Artificial intelligence is being mystified and misunderstood, says Chris Boos, founder of Arago and a member of the Digital Council of Germany’s Federal Government. An impartial look at the opportunities of AI is vital – for everyone who is interested in how and why this technology is transforming our society.
Artificial intelligence is on the verge of outsmarting humans. We will soon be superfluous. Watch too many commercials and dystopian sci-fi movies, and that is exactly the idea you’re likely to come away with. These fictional worlds are full of robots clunking their eyelids at us, programs giving human managers their instructions for the day, and all-powerful neural networks dispensing wisdom – or hatching evil plots.
These are all notions born of human imagination: fun advertising and scary entertainment, no more and no less. The problem is, such ideas are weaving a narrative that masks what AI really is now, and will become in the foreseeable future. They mystify and exaggerate AI, making us more likely to think: hmm, this one technology will resolve many of our problems – or it will resolve all our problems at once by eliminating humans altogether.
Comparing AI and human intelligence does not work. This is a false analogy. Machines do not understand anything.
Yes, artificial intelligence is changing a lot, and yes, AI is in my view the vital technology that will take us from an industrial to a knowledge-based economy. But it won’t happen without our help. AI will not be the only thing that (I hope) improves our lives. And lastly, there is nothing whatsoever to indicate that machines might develop a sense of self –
whatever that may be (we don’t actually know).
AI is a tool that helps us to optimize automation. Since humans first built machines, we have trying to automate processes. So we now have machines doing things we could do ourselves, but machines can do them better, faster, and more efficiently. Meanwhile, we have brighter and better ways to occupy our time. We might even choose to do nothing. AI takes
automation to a new level, with efficiency and productivity gains that are truly incredible.
What is Artificial Intelligence?
So what exactly is Artificial Intelligence? In my opinion there are three major functional blocks. The first is “directed action”; the second “knowledge transfer”, in other words language; and the third is “problem-solving”: having a set of options available and picking the best one to deal with a specific task.
We map these functions to algorithms:
machine learning, statistical procedures and machine reasoning. Once this is complete, we bolt on the method by which the machine will learn. That can be with people, through data or by the machine itself experimenting: there is assisted and unassisted learning, and also reinforcement learning. When we put all of this together, we have a good model to define the potential of AI and to stake out the limits within which the technology will develop.
AI takes over time-consuming routine maintenance work (Photo: iStock.com/miriam-doerr)
It is immediately clear that comparing AI and human intelligence does not work. This is a false analogy. Machines do not understand anything. We, in contrast, have a fundamental understanding of the world, we understand more or less consciously how it works. Our consciousness generates understanding, we think about chains of causality, we recognize how things are related. A machine can calculate correlations, but the existence of a correlation does not mean there is a genuine connection. For example, despite the indubitably strong correlation between babies and storks, we know it’s not storks that bring the babies. Machines can work with causality, but they cannot work out causality for themselves.
The end of admin jobs – social intelligence
Let’s come back to what AI is good for. A case in point: I am certain that office-based
administrative tasks in the broadest sense will be automated within the next few years, in the same way we are already seeing in modern factories. People working in public-sector administration or in back-office roles in business perform tasks and processes that are more or less pre-defined. Human staff are needed in these roles because there is still too much imprecision in the process descriptions and definitions. If we leave this imprecision to the machines, and teach them how the processes work, they can do the job at least as well as the current human administrative staff.
Letting AI take over administrative processes does not mean we will have less work. There is so much to do.
This thought scares people. What happens to the admin staff? I counter with a question of my own: does anyone regret the loss of jobs that involved exhausting, mind-numbing shifts on the assembly line? No, and besides, these ex-factory workers are now working in administration. Plus, if the use of AI delivers, say, 80 percent efficiency gains and the economy does not shrink, will there then not be more time for new and different things? Letting AI take over administrative processes does not mean we will have less work. There will be other work dedicated to resolving other problems. Problems that are arguably more urgent – such as climatechange, education or elderly care. There is so much to do.
Individual and flexible – AI in practice
Take business, for example, specifically logistics, which accounts for around 30 percent of total economic output. Every manufacturing company is constantly figuring out how many parts it needs to stock, and when and how parts need to arrive. Businesses make complex logistics plans to keep workflows on track, and they also factor in a certain amount of buffering of parts and goods – because creating the plan takes time, and they have no desire to make a fresh plan every time something changes.
This is the sort of scenario where machines and AI can excel. A machine can within a minute devise a plan that would take a human two days. This takes us straight to the subject of real time: just-in-time is yesterday’s news. With just-in-time logistics, businesses try to bend reality so they do not have to make a new plan. But the late arrival of a few trucks at the factory can feasibly cause the near-collapse of the entire system – or at least drive up its costs exponentially. Just-in-time is in fact a highly rigid system. With AI, companies gain complete flexibility and the ability to design processes truly around their needs.
Companies have to make an effort to automate more and more. If they do not, they will not be tomorrow.
All this is already possible. Organizations just need to begin transferring to machines the tasks they still carry out manually. The only way forward is to actually do it – and not to be slow or hesitant about it. Change always means disruption, and sometimes it can mean knocking down and rebuilding.
However, organizations should not start with a huge project. Instead, it is best to begin with areas that promise distinct advantages: the functions, activities or processes that produce more data than they consume, because it is possible to simplify these processes in incremental steps. And because this will make learning faster, more can be completed in the same time, or the machines can be taught more complicated things. With this approach, it is possible to transform the entire business within a definite timeframe.
This is not easy for established players with more traditional ways. They have to make an effort to automate more. If they do not, they will cease to exist – or belong to another company that is better at automation.
Many entrepreneurs, politicians and also trade unionists have understood this. Fortunately, there is also a growing openness in society, as well as the will to learn and take into action. And that’s exactly as it should be – because it’s only AI.
Lead image: iStock.com/3alexd
Sidebar: Steffen Kugler