There’s an article I’ve sat down to write a hundred times. There’s a mountain of scraps, disjointed words and phrases I clamber over seeking an altar of perfection -- but perfection doesn’t get the job done.
When I was a kid, when I lived in Germany, my Dad picked up litter along the road and around the riverbank near the Rheine River where we lived. He went through daily, with a garbage bag, picked up trash and floating debris to clean up the environment around him. He was never obliged to do this for anyone, he didn’t owe anyone, he didn’t want to impress anyone – hell, it wasn’t even his country. But he did it because he wanted to do something, it was something he cared about a lot, and he wanted to make things a little better going forward.
The current driving trend in AI is the emergence of agentic AI, which conceptually is AI replacing the role of workers autonomously in performing a task. This is no secret, in an interview with right-wing podcaster Theo Von, Sam Altman stated he believed the purpose of AI was to give people “the agency to co-create the future together.” In a society where people scarcely avert their eyes from their smartphones like meerkats on the Serengeti, it might seem broadly tantalizing in the corporate space to swap human resources for synthetic ones; as natural a transition as an automatic transmission replacing a manual.
But there’s a fatal flaw with this line of reasoning. Procedural intelligence, the role of understanding how something works, its fundamental steps (and equally perilous footfalls), is not something so cut and dry. Consider replacing a battery, the steps in which you must disconnect and reconnect the electric current must be done in a specific order of operation, or death or serious injury will result. LLM’s are amassed knowledge culled from the Internet, stripped from context, arranged in the specific cue of the next, most likely output – a guesstimate. The balance at which it is providing you with the correct answer for the given question you are providing it probably based on the best Redditor (and who knows if they even survived after humble-bragging in complete confidence.)
The concerning issue with these models is they create the illusion of expertise. An equally beneficial measure of procedural knowledge is the ability to improvise. Even in a mundane, paperwork-mill office environment, files are renamed, networks disappear, UI’s get rearranged, functions are depreciated -- each time any miniscule underlying changes shift like sands, the information store in an LLM changes from an asset to a liability. They’re looking at a map, a caricature of a coastline, when the truth is infinitely more fractal.
In his book, “Normal Accidents: Living with High-Risk Technologies,” Charles Perrow introduces the notion of systems accidents, where tiny miniscule errors add up to catastrophes, this same theme explored for crude and violent effect in the film, “Final Destination.” The thing he is trying to say in his writing is that if enough small errors are introduced in different places in a daily process, eventually ‘the holes in the cheese’ line up in such a way that they cause disaster that could never be predicted.
I don’t think the creators of these models are blind to these problems, but I do believe they are wagering that the public is willing to accept the negative feedback of the mistakes that will ensue as something that can be inured while they operate in the real world. Certainly there are repetitious, joyless tasks that need to be automated. But how much of our highly interconnected, highly complex world is strictly within the confines of these tasks? These tools are solving for the wrong problem.
Instead of introducing new variables into a world falling apart at its seams, we need humans with the drive, the empathy, the accountability, and the expertise to step up and acknowledge that the only people who will save us are us. Deferring accountability to AI misses one of the oldest adages of ‘A Computer Cannot Be Held Accountable – Therefore a Computer Should Not Make a Managerial Decision.’
The deliberate architects of these technologies are basing their decisions not because they’re unaware of this history but because they want an excuse to defer for their own psychopathic choices.
One of my favorite shows is “Halt and Catch Fire,” which is the story about how human engagement drove the story of technology. It pretends to be a show about the history of innovation, but it’s cleverly really a show about human beings and how they inspire each other and how communication reshaped the world for the better.
The truth is human society is an endless thread of reciprocity, each generation bound to the next, echoing down through time; and we are stronger bound to it than tearing it apart.