Artificial intelligence: the new arms race?

Will robots soon be doing our grocery shopping? Will they buy clothes you are guaranteed to like, that fit perfectly and get you ‘likes’ from your followers? Will they make medical diagnoses and decide on the treatment? Artificial intelligence (AI) – is it science fiction or reality?

Making decisions

“Artificial intelligence is the ability of machines to make decisions themselves,” says Steve Muylle, Professor of Digital Strategy. “Some people would add: better than humans. But essentially it is about making decisions, decisions that initiate action. So it is not about data analysis or supporting decisions with reports or graphs.”

AI was one of the hot topics at the World Economic Forum in Davos, along with immigration and globalisation, and AI is here to stay in mainstream media too. Each year, the transmission, processing and storage capacity of digital technology doubles, which means that things are possible today which were unthinkable a couple of years ago.

Self-learning system

One example that appeals to the imagination is DeepMind, taken over by Google in 2014. It is a self-learning system that managed to master the computer game Atari Breakout in a mere four hours. For those who do not know the game, the aim is to destroy a wall built of various layers of bricks using a bouncing ball that you hit with a bat at the bottom of the screen.
“DeepMind was given no more information than a player sees on the screen, the score and the instruction to maximise that score,” explains Steve. “The most impressive thing was that DeepMind was not programmed to play the game, but that the algorithm learned the best strategy based on feedback in the form of the score: it dug a tunnel bounced the ball off the back wall to break the wall from behind. Precisely what an Atari Breakout champion would do.”

Broad field of application

DeepMind demonstrates what we understand what we mean by ‘artificial narrow intelligence’: AI within a clearly defined context, with clear objectives, which enable an algorithm to learn quickly. ‘Broad/general artificial intelligence’ is related to more complex, unpredictable situations and less delineated, real-life problems.

“We see all kinds of initiatives going in this direction,” says Steve. “IBM is using its super computer Watson in various areas of application, including medicine. For example, radiologists are already supported by AI today. The current image-processing algorithms are gradually improving radiologists’ ability to make the correct diagnosis.”

“In retail, you have chatbots that function as sales assistants,” he continues, “just like in a regular shop, when you explain to a salesperson what you are looking for. At an online store you speak to the chatbot, which then offers suggestions. After a while, the bot learns your needs and preferences and can act as a fully-fledged shopping assistant. With one major difference: a software bot can get to know thousands of customers through and through. A flesh-and-blood sales assistant can’t manage anything like that number.”

Incidentally, the development of chatbots demonstrates to another interesting trend in AI: “All the major players have been investing heavily in natural language processing in recent years: technology that enables you to have a real conversation with a machine or computer. Amazon’s Alexa has completely penetrated the US market and is also being offered in Germany. It has not yet reached Belgium.”

Small step

Are all these examples related to broad AI? Not really, Steve says. “At the moment, we have applications that enhance and support humans, but nothing yet to replace them.”

Nevertheless it is only a small step from a virtual shopping assistant to a virtual personal shopper that actually buys your clothes within a certain budget. And what about robotic process automation, whereby software robots take over certain tasks? That is already the case nowadays at banks with various legacy systems that cannot be integrated in the traditional way, but that do need to exchange all kinds of process information with each other – for example for lost credit cards, and credit applications etc. “That robot software learns how to handle certain processes from specialised bank staff. At present no two processes are the same. Initially, the robot software will imitate the trainers, but it will gradually start learning for itself and find new working methods. That is precisely where the major potential of AI lies.”

Ethical issues

Steve still sees a crucial difference between humans and artificial intelligence: “An artificial intelligence system has no self-awareness. As a human, we can discuss the objectives of what we are doing, but artificial intelligence systems cannot do that yet.”

And that brings us to the risks of AI. What if it fell into the wrong hands? An army of killer robots with no scruples about eliminating certain targets would no longer be in the realm of science fiction. “As I said, we are not there yet, but are we going to allow it to get that far? That is the ethical question we have to consider. And is everyone equally ethical? You can imagine that major powers will not allow themselves to be stopped by ethical considerations if there is a risk they will be taken over by another major power. Remember the nuclear arms race.”

They are taking our jobs!

The people warning us of the dangers of real self-learning systems are not to be sniffed at: they include Stephen Hawking and Elon Musk. The latter also expects robots to present a threat to employment. “That was an important issue at Davos. What will we do with the millions of people who can be replaced by robots? For now people are exploring various lines of thought, from a robot tax to a basic income for all. But how do find meaning in a life without work? We are far from figuring out the answer.” 

Tension

How does Steve see it all evolving? “I don’t have a crystal ball, but I do expect AI to develop greatly – it is unstoppable. But, besides the ethical issues, it will lead to tension between politicians and the business community: between job creation and value creation for shareholders. The demographic situation in Japan, with an increasingly population and an acute shortage of manpower, is such that they are investing heavily in robots and AI. In other parts of the world, however, there is a surplus of manpower…”

Is he optimistic? Steve hesitates. “Good question, I am naturally optimistic but in certain circumstances I am not always sure what to think. Even the experts are divided, according to a study by the PEW Research Centre: 52% of respondents expect that jobs lost to automation will eventually be compensated by new ones, and 48% thinks otherwise. The crucial question is not what is possible with AI, but what we will do with it,” he concludes.

Related news

  1. From Robotic Process Automation to digital transformation

    Date: 29/11/2017
    Category: Opinions
    From Robotic Process Automation (RPA) to artificial intelligence (AI) and cognitive computing: the digital revolution is a fact, and that is good news. Karim Hajjar from Solvay, Robert Boute from Vlerick Business School and Thierry Mortier from EY talk about the challenges and opportunities on the road to intelligent automation.
  2. Insurance disrupted

    Date: 24/10/2017
    Category: Opinions
    Traditional insurance products can be rigid and complex. They require customers to fit into categories defined by underwriters and to pay premiums for areas of cover they may not even need. But this is already changing. Digital technologies are all set to make insurance highly personalised… and as easy to access and use as an app like Tinder. Welcome to Insurtech – the new era of insurance.
All articles