Current media coverage of AI seems to assume that it has just suddenly appeared out of the blue. Instead, the prevailing form of machine intelligence is the direct result of a series of decisions that we have made over the past decades. These were shortcuts aimed at addressing various technical (and business) problems, and that are now behind many of the current concerns about the impact of this technology on society. A major shortcut was taken with the creation of the very first statistical language models, and we will describe how that step was the first move towards statistical AI, how it challenged previous assumptions, and how it reflected a new mindset that was starting to emerge among AI researchers. When business models, data availability and scientific paradigms became aligned, the current revolution started. Understanding how those technical shortcuts limit the options of regulators will be essential to safely co-exist with the present form of AI.
Nello Cristianini is a Professor of Artificial Intelligence at the University of Bath, specialising in machine learning. Before that, he worked at the University of Bristol, and the University of California, Davis. He is the author of “The Shortcut - why intelligent machines do not think like us” edited by CRC Press (2023)