In an early AI experiment, researchers allowed two chatbots to converse freely. Before long, the two bots shifted away from English and began developing their own shorthand, a form of communication optimized for machines rather than humans. They could have gone on chatting that way for years if given an unlimited power supply.
But that’s not how most of us are currently using AI. There is almost always a human in the loop. For now, anyway.
I recently attended a webinar on the use of AI technology in documentation. The presenter was blunt about the future of tech writing. If you’re doing this work solely for the love of writing standardized documentation, forget it. You’ll soon find yourself out of work. Technical writing today is mostly about producing content that AI systems can easily process and interpret. And the best way to produce such content is with LLMs prompted for that specific purpose.
The role of humans — for now, anyway — is simply to oversee this interaction between agents in case they go rogue or start hallucinating. But as AI improves, we’re told, such errors will be less likely. And in any case, by then, interactions will be so rapid, large scale and complex that it might be impossible for any human to intervene.
Documentation work may be the leading edge of this shift, but the same logic is spreading into other industries as automation improves. For now, hotel rooms still need people to clean them. Restaurants need cooks. Massages need masseuses. But given the right machines and training models, life could start to look a lot like a futuristic cartoon of robot servants, push-button meal prep and self-pumelling massage tables.
I’ve made it a point when I go shopping not to use the self-service checkout whenever possible. Maybe I’m out of step with the times, but I prefer to interact with a human. Some argue that it’s a cost savings. But I don’t see those machines bringing down prices. And now my cashier neighbor is out of a job.
In the documentation webinar on AI, the presenter spoke of the increased efficiency of using AI and the measurable real-world outcomes of those efficiencies. If you’re only measuring results, it doesn’t matter whether a human is involved or not. If your business is delivering a better end-user experience by implementing AI, then why would they forgo that in order to support someone’s love of tech writing?
I don’t know where the future of AI is heading. There will be some sort of balance between human and AI-driven processes rather than a complete replacement of one by the other. But AI will be involved in many aspects of our work.
What I do know with some certainty is that humans are social creatures, full of messy emotions. Open up any successful CEO and you’ll find a swarm of insecurities, frailties, and a hunger for human connection — just like the rest of us. AI isn’t going to change that.
I think we’re passing through a phase in society where we’re experimenting with what it’s like to remove humanness from the equation. Driverless cars. Teacher-less classrooms. Receptionist-less hotels. Composer-less music. I don’t think we’re going to be able to completely adapt to that sort of world without going a bit crazy.
I attend events where, ironically, groups of real people gather to have drinks, discuss and contemplate a future where AI agents do all our interactions without us. In my own work, I’m already seeing a quiet backlash to AI-generated content. People increasingly seek out work that feels unmistakably human — not necessarily perfect, but personal, imperfect, and alive.
Efficiency isn’t the only value worth optimizing for, and a civilization that forgets that will impoverish itself in ways that don’t show up in productivity metrics.