The aviation industry has a long-standing and well-founded reputation for being at the forefront of technological innovation. When it comes to artificial intelligence (AI), however, it seemed to be lagging behind somewhat, at least compared to other tech-driven industries. But as a recently published report clearly shows, the pace of AI deployment is now rapidly accelerating in the aviation industry too.
As European Commissioner for Transport Adina Valean observes in her welcome message for that recent report: “In comparison with the world around us, where rapid change is being powered by the broad availability of internet and an ongoing digital revolution, the aviation industry has been slower to take up technological, and in particular, digital innovations.”
Black box problem
There are plenty of good reasons, though, for this slower take-up. As Ms. Valean herself is quick to point out, “AI poses challenges of a technological, safety, security, ethical and legal nature.” That’s a lot of challenges to deal with, to be sure. But the main challenge remains quite obviously the safety-critical nature of the aviation industry itself. Never mind then that various levels of automation are already present in that industry, alongside massive data streams or so-called Big Data - a key prerequisite for the successful use of AI. In aviation, as in any other transportation industry, safety and security has always been, and will always remain, of paramount importance.
In other words: as with the adoption of any new technology, trust is the main issue here. In the case of artificial intelligence, however, this tends to be an even bigger issue than usual, since AI technologies make recommendations or decisions based on information that is quite difficult, if not downright impossible, for people to comprehend. For many people, in fact, AI remains somewhat of a black box: you put something in, but you don’t know exactly what will come out or why.
Trustworthiness is the trust we place in a technology. In order to give confidence to a new technology, we must somehow understand it first. Moreover, we must be able to justify why certain choices are made by that new technology and why specific decisions are taken. That is the explainability of AI: a second essential component for building trust, that has already been dealt with in XAI: making algorithms great again.
Questions about the trustworthiness and explainability of AI arise from the fact that we have little insight into the algorithms behind it. Hence the need for a system that can explain the decision model behind these algorithms and show us which input and parameters have been important in making a certain choice. By displaying all this clearly and visually, the barriers to using AI could be substantially lowered, so that people with a less technical background could also start working with the technology.
Last but not least, the actual ‘intelligence’ in AI is still somewhat inherently limited today by the fact that it only looks at past decisions and makes a series of new decisions based on them. Therefore, it is extremely important that the historical data entered does not contain errors and that it is not biased.
Safety and security issues notwithstanding, the recent report on the use of AI in aviation clearly “shows that AI is being trialled extensively, and increasingly introduced into service provision”, according to EU Commissioner Valean. As always, for any new technology to succeed, it is important to find the right balance between different and sometimes conflicting interests.
It is at least as important, though, to find the right business case. At Sopra Steria, we believe the aviation industry offers many opportunities for AI deployment, especially in cases where that technology can set up a collaboration between man and machine. At this stage, we mainly see AI as a human advisor, with people retaining control and responsibility at all times. Rather than a single system that solves everything itself, we prefer to think of AI as a system that is ideally suited for supporting people.
In my next blog post, I will take a closer look at the challenges in the aviation industry that AI technologies can help to address.