“It’s a truism in the AI world that as soon as a technology becomes reliable, it’s no longer considered to be AI. Machine translation used to be the most interesting problem in AI, the centerpiece of scientific efforts in the 1950s and 60s – it is now rarely discussed because statistically-based translation systems work very well if they’ve got sufficient data to extrapolate from.
“As AI starts to work, it becomes normalized, and ceases to be seen as ‘AI.’ As a result, it’s hard to know what we’ll consider to be AI by 2040. It’s likely that many debates about AI will have been resolved. We will likely understand what our societal comfort level is with automated vehicles, for example. This is not necessarily a guarantee that all driving will be automated, more a sense that we will have established what parts of driving are automated (highways, dense urban areas) and which require human control (rarely-traveled rural roads, challenging weather conditions, for example).
“The set of issues that are controversial will shift from year to year, as some AI applications become ordinary, others become tools used by humans and a small set remain the locus of debate. While this sounds like an affirmative embrace of AI. I don’t much like the future I’m describing.
“This next period of AI will be one of sorting; some tasks will be automated entirely, some tasks will require skilled humans to work with automated tools and other sets of tasks will remain curiously untouched. Almost by definition, the interesting topics in AI are the controversial ones: Can we trust an AI that hallucinates to write meaningful and significant texts? Should we allow technologies that are opaque and difficult to discern to predict our behavior and to act on our behalf, move objects in the physical world, spend money?
AI will continue to become ordinary in ways that we don’t question sufficiently. … My call is to ensure that as AI becomes ordinary we do the hard work of understanding what presumptions we are encoding within our systems.
“My prediction is that the set of issues that are controversial will shift from year to year, as some AI applications become ordinary, others become tools used by humans and a small set remain the locus of debate. While this sounds like an affirmative embrace of AI. I don’t much like the future I’m describing.
“AI will continue to become ordinary in ways that we don’t question sufficiently. Built into every AI or machine learning system are the assumptions, values and biases of the data a system has been trained on. The more ordinary and unspectacular an AI system is, the less likely we are to interrogate these biases and work to mediate them. My call is to ensure that as AI becomes ordinary we do the hard work of understanding what presumptions we are encoding within our systems.”
This essay was written in November 2023 in reply to the question: Considering likely changes due to the proliferation of AI in individuals’ lives and in social, economic and political systems, how will life have changed by 2040? This and more than 150 additional essay responses are included in the report “The Impact of Artificial Intelligence by 2040.”