When I first became interested in artificial intelligence, AI, the technology was more fantasy than reality for most people. Driverless cars and intelligent household robots were imagined, but few believed that they would become a reality in the near future, says Thomas Bolander, Professor at DTU Compute. He researches artificial intelligence and has provided feedback on the EU Artificial Intelligence Act. In this article, he provides an overview of how far the development of artificial intelligence has come.
Where are we in the development of artificial intelligence?
We have reached a development level where we can build artificial intelligence models that are incredibly large and can do a lot of things, but we have by no means reached the limit, perhaps quite the opposite. Personally, I think this is only the beginning. Artificial intelligence has gone from being a fun science-fiction element that many deemed unrealistic to something most people now have to deal with. Yet the potential of artificial intelligence is so great that we cannot even fathom the extent of it. It's a bit like a Swiss Army knife that just keeps getting new types of tools that we have to learn to use and understand.
What has created the massive attention?
ChatGPT has taken the world by storm, of course. Similar artificial intelligence systems have long been known, such as Apple's Siri, which has functioned as a personal assistant since 2011, but which was perhaps less impressive initially. We have now reached a place where much bigger and better models have been developed. You could definitely say that artificial intelligence has become a trend - one which even HM Queen Margrethe mentioned in her last New Year's speech. That's primarily because we suddenly have a tool like ChatGPT, which is too good to ignore.
What significance does it have that development is mainly driven by private companies?
There are clear advantages and disadvantages. If private companies like OpenAI (which is behind ChatGPT, ed.) and Google did not put so many resources into developing these artificial intelligence models, we probably would not be as advanced as we are today, and then we might not have discovered how great the potential of artificial intelligence actually is. It is clearly a resource issue - it's about money and the opportunity to test your technologies on a large scale. Conversely, of course, there are also some concerns about the kind of companies that develop these technologies. Do they want to develop for the common good, or will money dictate whether they have to compromise on morality? Therefore, there may be a concern that development is moving too fast for society to keep up and adapt. It will be a question of regulation and ensuring that we develop technologies that actually benefit us.
How do we regulate this massive development?
Policy and regulations sometimes lag a little behind when development moves this fast. Sometimes technology manages to do some damage before you become aware of the problem. We need to take into account that artificial intelligence is a very complicated technology with almost infinite layers to it. We don't necessarily know what we're dealing with when we talk about artificial intelligence, and we don't really know its limitations either. That means we now have a technology in constant development, with no clear limits to how we can use it. And consequently, that also means we need to place quite high demands on ourselves when it comes to regulating artificial intelligence. If, for instance, an algorithm makes a model act racist or sexist, that means the model has shortcomings, and then we must pull it back and reassess it. Because everything is still so new, we experience lots of teething problems. It's imperative that we push forward and develop new technologies to get rid of these problems. It can be compared to the invention of the car. If we imagine that a car originally had no side and rearview mirrors, which would have led to a lot of accidents when reversing or changing lanes, then there are two ways to address this issue: We can regulate and make it illegal to change lanes and to reverse, or we can invent side and rearview mirrors. We will probably have to do both when it comes to artificial intelligence.
Where will we be in five or ten years?
That is a slightly unfair question. Ten years ago, I don't think many people would have foreseen a technology like ChatGPT, and those who did might have foreseen a lot of other things that never came into being. However, I believe that the idea of a computer as your personal assistant that can interact on human terms is an area that will continue to develop significantly. Now that artificial intelligence is squarely in the spotlight, development is also faster than ever. This can lead to a faster advancement of technologies that would otherwise not have been developed until much later. For instance, it might only be a matter of time before a household robot hits the market. It has always been the dream to have a robot take care of our daily chores, and this is where artificial intelligence differs from other technologies. While a dishwasher will only be able to do your dishes, a household robot would be able to vacuum and cook as well. But whether they will emerge in five, ten, or twenty years is hard to say. Sometimes the complexity of some technologies surprises you. While they may work well in limited areas - such as warehouse robots at Amazon - it is much more complicated to operate on a larger scale among humans. So we shouldn't underestimate the complexity of having to implement that technology in the real world, where it must be able to understand people and conform to human norms. But we will definitely get used to artificial intelligence being part of our everyday lives, just as we have done with cars and computers.
Why should we develop artificial intelligence at all?
Just like with any other technology, we do it to make life easier and better. We want better health, better food, more free time, and fewer boring chores. All these areas are something that artificial intelligence can potentially help with, so naturally, that's the way development is headed. It obviously varies from person to person whether you want to chop your own vegetables or have a robot do it for you. Therefore, artificial intelligence - like many other technologies - also places a great responsibility on humans. If artificial intelligence can solve many of our cognitive tasks, it will be our own responsibility to stimulate our brains, in the same way that it's our own responsibility to be physically active if we spend a lot of time in front of the computer or driving in our cars. In addition, artificial intelligence has great potential to solve major societal challenges, including in the health sector, where it may become an important tool for diagnosing and treating patients.
