Column
Damian Trilling
What a Paranormal Fair Teaches Us About AI
About 20 years ago, during my studies, I took a course on report writing and had to go to a paranormal fair. A woman was selling a box (an expensive one) that was supposed to do something with the ‘energy’ of your drinking water. My question about how this worked was not of interest for her: “It doesn't matter how it works. You don't have to know how an engine works to drive a car, do you?”
“Let's talk a little less about what artificial intelligence does and a little more about how it works.”
This statement has stuck with me – and not just because I happen to know how a car engine works. There was a lot of interest in her device, and I was fascinated by how much interest there was in the what, but how little interest there was in the how.
Now, almost two decades later, this amazement strikes me again. Everything and everyone seems to be working on ‘something with AI’. Admittedly, language models such as GPT-4 or models that generate images, sound and even videos are extremely impressive.
Both in academic education (“Use AI to make your courses more efficient!”) and research and research applications (“We will build an AI tool to do X!”), you can't avoid it.
But because of this hype, it's all too easy to forget to ask the question of how this all works. Now, I'm not saying that everyone should be able to understand every mathematical detail of such a model, but without at least a basic understanding of the principles behind it, it quickly seems like black magic. It surprises me how quickly many – including academics – trust a so-called AI tool, even if they don't know how it works. To make it concrete: you can't do critical research into, let's say, the role of big tech companies, while at the same time uncritically using a tool you don't understand.
Let's - especially as humanities scholars too - talk a little less about what artificial intelligence does and a little more about how it works. Dismissing the technology behind it as something for betas is a luxury we can't afford.
Damian Trilling (1983) is Professor of Journalism Studies. He, too, works a great deal with AI. In the NEWSFLOWS project, his team uses language models to identify news events and create news recommendation systems, and in the TWON project, a team of eight international partners is developing artificial 'users' of social networks to be able to study them.
magazine for humanities alumni june 2024