Transcript
the same intelligence shift, and that's how AI helps support and is leveraged in organisations. Today we're at the McLaren Technology Centre, and I'm joined by Andrew McCutcheon, Head of Data Science. Thank you very much for having us here today, Andrew. So modern Formula One environments generate a huge volume of data. Where is it distinguished between what's meaningful and insightful data to just noise? In some ways I might turn that question around to say that actually we generate a huge amount of signal, and sometimes finding the noise is the problem that we're trying to solve. So for example, in a race we get a large number of sensors, we have hundreds of sensors on the car, and we use those to derive thousands of extra channels and metrics. And then what you're really interested in is, is something going wrong? Is there something that's happening that I need to pay attention to, or is it just racing as normal? And so all of those signals are coming in, and we have engineers who look at the screen and look at the telemetry as it goes through the race, and you're trying to assess, is there something that I need to react to? And this is, I suppose, very similar to the security world as well, where you have most of the time things are fine and you don't need to react to them, but on that 1% of time, something in that data is telling you there's a problem and we need to react to it. And so that is actually where AI is quite good. It can spot these patterns, but it is also a real challenge for AI. So AI is very good if you give it an example and you say, find me this. It can go away and look through all the data and find patterns that it recognises. If you just say, find something you've never seen before, that is a much harder problem. The AI really needs to understand everything about the data to understand if something new is different or just as expected. But it can come and detect some really quite subtle patterns, and this can be quite critical to us being able to understand that there's a problem coming and to be able to react to it before it takes us out of the race. So it might be a subtle shift between the timing of two signals that would normally be perfectly in sync, but now has started to drift out of sync and it knows from previous experience and from previous data points that this indicates that something might be going wrong. So we use AI a lot to try and anticipate when something's going to go wrong, to detect it amongst all of that data, and then we can react quickly in a way that hopefully means that we mitigate the problem before it takes us out of the race.
