RPA and AI are key technologies of our time, but will they endure?
Let’s do a quick thought experiment. Imagine you are a User Interface Developer for a large software vendor such as SAP or Oracle. You spend all of your time creating screens and journeys so that users of your software have the most effective and efficient experience. You think hard about screen design, typefaces, workflows and colours because these are all important to improving how a user interacts with the software. But then one day you suddenly realise that there are more robots using your software than there are humans. Why bother then creating all these fancy front-ends, when the robot doesn’t care about all of that? Why not just have everything in a simple table, in black and white, or why even display anything at all, since robots can’t ‘see’ anything. Let’s just connect the systems and data together with bits of code. And then suddenly we are in the domain of APIs and Web Services. And the robots become redundant. Could this be final scenes for RPA?
Now, I may be exaggerating a bit to make the point, but this does demonstrate how Robotic Process Automation (RPA) potentially could become just a passing phase of automation, as systems transition from being worked by humans to working with each other. It is difficult to predict when this might come about (and things always happen much slower than everyone expects) but in the meantime I predict we will see RPA increasingly becoming an everyday standardised tool that businesses use to increase efficiency and reduce costs. Perhaps it will even ship with Microsoft Office?
The same evolutionary commoditisation happened with outsourcing, which was my previous consulting specialism. Back in the day we used to have ‘outsourcing advisors ‘and do ‘outsourcing projects’, but now outsourcing is usually part of the wider IT or transformation strategy of businesses, and it’s done by people within the business. This is the journey that RPA may well take, and we’re already starting to see that happen with the growth of in-house RPA teams.
The next decade
If there is an important role for RPA in the next 5 to 10 years, it will not be as the primary automation tool, but actually as the support act to artificial intelligence (AI). The value from AI can be orders of magnitude greater than RPA will ever achieve because, rather than simply delivering labour arbitrage, AI augments peoples’ capabilities so that they can make better business decisions. And doing something that, for example, increases revenue by 10% is going to have a far more impactful benefit than doing something that reduces labour costs by 10%. AI should therefore be the lead technology, but it will need to be fed data in order to be valuable. And we all know that RPA is great at connecting data sources, collating data and feeding it to the right (AI) systems. If AI is the schools, then RPA is the school buses.
What about the future of AI?
Will it also shine brightly then fade into the technological twilight? Artificial Intelligences’s evolutionary path is actually very different. Although it’s been around in various forms since the 1950s, we are still very early in the journey, but with the technology developing at an exponential rate. What we have now is the perfect storm of ubiquitous data (which AI feeds off), storage costs for all this data that is so cheap that they almost become irrelevant, the processing power to run complex models in minutes rather than days, and everything connected together(including access to publicly available data training sets). AI is ready to really lift off.
But before we get carried away and start to imagine sentient machines that will take over the world, we need to remember that everything that AI does is very narrow. That means that each AI model can do one thing, and one thing only, very well. An AI trained to recognise pictures of dogs can’t read text. It can’t even be used to recognise pictures of cats – the system would need to be completely wiped and retrained using cat pictures instead of dog pictures. In some cases, AI can exceed human capabilities in a single capability, such as predicting when certain pieces of equipment will fail, but what it can’t do is join all these narrow capabilities together to create real understanding or cognition, as our own brain can do so well. We are very far off this state of affairs (called Artificial General Intelligence) and some would argue that it may never happen. But there are early signs that we are moving in that direction…
How this may come about…
I want to relate the story of AlphaGo, an AI that was designed by DeepMind to play the ancient board game of Go. Many of you will have read a few years ago about the amazing achievement of AlphaGo beating the best human player in the world at Go(to put this into perspective, there are more possible moves on a Go board than there are atoms in the Universe). But that wasn’t the really interesting part of the story. To train AlphaGo, DeepMind first showed it thousands of games where humans had played humans, and it learned its tactics from there. Afterwards they got it to play a version of itself, which it could do much, much quicker, and that’s how it became better than the best player in the world. But then they developed a second version, called AlphaGo Zero. AlphaGo Zero was essentially the same machine, but it only learnt from playing a version of itself – it didn’t use human games as training data this time. So at first AlphaGo Zero was much worse than humans and much, much worse than AlphaGo. But very quickly it learnt how to play the game well, and over time (just a few days) it became better than humans and, just like its elder sibling, beat the reigning world champion.
Now, the really interesting part is that AlphaGo Zero started playing moves that human players had never seen before. Because it was unencumbered by human ways of doing things, it came up with its own approaches. In effect it had developed new moves – it had become creative. Now, this can be a very scary proposition, but it can also be very exciting. The human players were able to learn from the way AlphaGo played and incorporate those tactics into their own games. If we can use AI like this in other disciplines – science, business, politics, art, sport – then it will surely have a huge positive impact on a great deal that we do.
So while everyone is very excited about the RPA project that they are doing right now, keep in mind the AI long game. Long live AI!
For more from Andrew Burgess on AI and RPA, sign up to the free weekly D/SRUPTION newsletter.