Laughing at AI Won’t Stop What’s Coming
“AI Fails” has suddenly become a popular genre of internet entertainment. Social media is full of clips showing ChatGPT advising someone to walk to a car wash instead of driving, image models mistaking an upside-down cup for a prank device, or AI-generated videos where actors sprout extra limbs and bodies twist into impossible shapes. The focus is almost always the same: the strange things AI gets wrong. The hashtag itself reflects something deeper than amusement. It hints at a quiet desire to see the technology stumble rather than succeed.
I’m not here to defend AI. Today’s systems are, at their core, machines trained to predict the next token in a sequence. But it is important to separate what these systems genuinely do well from what they do poorly, and to understand why. It is equally important to remember that we are still at the very beginning of this technological shift, and the capabilities are evolving at a remarkable pace.
Today’s foundation models may resemble hyper-powered autocomplete machines, but scale has changed what that actually means in practice. Given a portfolio of patents, these systems can now identify and analyze thousands of products, determining which products are most likely to infringe which patent claims. That is a task that previously might have required a team of specialists working for months. And that is only one example. The same models are now drafting legal briefs, summarizing hundreds of pages of medical records, translating between languages in real time, generating production-quality software code, designing marketing campaigns, and tutoring students across entire curricula. In the software world, developers describe completing complex projects in weekend “vibe coding” sessions using tools such as Claude Code. Whatever else these systems are, trivial they are not.
Most people sharing “AI Fails” videos believe they are exposing the limits of artificial intelligence. In reality, they may be revealing something else entirely: how uncomfortable we are with the speed at which it is improving.
This is really happening
In many ways, the “AI Fails” phenomenon feels like a form of denial. It is easier to laugh at the mistakes than to confront the possibility that this technology may fundamentally reshape how we work and live.
But denial rarely works for long. In some ways the reaction resembles that of horse owners at the dawn of the automobile. Early cars frequently broke down or became stuck in the mud. Horses were sometimes used to pull them out. Skeptics pointed to these failures as proof that automobiles would never replace horses. After all, the machines could not yet do what a horse could do. History, of course, had other plans.
What do humans actually do?
As my wife puts it: “AI is forcing us to consider what our value as humans really is.” Even in its infancy, the technology is revealing that many tasks we once considered uniquely human, from bookkeeping to drafting documents to writing code, can now be partially automated. A great deal of modern knowledge work turns out to be structured pattern recognition, and machines are increasingly good at that.
But automation of tasks is not the same as replacement of people. Humans still define the problems worth solving. We set the goals, interpret ambiguous situations, and decide what outcomes matter. Machines can propose answers, but they do not decide what questions society should ask in the first place.
In practice this means the human role shifts rather than disappears. We become editors, supervisors, and architects of systems rather than the sole performers of every task. Judgment, context, responsibility, and moral reasoning remain stubbornly human domains. So do the less measurable qualities that drive human achievement: talent, ambition, creativity, and the desire to build something meaningful. Machines may generate options, but humans still decide which ones should exist in the world. And when it comes to creating great art, music, and culture, the human spark remains unmistakable.
This is why the endless stream of “AI Fails” clips misses the larger story. Every transformative technology begins with visible limitations. Early airplanes barely stayed in the air. Early computers filled rooms and crashed constantly. Early automobiles frightened horses and broke down in the mud. What mattered was not what the first versions could not do, but the direction the technology was moving.
Ridiculing AI will not slow that trajectory. The more important question is how we choose to shape it. These systems will not define human value for us, but they are forcing us to examine it more carefully than we have in generations. If machines can increasingly perform our tasks, the question becomes unavoidable: what does it truly mean to contribute as a human being?