Devorah Fischler
Michael Kearns, National Center Professor of Management & Technology in the Department of Computer and Information Science, draws on decades of leadership in the field of artificial intelligence in an interview with Malika Bilal, host of Al Jazeera’s flagship podcast, The Take.
In the episode, “The Human Cost of ChatGPT,” Kearns discusses what makes the newest generation of ChatGPT different from previous iterations of the same technology, noting the explosion of ethical issues prompted by the accelerating pace of AI.
ChatGPT is, at a high level, “a very powerful statistical model” that uses “a massive collection of actual textual language data” to complete a human-generated prompt with the most likely continuations.
“It never worked this well before,” says Kearns, “because we never had so much data before.”
“I do feel something pretty fundamental has happened here,” continues Kearns. “Whether it will really be that people won’t need to learn to write well anymore, I don’t know. But I do think from a science standpoint that they have really cracked some difficult problems.”
When it comes to ethics, however, problems are proliferating faster than they can be explored and resolved.
“The number of ways things could go wrong is…[more than] innumerable. The ethical issues proposed by ChatGPT and similar large-language models [challenge] what people like me have thought about the topic before.”
With the business of AI outpacing research into its effects, Kearns expresses concern for the future.
“My point here is that this is just going to be a big game of whack-a-mole. As scientists, this is not the way we want to proceed.”
To hear the interview in its entirety, listen to The Take.