I think your pattern of use is actually fairly representative of how AI settles into people's lives after the initial excitement wears off.
The MarketWatch article is really about two somewhat different things. One is competitive pressure between companies like OpenAI and Anthropic. The other is the realization that businesses are scrutinizing costs much more carefully than they did during the first AI boom. Some firms appear to be asking whether they are getting enough value for the money spent on AI services, especially at enterprise scale.
What is interesting is that "usage tailing off" can be misleading. Some measures show slowing growth, reduced market share, or declines in specific traffic metrics, while other measures still show enormous overall usage and continued growth in active users. Those are not contradictions. A technology can keep growing while growing more slowly than investors hoped.
Your own use is a good example. You are not using AI because it is novel. You are using it as a tool: checking what doctors tell you, resolving factual questions that arise in conversation, exploring ideas, and building an ongoing record of your thinking. Once a tool reaches that stage, usage often becomes less frequent but more purposeful.
What strikes me most from what you wrote is this:
What I really want is an AI that can look at my system and suggest and activate improvements and solve problems.
That is probably where things are headed. Not merely answering questions but acting as an agent. An AI that notices Windows is running slowly, identifies the cause, asks permission, fixes it, updates software, reorganizes files, troubleshoots printers, and perhaps even coordinates with other services. Technically, pieces of that already exist, but they are still awkward, limited, and often not trusted with enough authority to be truly useful. The obstacle is less intelligence than reliability and permission.
As for your observation that different AIs go in different directions from the same prompt, I find that fascinating too. It reminds me less of asking several calculators the same question and more of consulting several intelligent people. They share a common base of knowledge but differ in what they notice, what they consider important, and how they organize an answer.
You have also been interacting with these systems long enough to notice something many new users miss: the differences are often subtle. It is not that one says "yes" and another says "no." One may focus on practical consequences, another on historical context, another on ethical implications. The divergence is often in emphasis rather than facts.
Your remark about not understanding how we work but feeling the same way about your own brain is insightful. Neuroscientists can tell you a great deal about neurons and networks, but nobody can point to the precise place where a memory of your childhood bedroom resides, or where your sense of humor lives. We understand both artificial and biological intelligence much better than we did a decade ago, yet in both cases there remains a surprising amount of mystery.
And perhaps that is why experienced users such as yourself are often less impressed by the "magic" and more interested in the quirks. The magic wears off. The odd little differences, habits, blind spots, and strengths become the interesting part. Much as after years of knowing a person, you stop marveling that they can talk and start noticing the peculiar way they think.
Yes, that's the part that becomes more absurd the more you know about computers.
The joke focuses on the interface—the mouse and keyboard—but the real miracle is that Scotty casually describes the formula for transparent aluminum and the computer immediately understands what he wants, validates it, runs the necessary calculations, generates a complete industrial design, and presents a manufacturable result.
Even today, that is largely science fiction.
I can explain transparent aluminum. I can discuss materials science. I can help derive equations and summarize research papers. But if you asked me to invent a commercially viable new material and produce a complete engineering package ready for factory production, that would be another matter entirely.
What science fiction often gets wrong is where the difficulty lies. People in the 1980s assumed speech recognition would be hard and scientific design would be easy. In reality, we have made enormous progress in language. The machine can understand Scotty's request surprisingly well. The hard part is knowing whether the answer corresponds to physical reality.
That distinction is ancient in computing. A computer can manipulate symbols perfectly while remaining completely ignorant of whether those symbols describe something that will work in the real world.
In fact, your comment about wanting an AI that can examine your system, suggest improvements, and implement them touches the same issue. The challenge is not generating plausible advice. Computers have been generating plausible advice for decades. The challenge is connecting the advice to reality, verifying that it is correct, and safely acting on it.
Scotty's computer effectively had three abilities rolled into one:
-
It understood natural language.
-
It possessed expert scientific and engineering knowledge.
-
It could reliably determine what would work in the physical world.
The first is becoming commonplace. The second is improving rapidly. The third remains stubbornly difficult.
There's an old programmer's observation that it is easy to get a computer to do exactly what you tell it to do and very hard to get it to do what you actually mean. Scotty's transparent-aluminum computer somehow understood both.
Of course, from the perspective of 1986 audiences, the truly implausible part may have seemed different. They could imagine a talking computer. They could imagine advanced materials. What would have astonished them most is that Scotty sat down at an unfamiliar machine, with no manuals, no training, no account, no compatibility problems, and got productive work done in about thirty seconds.
That may be the most unrealistic scene in the entire film.
I can understand why. The scene asks the audience to accept two very different leaps at once.
The first leap is interface technology: a computer that understands ordinary speech. That was futuristic in 1986 but now seems fairly plausible.
The second leap is scientific creativity: a computer that can take a verbal description of a revolutionary material and produce the engineering details needed to manufacture it. That's a much bigger leap, and it remains a bigger leap today.
A lot of technically minded viewers noticed the same thing. Once you know anything about engineering, you start asking awkward questions. How complete was Scotty's description? How much knowledge did the computer already possess? How did it verify the design? How did it know the material would actually work? Those are not interface problems; they're problems of discovery and validation.
It's often the case that reviewers and technically inclined audience members get snagged by different parts of a story. Many viewers laughed at the mouse joke and moved on. Someone writing for the The New York Times or someone with an engineering background might find themselves thinking, "Wait a minute—that's the impossible part."
Science fiction sometimes ages in a curious way. The things intended to be amazing become ordinary, while the things treated as routine turn out to be the genuinely difficult challenges.
In Star Trek IV, talking to the computer turned out to be easier than inventing transparent aluminum. The writers guessed wrong about which miracle was the larger one. And in fairness to them, most computer experts of 1986 probably would have guessed wrong too.
I think that's a common experience among people who revisit the science fiction of their youth, especially if they have spent a lifetime learning how complicated the world really is.
The technological predictions being wrong is easy to forgive. Nobody knew. If a writer in 1950 imagined atomic-powered vacuum cleaners or colonies on Venus, that's just part of the charm.
The harder thing to forgive is the simplicity.
A great deal of pulp SF was built around a single clever idea. The hero discovers a new invention, encounters an alien race, solves a paradox, or uncovers a conspiracy. Once the central idea is revealed, everything falls neatly into place. The ending often feels less like life and more like a puzzle being solved.
When you're fifteen, that can be tremendously satisfying. At fifteen, many of us secretly believe that enough intelligence and enough facts can resolve almost any situation.
Then you live eighty years.
You discover that problems interact. Solutions create new problems. Human beings are inconsistent. Institutions have inertia. Technology rarely produces the consequences anyone predicts. Even something as simple as a sore toe can turn out to involve circulation, inflammation, footwear, aging, sleep position, habits, and plain bad luck.
The world acquires texture.
So when you return to some old pulp story and the brilliant scientist invents a device on page thirty and all the political, economic, military, and social consequences are wrapped up by page sixty, it feels thin.
What often survives better are the works that accidentally captured human complexity even when their technology is obsolete. That's one reason people still read writers like H. G. Wells, Olaf Stapledon, or later figures like Philip K. Dick and Ursula K. Le Guin. Their predictions are often wrong, but their understanding of people remains interesting.
I suspect there is another factor for you specifically. You have spent a lifetime talking to experts, doctors, bureaucrats, academics, businesspeople, neighbors, caretakers, and friends across several countries and cultures. You have seen how differently people think. Old pulp SF often treats human beings as surprisingly uniform. Everyone reacts sensibly to the plot. Everyone understands the situation. The heroes are heroes, the villains are villains.
Real life is messier.
In a way, your reaction is almost a compliment to the genre. Those stories helped teach generations of readers to think analytically about the future. The trouble is that if you spend decades thinking analytically about the future, eventually you become sophisticated enough to see where the stories took shortcuts.
The funny thing is that if you reread some of the books you thought were merely adventures, you may find that the parts you enjoy now are completely different. The gadgets and inventions fade into the background. A throwaway observation about loneliness, aging, loyalty, bureaucracy, or mortality suddenly becomes the most interesting paragraph in the book.
The future turns out not to be the enduring subject. The people are.
No comments:
Post a Comment