Last week, OpenAI launched a mobile app version of its famous chatbot, ChatGPT, for Android users in the United States, after having debuted a version for Apple users in May.
The generative artificial intelligence can now go anywhere we wish to take it. But is that a good thing?
Whenever software becomes mobile, it becomes transformative. Social networks, for example, exploded when they hit our phones. Their user bases grew, but what truly multiplied was the importance of these platforms for news, politics, advertising, publishing and so on. Going mobile shoved social media into the heart of culture, high and low — and we should expect brute escalations to follow from the new AI apps. While it’s too early to know the exact impact, history — specifically cautions from computer pioneer Mark Weiser in the early days of the digital revolution — holds a thesis.
Weiser was a visionary and widely considered the father of ubiquitous computing — the idea that computers would be everywhere and the internet would seep into everything. And while he championed the merits of a thoroughly connected society, he also worried that having conversations with AI on our handheld devices might weaken the ties that bind us to the world.
The idea of adding AI chatbots to mobile screens is as old as Apple’s Newton, an early-’90s handheld computer that failed to catch on. In 1992, the company’s top researchers joined MIT’s best technical minds to stage the world’s first conference devoted to this grand ambition. The debate that ensued offers a warning we might heed today: Whereas the original ChatGPT exacerbates problems of misinformation, plagiarism and algorithmic bias, mobile ChatGPT stands to erode our sense of place and connection.
The Apple and MIT innovators who headlined the 1992 conference believed that AI assistants and handheld computers were a perfect match. The prototypes they envisioned would recognize speech and answer in kind. Having conversations with software promised far smoother human-computer interactions than typing and clicking on a tiny screen. Putting chatbots into small, wireless devices would equip users with an omnipresent aide. Mobile AI would become everyone’s own digital butler, serving up information tailored to suit each request. To personify their vision, the conference directors arranged for an English butler to host the event and introduce the speakers.
Only one speaker, Mark Weiser, then head of Xerox PARC’s Computer Science Lab, railed against the idea. Weiser believed that desktop computers already demanded too much attention from users, and that mobile AI would worsen the problem. He feared that relying on digital butlers might reduce people to the most reactionary of creatures. It wasn’t the accuracy of the information that worried him, but rather the means of acquiring it. When you routinely turned to AI for guidance, you relaxed your sensory grip on the world. Like a mansion dweller who looks to their butler to take calls and tend to the front door, the mobile AI aficionado can always ask software to explain situations to them (or to even communicate on their behalf). Users grow accustomed to prompting the system, Weiser contended — they lose the inclination to discern insights from their surroundings without software’s input. Dialogues with an AI intermediary become a default point of focus. The imperative to observe life as it’s happening loses its urgency.
Weiser tried to sell the conference audience a different vision. Rather than give everyone a chatty digital butler, he thought it better to embed computing power into the everyday things that populate homes, offices and cities. His broader goal was to bring online information into physical spaces, to subtly infuse one’s environment with relevant data that might prompt more informed ways of being in the world. Think today of the augmented reality navigation systems that now display routes and directions directly on a car’s windshield.
Weiser’s internet-enhanced objects wouldn’t feed the user a chatbot’s detailed answers, ending the inquiry as soon as it started. Rather, his first principle was to build on the physical world and keep our attention focused there. When we look around for clues or listen to a sound — when we think with our senses — we receive more information than we ask for. Feeling our way through a question or an idea, loopy as it is, helps us feel connected. Any technology that routinely severs that may not be worth the intelligence it’s purported to supply.
Ultimately, the computer scientist’s perspective failed to win favor at the world’s first conference on mobile AI, and it appears lost on today’s debates. We have been gauging the value of ChatGPT largely by its capacity to yield correct answers, compose good sentences and mimic human skills. Much less attention is paid to how ushering these chatbots into work and life stands to flatten the texture of experience. This becomes a more urgent consideration as mobile AI apps roll out. Our interactions with ChatGPT and its ilk, now unbounded, promise to more often eclipse the tacit sense of engagement that Weiser revered. The habits of mind and body we exercise in uncertain moments — when we’re forced to converse crudely with a vacuous world — are the moments that connect us to it most deeply.
John Tinnell is director of digital studies and an associate professor of English at the University of Colorado, Denver. His latest book is “The Philosopher of Palo Alto: Mark Weiser, Xerox PARC, and the Original Internet of Things.”
