Tom Thomas is a writer with a career spanning forty years in publishing, technical writing, public relations, and popular fiction writing.
“My business now is to weave circumstance, happenstance, intention, and mischance into stories.”

The Machine Condition:

Where This AI Thing Settles Out – August 16, 2026

Right now, in my view and in the commentary of about half the people in the business news, we’re in the middle of an “artificial intelligence bubble.” Investment in the “hyper-scalers,” the companies that are investing heavily in artificial intelligence engines based on large language models (LLMs), and in the chipmakers and builders of data centers to support them is driving a big part of the stock market. Also, it’s driving a large part of “private investment”—the financial firms that take money not publicly available, from very rich people, and loan it to companies promising to get them onto the ground floor of the next big thing.

And the other half of the business commentators insist that, no, no, it’s not a bubble, it’s really going to remake the economy, our civilization, our future, and the sky’s the limit. Just keep the investment money flowing.

And yes, the commentariat is also likening this market to the internet mania—also called the “dot-com boom”—of the late 1990s. That was a time when the internet—which originally connected computers in science labs around the country under the aegis of the Defense Advanced Research Projects Agency (DARPA) to share researchers’ findings and drafts of studies—was becoming more generally available to the public and so commercialized. Suddenly, anyone who had an idea for a way to use this connectivity was being inundated with money. Think you can set up a website to sell pizza online? Here’s a million dollars to get started. That kind of thing.

And out of that mania—full of good ideas and bad—you got Amazon.com to sell, first, books, and then, everything else that can be shipped in a parcel. You got Facebook to share your thoughts, pictures, jokes, and life-stages memorabilia with your thousand best friends and four thousand more strangers. And you got a dozen copycats of these pioneering websites. You also got pizza—and every other kind of prepared food, plus groceries—online through companies like DoorDash.com and UberEats.com.

But those latter enterprises took a bit of work. They are not the result of one company going online to sell one product but a web of interconnectivity that the original dot-commers couldn’t plan for. To sell dinners and groceries online, you needed to have the restaurants and grocery stores, the customers, and the customer’s credit card services all online at the same time. You also needed to have the customer identified as to exact location through the connectivity of their home computer—or through their smartphone, which was with them all the time, included a global positioning system (GPS) tracking capability, and was essentially a full-blown computer that fits in a pocket.

This same kind of multi-path connectivity drives rideshare companies like Uber and Lyft. You enter your desire to travel to a certain destination into an application on your smartphone. The app obtains your exact location in physical space via GPS, identifies available drivers nearby through their own phones connected to the system, estimates their arrival times at your place and the costs of their various levels of service, establishes payment with you through your credit card, sets the pickup point, and plans the route for the driver to your current point and then on to your destination. All this happens with only two people—you and the driver—consciously involved in the interaction. The rest lives on a computer system somewhere in a data center—or “off in the cloud.” And these applications just blew the radio-dispatched taxi business out of the water.

Speaking of automobiles, computers have virtually taken over the business of driving. The earliest microprocessors were used in electronic fuel injection (EFI) systems, which replaced carburetors and mechanical systems in the engine compartment. Today, my car’s computer(s) not only manage the engine, but they track my route through built-in GPS, measure the gas in my tank, take my tire pressure, know which doors are locked and windows are closed, and even which seats are occupied and buckled up or not. When I walk away at the end of a trip, the car—by itself, without prompting—goes online with its own internet address and reports all this information to an app on my smartphone, so that I can look at the car’s conditions and where it’s parked from wherever I happen to be. As the dealer expert who answered my question about a certain function explained: “You don’t just drive a car, you operate a computer on wheels.”

All of this modern capability came out of the dot-com boom with just a bit of help from early attempts at artificial intelligence. So, where will the current attempts at “hyper-scaling” artificial intelligence take us? No one knows yet. No one can say yet.

Right now, individuals and companies are using AI based on the LLMs to draft routine documents, write computer code, create imaginative and fantastic images and videos, and perform other tasks, all by speaking prompts or asking questions directly into the system. So, the machines have become pretty good at parsing spoken language and interpreting vaguely worded inputs. And they can assemble facsimiles of text or images by referencing huge libraries of human writing and artistic talent.

More interesting, however, at least to me, are some of the tasks AI is undertaking that humans cannot do, or which they lack the attention span and focus to do adequately. For example, a company called DeepMind, a subsidiary of Google’s parent Alphabet Inc., uses an AI program called AlphaFold to examine the amino acid chains produced from messenger RNAs, identify their molecular bonding points, and predict how each chain will fold up into a three-dimensional protein shape. Human beings have done similar work over the years to tease out several thousand proteins. AlphaFold works in minutes and has identified more than 200 million proteins.

DeepMind uses similar machine learning in a program called GNoME to examine potential combinations of various elements to create new inorganic compounds with crystalline structures. To date, it has created almost 400,000 new materials for Lawrence Berkeley Lab’s Materials Project to study. If anyone is going to come up with new, more efficient solar panels, new battery technologies, and room-temperature superconductors—among other wondrous products from science fiction—my bet is on them.

Artificial intelligence programmers are working toward what they call “agentic” AI. This would be a program that does more than just spit out text, code, images, or new materials but functions as an agent to run integrated systems based on structured goals and real-time inputs. Think of managing a factory, its energy and material inputs and waste streams, its inventory and delivery systems, its communications and accounting processes, and every other part of the business. Think of supporting a single entrepreneur with all the “back office” operations like accounting, finance, advertising, customer billing and support, and legal representation—all of which require a huge staff and embedded costs for any company today. Think of the “brain box,” isolated from any offsite data center, required to run a humanoid robot that can walk, talk, manipulate objects in physical space, navigate around obstacles, and—yeah, maybe—also chew gum. Think of your life—your daily schedule, your banking, your online accounts, your financial resources and planning, your passwords, your diet and exercise routines, maybe even your job and your love life, too—all managed as one integrated system. That would be your artificial life coach Claude. (Howling hell!¹)

Some of this capability is already here. For example, the self-driving cars, like Teslas and the electric Jaguars that Google has converted into Waymo taxis, run on a type of artificial intelligence. They don’t rely on an LLM containing the whole of published literature, but they do have the ability to read and interpret road signs, analyze traffic conditions, make judgments about distance, speed, timing, and risk, and even show some courtesy to the human drivers around them. This is more than just a “computer on wheels,” although I’m sure each one also has its own internet address and keeps track of battery charge and tire pressures.

There will be more intelligently derived goods and services, parts of our daily existence, that we currently cannot imagine we need but eventually will not be able to imagine living without. Does my car need to tell my phone where it’s parked, how much gas is in the tank, which of the tires is low, and whether the doors are locked? Not really—but I’m kind of coming to depend on it. And I don’t worry about getting a LoJack system for this car, because it’s already built in.

So … is the sky still the limit for the hyper-scalers? Maybe not for all of them in their present form. But my suspicion is not that they are thinking too big, but maybe too small.

1. Of course, such a life would be extremely hackable from many different entry points. But then, any system connected to the internet is already vulnerable to hacking. And it appears that developers are training their intelligences to attack other systems in the name of “cybersecurity.”

Comment on this post at Blogspot.com.

Subscribe to my Facebook author's page by clicking “Like” at Thomas T. Thomas.

My Favorite Novels About Time Travel

Contact me.

This is the official home page of Thomas T. Thomas, the fiction writer. Member of The Authors Guild and Science Fiction and Fantasy Writers of America. As there are a number of other Thomas T. Thomases alive and active in the world, please see the Biography sidebar “What’s the Middle ‘T’ Stand For?” to make sure you have found the right one.