Marvel: A Lazy-Ass Superman

For 40 advanced chapters, visit my Patreon:

Patreon - Twilight_scribe1

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Although Henry wasn’t actually in Alaska anymore, Tony Stark had confirmation that a certain alien was still alive.

The reason was simple: on the secret message board they had once used to exchange information, the former West Coast employee had left one final message after resigning.

The site used a randomly generated URL.

After tracing the domain registration and resolving the IP address, Tony discovered that the server was physically located at Henry Brown’s rented residence in Los Angeles.

Opening the webpage revealed an extremely crude browser-based chess game.

There was also a chat box beside it.

Not entirely sure what he was looking at, Tony decided to give it a try and see how good the creator was at chess.

The result was terrible.

Absolutely terrible.

The moves followed the rules of chess, but beyond that, it was little better than random play.

Tony Stark—a genius who had barely played a handful of games in his entire life—easily checkmated his opponent.

The moment the king fell, the board instantly reset to its starting position.

Only now the colors had switched.

The first move belonged to the other side.

However, the opening moves from White immediately annoyed Tony.

White simply copied the opening sequence from Tony’s first game while completely ignoring the responses Tony had made.

After capturing a single piece, White’s play devolved into utter nonsense.

So Tony typed an insult into the chat box.

> What kind of garbage is this?

A reply immediately appeared.

> Hello. I am a self-learning artificial intelligence, Miss_A_111198.

Tony fell silent.

So the thing he had just played against wasn’t a person at all?

A self-learning artificial intelligence?

Tony proceeded to enter a series of questions.

Every answer followed the same pattern.

The entity identified itself as a self-learning AI called Miss A.

As for the string of numbers at the end, it looked suspiciously like a version date.

When Tony asked:

> What are your rules?

Miss A immediately produced a lengthy response.

In reality, it was simply a complete explanation of the basic rules of chess.

Tony quickly discovered that any question containing the keyword ”rules” triggered the same response.

There was a third keyword as well:

”game record.”

Whenever he entered that phrase, the computer instantly reconstructed the chess match they had just played.

Then the board returned to the exact state it had been in several moves earlier.

Driven by a growing suspicion, Tony completed a second game.

Again, he won effortlessly.

Afterward, he realized he could freely retrieve the records of either the first or second game.

At that point, if Tony Stark still couldn’t figure out the purpose of this web-based chess program, then his reputation as a genius would have been undeserved.

The chess itself wasn’t the point.

Winning wasn’t the point.

The point was the artificial intelligence behind it.

The earliest AI research had emerged from game theory in the 1950s.

Put simply, it focused on finding optimal solutions among numerous intersecting choices.

In mathematical terms, it resembled the shortest-path problem in graph theory.

By the 1970s, AI research had shifted toward knowledge engineering.

The goal became understanding complex human decision-making and helping generate conclusions.

In simple terms, researchers attempted to classify an infinite number of problems into a finite number of solution categories.

Whenever an existing solution couldn’t handle a new problem, another category would be added.

The system expanded continually in this fashion.

Applied to everyday life, it resembled automated factory machinery capable only of performing a predefined set of actions.

People believed that if enough actions were programmed into a machine, it would eventually behave like a living human being.

For a very long time, this was humanity’s vision of artificial intelligence.

Including things like Skynet and its Terminator armies from The Terminator.

Even Dum-E—the robotic arm Tony Stark had created as his undergraduate thesis at MIT when he earned his bachelor’s degree at sixteen—was based on this philosophy.

It wasn’t until neural-network theory gained traction in 1991 that machine learning truly entered the computing stage.

Even then, machine learning remained only one branch of AI rather than the dominant approach.

The computational cost was simply too high.

Available hardware couldn’t provide the necessary processing power.

Only in the twenty-first century, when computer hardware advanced sufficiently and concepts like deep learning and big-data analysis ceased looking absurdly impractical, did learning become a central concept of real-world AI development.

Before that, robots capable of independently learning new things existed only in science fiction novels, films, television dramas, and games.

The prevailing scientific belief was straightforward:

If humans hadn’t explicitly designed a behavior, a robot couldn’t perform it.

And in practice, that was generally true.

However, in the Marvel Universe, Tony Stark had earned his doctorate in artificial intelligence from MIT several years earlier.

The centerpiece of his dissertation had been J.A.R.V.I.S. (Just A Rather Very Intelligent System).

Built upon the foundations of Dum-E and neural-network theory, J.A.R.V.I.S. was itself a learning AI.

Running on a supercomputer platform, J.A.R.V.I.S. could observe interactions with Tony Stark, memorize new behavioral patterns, and apply them to similar situations later.

Tony never worried about accidentally creating a ”Skynet.”

Because fundamentally, J.A.R.V.I.S. was still an advanced command-and-response system.

It simply had an enormous number of functions.

It didn’t decide to do things on its own.

Given an instruction, it searched for the best method of accomplishing the task.

If Tony corrected it, it quickly learned and incorporated the correction into future decision-making.

Confident in the maturity of the system, Tony copied a version of J.A.R.V.I.S. into the Hollywood Kid supercomputer and had it play against Henry’s Miss A.

The two AIs played astonishingly quickly.

Had they not been limited by internet transmission speeds, they probably could have completed several games per second.

Even so, the simplistic web-based board still required time to load and refresh.

At several seconds per game, the pace was already so fast that Tony struggled to follow it.

Eventually he found a workaround.

Instead of displaying every move, he let the two AIs play in the background and recorded only wins and losses.

Watching J.A.R.V.I.S. steadily rack up victories, Tony felt quite pleased.

Then he went to bed.

He wanted to see just how thoroughly J.A.R.V.I.S. could crush Henry’s Miss A before some kind of surrender mechanism activated.

What he hadn’t expected was to wake up feeling as though the sky had fallen.

By the time he stopped the automated matches, the two AIs had spent roughly seven hours playing nearly one hundred thousand games.

J.A.R.V.I.S.’s overall win rate was only 30%.

And even that thirty percent was concentrated almost entirely in the early stages of the experiment.

Which meant that later on...

J.A.R.V.I.S. had been utterly massacred by Miss A.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

🎉 Power Stone Goal Announcement! 🎉

I’ll release one bonus chapter for every 500 Power Stones we hit!”

Let me know what should I do

Your support means everything—let’s crush these goals together! Keep voting, and let the stones pile up! 🚀

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~