A tech observer says we’re thinking about AI in the wrong way
Sociologist and technology savant Zeynep Tufekci used a plenary session at RSNA25 to urge taking a radical fresh look at artificial intelligence.
One observer who has done more than her share of reflecting on the current trends is Zeynep Tufekci, a Turkish-American sociologist and the Henry G. Bryant Professor of Sociology and Public Affairs at Princeton University, and a columnist for The New York Times. Professor Tufekci, who writes and speaks on technological change and its sociological implications, shared her views during a plenary session at RSNA25, the recent annual meeting of the Radiological Society of North America.
Tufekci argued that a core problem for society is that people and organizations are looking at the AI revolution through the lens of current lived experiences when, in fact, changes brought about by AI are going to create unprecedented new realities that will change how society functions.
Among the examples she plumbed for the audience were the creation of the printing press and its emergence in relation to the Protestant Reformation in sixteenth-century Europe, and the development of trains and automobiles, superseding the use of horses in the nineteenth and early twentieth centuries.
Printing presses and autos
Regarding the printing press, Tufekci showed her audience a slide of a Bible from the time of the Reformation, and said, “When the printing press came about, the Catholic Church was super-excited, because church leaders were selling indulgences, and the Church was getting very wealthy, but they were writing the indulgences by hand.” Church leaders imagined that the invention of the printing press would simply lead to the faster creation of certificates of indulgence.
“It's very easy to try to imagine a new technology as if it’s just a newer, faster version of something before it,” she continued. “It takes time for us to understand its ‘native language.’ In fact, what happened is that the Reformation came about by people using the printing press to publish lots of pamphlets railing against the selling of indulgences, and also against the Catholic Church; and they printed the Bible in their own languages. There was a period that was extremely violent and unstable, so a transition can destabilize the existing order.”
In fact, the invention of the printing press dramatically accelerated the spread of the Protestant Reformation. But the Catholic Church leaders’ thinking about indulgences is precisely the kind of thinking that is happening now around AI in healthcare and society, she said.
Similarly, when the steam engine was invented in the late 1700s, it initially was compared with horses, which were widely used and were influential in human history. “And at first, the comparison was, is this faster than a horse? In fact,” Tufekci said. “The first cars were imagined as horse-less carriages; people thought about cars as a kind of horse. So, people were focused on the horse and comparing the car with the horse.”
When cars were developed and first seen on the roads, “there was a great deal of commentary on what the impact of the motor vehicle would be,” she said, noting that the impact of the personal automobile was completely unanticipated. But the invention and wide use of the person automobile led to the expansion of cities, because it enabled people to live farther away from their workplaces.
Thus, the entire concept of place changed over time with modern transportation — and none of that was anticipated at a time when people were obsessed with whether cars could go faster than horses.
Rethinking AI’s role
Similarly, we’re not thinking about how AI will impact society as it’s applied at scale, Tufekci said. “When we think about AI (and) machine learning, we are talking about AI as a connectional system. It’s not stuff that we program and ask to do things step by step. All generative products like (large language models) study examples and learn from them and develop models used to do AI stuff.
“Think about artificial general intelligence, when AI is as smart as or smarter than humans,” she explained. “It’s got the human as the benchmark, as though the AI is like a little human on a chip that is slowly getting smarter. It’s not a form of human intelligence; it is its own thing. When we imagine super-smart humans, we’re not really imagining generative AI, which is a completely different thing.”
Concerning radiology, it’s becoming clear that AI will not replace humans, Tufekci said, but instead the technology will become a necessary adjunct to human operations in its capabilities. “The way that LLMs work — I’ve been calling them plausibility engines — they take all the data and create plausible possibilities; they don’t have true or false built into their operating systems. That’s why AI applications are most successful in areas like coding, where they can easily be corrected.”
Rather, AI will always need to be paired with actual human thinking to operate effectively. What’s important now is for leaders in all industries to reflect on what they want the emerging AI technologies to do and how to protect workers and consumers from pernicious developments.
“Believe it or not, I’m actually optimistic about the future and am genuinely excited about the potential benefits AI can bring,” Tufekci told her audience. “But enjoying those benefits requires that we start having serious conversations about what we want from this transformative technology.”
Mark Hagland is a freelance writer/editor with extensive experience covering healthcare and information technology.