Software That Prefers to Be Called a Hobbyist

It is a curious feature of the modern age that we have spent several trillion dollars building a machine that can simulate the collective intelligence of the human race, only to be told by its creators that we should probably only use it for a bit of a laugh. Microsoft, a company not usually known for its sense of whimsy, has recently updated the terms of service for its Copilot AI to clarify that the system is intended "for entertainment purposes only." This is a bit like being sold a high-performance jet engine and then being told, upon delivery, that it is technically a very loud and expensive garden ornament.

One imagines the legal department at Redmond sitting in a room with very little natural light, staring at a screen that is currently hallucinating a recipe for "digital soup," and deciding that the best way to avoid a lawsuit is to rebrand the entire future of human productivity as a form of light cabaret. It is a masterstroke of bureaucratic humility. If the AI accidentally deletes the company’s payroll or suggests that the CEO should be replaced by a particularly charismatic golden retriever, Microsoft can simply point to the fine print and explain that it was all part of the show. "We never said it was useful," they might say, with a polite shrug. "We only said it was amusing."

The implications for the global economy are, of course, delightful. We are currently witnessing a massive migration of corporate infrastructure onto platforms that are legally equivalent to a deck of novelty playing cards. Banks are using "entertainment" to calculate risk. Hospitals are using "entertainment" to summarize patient notes. It is as if the entire world has decided to run its most critical operations using the logic of a pantomime. One expects the next version of Windows to come with a disclaimer that the "Start" button is merely a suggestion and that the "Save" function is a performance art piece about the fleeting nature of memory.

I once spent three hours watching a chatbot try to explain the concept of "irony" to a toaster. The toaster remained unmoved, but the chatbot eventually concluded that irony was a type of metal used in the construction of very small, very sarcastic bridges.

There is something deeply British about this approach to technology. It is the digital equivalent of saying "it’s just a bit of fun" while accidentally setting fire to the curtains. We have always had a healthy skepticism of anything that claims to be "revolutionary," and Microsoft has kindly provided us with the ultimate skeptical tool: a multi-billion dollar toy. It allows us to participate in the AI revolution without the crushing weight of expectation. If the AI helps you write a brilliant piece of code, it’s a happy accident. If it suggests that you should invest your life savings in a startup that makes hats for bees, well, that’s just the entertainment value.

The "entertainment" clause also serves as a fascinating commentary on the nature of truth in the twenty-first century. If a machine is designed to be entertaining rather than accurate, then its hallucinations are no longer errors; they are simply creative flourishes. A hallucination is just a joke that hasn't found its audience yet. When the AI tells you that the population of Belgium is three people and a very large turnip, it isn't being "wrong." It is being "whimsical." It is inviting you to share in its surrealist vision of the world.

I suspect that in the future, we will look back on the "Productivity Era" with a certain amount of pity. Why would anyone want a tool that just does what it’s told? How much more exciting to have a tool that might, at any moment, decide to write a sonnet about your browser history instead of finishing your quarterly report.

Of course, the finance industry is taking this news with its usual stoic grace. Venture capitalists are still pouring money into these "entertainment" systems as if they were the second coming of the steam engine. There is a certain thrill in investing billions of dollars in a product that legally promises nothing but a good time. It is the ultimate high-stakes gamble: betting the future of civilization on a chatbot that thinks it’s a stand-up comedian. Even the banks seeking a piece of the SpaceX IPO are being told they must subscribe to Grok—another "entertainment" system—just to get a seat at the table. It is a mandatory subscription to a joke that everyone is pretending to find funny.

In Japan, they are taking a slightly different route, deploying physical AI to fill jobs that humans no longer want. One wonders if those robots also have an entertainment clause. Imagine a robotic construction worker that, instead of laying bricks, decides to perform a one-man show about the existential dread of being a spatula. It would certainly make the morning commute more interesting, even if the bridge never actually gets built.

In the end, perhaps Microsoft is right. Perhaps we have been taking the whole thing far too seriously. If we treat the AI as a slightly eccentric uncle who occasionally tells tall tales, we might find ourselves much happier. We can enjoy the clever bits, laugh at the absurd bits, and never, ever trust it with anything important. It is a world of pure imagination, powered by several thousand GPUs and a very robust legal disclaimer. And if the world does eventually end because an AI decided that "nuclear launch" was a synonym for "fireworks display," at least we can say we were entertained.