for the organization.
Smith: For HR, we’re using AI to help remove unconscious bias in job descriptions and recruitment. So, we’ll run the descriptions through an AI and it’ll tell us if we’re overly masculine or feminine in how we’re writing our job descriptions. We will also leverage gamification to help determine things like personality, cultural value, and fit with the organisation as well. An area where I see great benefit is running a video interview through a translation engine to get a view on how good somebody’s English and communication skills are. And if you’re hiring in volume in say, somewhere like India or China, you can save a lot of time with these upfront tasks which generally take a lot of resources.
Korngiebel: How do you ensure you can trust your data?
Smith: It’s about choosing the right data and the right datasets so you get the right outcomes, not false positives. If you just put machine learning on top of your legacy hiring data, and then decided what we’ll do is look for the people that we hired that were the highest performers, the chances are you’d not get results that align with your wishes to be a diverse employer. The machine learning algorithm would just repeat history and maybe lead you to hire as you did 5-10 years ago! You need to be smarter about how and where you use machine learning, and what data you expose it to.
Korngiebel: Finally, can you talk about other technologies that you’re excited about and that are influencing your business in new and interesting ways?
Henderson: I think from my perspective, Snowflake, the data warehouse, is really interesting because it allows us to join disparate data sources and makes it really easy to do things with data. So, I’m loving that in the near term. Also, I guess voice, with Alexa and similar devices. It’s just the start. You’ve got Amazon with at least 5,000 linguists making Alexa better.
Smith: I think data is undoubtedly the main area, particularly predictive tools. I’m not an IBM employee, but I saw some stuff that blows my mind away around some of the Watson technology. Another area that I think is cool is robotic process automation (RPA). And, the next stage of RPA is cognitive robots. So, this is more than replicating end users. These cognitive robots have the ability to scan and learn from documents and from unstructured data and “dark” data. And that will allow us to automate so much more and uncover so many additional capabilities. I think particularly around unstructured data and processing for the back office, this is going to be huge.