In the time between the conceptualization of this issue last summer and its production this year, artificial intelligence (AI) has exploded onto the consumer market with a dizzying force that is as exciting as it is fear-inducing, making this issue timelier and more relevant than our editorial team could have imagined. As a self-pronounced Luddite, I embrace technology slowly and reluctantly; yet whether one embraces it or not, there is no denying its increasing presence. ChatGPT, for example, was adopted faster than any other technology in history, boasting over 100 million registered users within two months of its launch. Like it or not, these technologies have arrived, and we will all be better served by understanding how they work as well as their benefits and drawbacks.
Generative AI — which is a set of algorithms capable of generating seemingly new, realistic content based on existing content or training data — had a sizeable presence at this year’s South by Southwest conference (see page 20 for more on this). The founding executive director of Wired magazine, Kevin Kelly, who now bears the title of “senior maverick” at the same publication, shared some intriguing and reassuring insights into this emerging technology in his keynote presentation, “The Universal Intern and Partner: How Generative AI is Changing How We Work.”
To begin, Kelly notes, we really don’t know how AI will be used in the future — which is true of any new technology — so discussions around it are bound to be limited and speculative. When the internet became widely available to the public in the early 1990s, for example, many envisioned that it would become “better TV,” giving users an increasing array of content to consume. But, unexpectedly, everyone started making their own content — particularly once smart phones arrived on the scene. “That was the revolution that even we didn’t see,” says Kelly.
Another key point to consider is that there is not one singular AI but rather an ecosystem of AIs capable of many types of cognition with varying abilities. But so far engineers have been able to synthesize only one type: pattern recognition and generation. They have not been able to simulate other types of psychological processes such as deductive or symbolic reasoning — at least not yet. Kelly explains that “artificial smartness” rather than “artificial intelligence” may more accurately describe the rote, mechanical tasks that AI is adept at handling. A calculator, for example, is far more capable at rapid calculations than most humans, but we would hardly call it intelligent. It has a very specific type of smartness that is useful, but it can’t match human intelligence in most other ways. Just as many animal species exceed our intelligence in some dimensions, the same will be true for machines.
But one significant issue that we face as humans is that we have an incredibly poor understanding of our own intelligence, both individually and collectively. Many activities thought to be quintessentially human are quite mechanical — activities like playing chess, searching for information, recognizing a face, or even painting a picture. But fear not, AI is not coming to take our jobs, says Kelly. In fact, the strongest results in complex processes like playing chess or medical diagnostics are achieved through a combination of AI and human intelligence.
While the lore of AI has loomed large in our collective imaginations, Kelly believes that we should consider AI not as a slave nor a pet nor a god, but as a universal intern. It can make tasks more efficient, but since it is trained on human-generated content — both the best and the worst — you’ve got to check its work. The output is very literally the average, wisdom-of-the-crowd kind of knowledge that might be considered creative but not truly innovative. And if the data used to train an AI system is biased, racist, sexist, or incomplete, it can lead to inaccurate or discriminatory solutions. Thus, it is imperative that AI be better than us — but what is “better” and who is “us”? And therein lies what may be the greatest potential of AI — helping us to discover our own humanity. So who do we really want to be? Once we have grappled with these questions, only then will we know what AI — and we — are truly capable of.