Above, Stanley Kubrick’s 2001: A Space Odyssey pitted astronauts against an out-of-control AI dubbed HAL.
Generative AI programs such as ChatGPT, a large language model, and DALL-E, an art creator, have captured the public’s attention in the last year and sparked a white-hot discussion about artificial intelligence. Some of the discussion has been productive. Much of it has been breathless.
Fed by decades of Hollywood movies on the menace of artificial intelligence, that mounting anxiety has grown into a near hysteria. In a Reuters/Ipsos survey in May 2023, 65% of Americans said they were concerned about the use of AI. In the same survey 52% of Americans agreed with the statement that “AI is bad for humanity.” And 83% of Americans in an AI Policy Institute survey in August 2023 believe AI could accidentally cause a catastrophic event.
Granted, some of those fears about the consequences of an unleashed AI come from the tech community itself and others skeptical of the programming logic behind it. But that critical discussion around the future of AI has added energy to a wave of oversimplified and generalized fears in public discussions. The public concerns about AI include limited trust in technology companies to self-regulate, and a fear of widespread job loss—83% don’t trust groups and companies developing AI systems to do so responsibly, according to a June 2023 survey by Ipsos. A separate survey by Reuters/Ipsos finds that 62% of Americans think that “companies that replace workers with artificial intelligence should pay a financial penalty to offset the increased unemployment.”
Public opinion research has focused on these general perceptions of AI and potential risks, with a heavy concentration of questions related to potential job loss and autonomous vehicles. Very little opinion research has gauged perceptions of AI’s potential benefits. Instead, most of the opinion research that has been done perpetuates a kind of doom loop of negative feedback.
While leaders do need to understand that the general public has significant concerns about AI, it also should consider them within the historical context of technology adoption. There is a common pattern of social adoption with all new technologies, from blissful ignorance to early adoption to irrational exuberance, then disillusionment, regulation and finally acceptance and technological maturity. Elevators and electricity are two examples.
When elevators were first invented, the public was resistant to entering them. Elisha Otis’ demonstration of his safety elevator with a braking system at the 1853 New York World’s Fair began to allay the public’s concerns.
Similarly, there was significant public fear of electricity at its advent. Benjamin Harrison, US President from 1889 to 1893 and the first to live in an electrified White House, echoed common fears by refusing to touch the light switches for fear of being electrocuted. Meanwhile, the new technology’s benefits were also misunderstood, with electricity being offered as a cure for various ailments, via consumer devices that gave users a mild shock.
As the public experiences the positive aspects of AI, fear will slowly give way to more nuanced opinion. AI will evolve from a tool, to a co-worker, to an advisor, to a friend.
That bumpy road toward the acceptance of an important new technology plays out consistently. There are a number of models for social adoption. The most commonly used is the diffusion theory: A technology is adopted by progressively larger segments of the population, starting with innovators and early adopters, expanding into the early majority and maturing with the late majority and laggards. By this model, GenAI is still very much in the innovator and early adopter phase of this model.
Most business leaders are also familiar with the Gartner Hype Cycle model, which posits that every technology progresses through five phases: an innovation trigger, a peak of inflated expectations, a trough of disillusionment, the slope of enlightenment and a plateau of productivity. Gartner places GenAI at the absolute apex of inflated expectations, just before the crash into disillusionment.
To pinpoint the exact location of the sentiment of the general public on any of these models is not easy. Unlike other technologies that have surprisingly burst onto the public’s imagination, humans have been speculating on artificial intelligence ever since Charles Babbage’s Difference Engine in 1832, the ancestor of the modern computer. In 1833, Lady Byron reviewed the operations of Babbage’s Difference Engine and referred to it as a “thinking machine.” We have been wrestling with the implications of thinking machines ever since.
Years of sensational movie plots, from HAL in Stanley Kubrick’s 2001: A Space Odyssey to this season’s The Creator, have primed the public for a deep fear of AI, in particular the inevitability of Artificial General Intelligence (AGI) or so-called Strong AI. In fact, 54% of Americans believe human-level AGI will be developed in the next five years, according to an AI Policy Institute survey.
Western culture struggles with AI in part because of the Frankenstein Paradigm. In 1818 Mary Shelley published Frankenstein or The Modern Prometheus, which many consider the first science fiction novel. Dr. Frankenstein creates “the monster” and it unleashes death and destruction. The story cemented a vision of technology that destroys us. But, what if we created a new technology, like AI, and it freed us?
If history is any guide, public opinion will get far worse before it gets better. The trough of disillusionment will be very deep. History suggests that some future event will trigger regulation and new social norms, and these will promote social acceptance.
History also suggests a winner-take-all arms race. Revolution in Military Affairs theory, or RMA, postulates that each transformative technology is weaponized and the power that dominates the weaponization of that technology gains hegemony. Chariots, bronze, iron, gunpowder, industrialization, air power, atomic weapons and cyber are all examples of weaponized technologies that produced a hegemon. We can anticipate the same pattern with AI. However, AI may be fundamentally different than previous technologies, in that it may develop at a far more exponential rate, giving the first mover an unassailable advantage. An AI arms race is unlikely to reassure a wary public.
As the public experiences the positive aspects of AI, fear will slowly give way to more nuanced opinion. AI will evolve from a tool to a co-worker and an advisor. AI will reduce the risk of accidents, save lives with early disease detection, invent new medicines and free us from boring work. Pew Research Center polling from 2021 found that the public is more excited than concerned about using AI to “perform repetitive workplace tasks” and “diagnose medical problems.” This gives us several clues to our future.
Similar to computing’s march from mainframe to personal computer, AI will also evolve from distant other to helpful friend—personal AI. Only then will we move beyond the Frankenstein Paradigm.