NEWS Newspapers and social media are full of headlines about AI, but – what is it really, and is it just hype? “We need to be aware of how systems work, and that it can do both good and harm”, say AI researchers Dimitri Coelho Mollo and Andrea Aler Tubella.
ImagePrivat
With headlines such as "Chat GPT: Are we replacing writers and coders?", it may seem that AI systems can think, reason, and behave as we do, and thus may soon replace us.
“I think it's high time to take an interest in big language models that could lead to significant changes in society, but also bring risks. However, we must be careful not to misunderstand what these systems are and what they can do. This is especially true for systems like ChatGPT,' says Dimitri Coelho Mollo Assistant Professor at the Department of Historical, Philosophical, and Religious Studies.
His main research areas are philosophy of cognitive science and artificial intelligence, and philosophy of science and Dimitri Coelho Mollo is now coordinating Understanding (and Explaining) AI, one of the research areas at Umeå University's new centre for transdisciplinary AI, TAIGA.
While automation has been going on for centuries and new AI technology is now bringing this process into new areas, AI technology is far from being a suitable replacement for human labor in most areas, argues Coelho Mollo.
“ChatGPT, for example, has major problems in producing accurate information, so it's a poor substitute for journalists and writers. Other models, such as DALL-E, could replace artists in some tasks, but is that desirable and fair? The models are trained on huge databases, which in a sense parasitize on human work, and so there are complex copyright issues,” Dimitri Coelho Mollo says.
It is equally important to remember that the AI systems that we use don't just "appear". “Building the models requires many people, such as engineers but also those who mark up the data. If there is a shift towards replacing humans with AI, we should ask ourselves who is doing the work behind this development, as well as what their conditions are, says Andrea Aler Tubella, Senior Research Engineer at the Department of Computing Science, and one of the members of the 'Responsible Artificial Intelligence' research group, which studies the ethical and societal impacts of AI, and develops tools and methods to precisely mitigate the negative effects.
“A lot of recent research focuses precisely on the impact that these systems can have, and what the consequences might be if they were to be introduced. A new regulation is also underway in Europe that aims to set strict rules for high-risk uses," says Andrea Aler Tubella.
The history of AI has been one of enthusiasm followed by disappointment, which came to be known as 'AI springs' and 'AI winters', says Dimitri Coelho Mollo. "We are now in the third AI Spring, depending on how you count, which started with the Deep Learning revolution about 10 years ago."
He believes that technology today has made many impressive advances. "Today's AI systems are able to succeed at tasks that might not have been imagined just ten years ago and we will see many new, creative suggestions being made about how to go about building AI systems."