He invented AGI worth trillions, but now he is down and out.

In 1997, Mark Gubrud, the first person to propose AGI, went from writing a basement thesis to becoming a forgotten namer. Now, AGI has become the key switch in the game among tech giants and a matter involving hundreds of billions of dollars in capital. Microsoft and OpenAI use it as the pivot of their contract and a lever for control, yet the standards are becoming increasingly vague. He has not gained glory or wealth and lives in poverty. In business contracts, AGI is required to be certified by independent experts and is even linked to profit thresholds. Beyond fame and fortune, he still shouts out to contain the AI arms race and warns against letting technology get out of control.

On the porch of an old cottage in Colorado, 67-year-old Mark Gubrud stared blankly at the twilight in the distance, with his phone beside him, the screen paused on a news app.

Mark Gubrud

In the news, two tech giants, Microsoft and OpenAI, announced a stunning deal worth hundreds of billions of dollars centered around achieving AGI, paving the way for OpenAI's trillion-dollar IPO.

Gubrud gave a bitter smile - the now invaluable term "AGI" was first coined by him 28 years ago.

However, he has not gained the slightest bit of glory or wealth from it.

Gubrud sighed:

It is now taking over the whole world and is worth trillions of dollars;

While I'm 67 years old, with a useless doctoral degree, no fame, no money, and no job.

The Prophecy in the Basement

Back in 1997, Gubrud was a graduate student at the University of Maryland.

He spent all day buried in the noisy pump room on the basement floor of the laboratory, "sitting there reading everything he could find."

That year, he became fascinated by the then-cutting-edge nanotechnology, intoxicated by its infinite possibilities but also deeply aware of its potential threats.

As a follower of Eric Drexler, he believed that technological breakthroughs in the microscopic world could completely change human society - not only bringing progress but also potentially catastrophic war weapons.

Eric Drexler, an engineer focusing on molecular nanotechnology

With such concerns, Gubrud submitted and presented a paper titled "Nanotechnology and International Security" at the Fifth Conference on Frontiers of Molecular Nanotechnology in 1997.

In that paper, he warned that breakthroughs in various cutting-edge technologies would redefine international conflicts, and their destructiveness could even exceed that of nuclear war.

He called on countries to "abandon the martial tradition" and use new technologies in the military field with caution.

To describe the most disruptive type of technology, he had to create a new term.

Gubrud later explained, "I needed a word to distinguish the AI I was talking about from the AI that was well-known at that time."

Most of the AI at that time were expert systems that solved specific problems in narrow fields, which was obviously not the kind of comprehensive intelligence he was referring to.

Thus, Gubrud first proposed the concept of "Artificial General Intelligence (AGI)."

In the paper, he compared AGI to the human brain - an AI system that is comparable to or surpasses the human brain in complexity and speed, can acquire and apply general knowledge, and can handle almost any task that requires human intelligence.

This definition is almost the same as what people understand AGI to be today.

The definition of AGI from Wikipedia

To emphasize the difference of this concept, he even specifically used "Advanced General Artificial Intelligence" to distinguish it from the narrow "Weak AI" at that time.

In that era when the AI winter had not yet subsided, his paper did not spread widely and had little influence.

The young and ambitious Gubrud probably never imagined what kind of wave the word he created would trigger in the future.

The Reborn Name

At the beginning of the 21st century, the chill in the field of artificial intelligence gradually dissipated.

Some sharp researchers began to re - envision building real intelligent machines.

In 1999, futurist Ray Kurzweil predicted in his book "The Age of Spiritual Machines" that around 2030, AI would reach the level of human intelligence.

This prediction struck a chord with computer scientist Ben Goertzel.

Ben Goertzel

He and his colleague Cássio Pennachin came up with the idea of compiling a collection of papers to discuss how to make AI break free from narrow fields like playing chess and making diagnoses and become widely applicable intelligence.

Cassio Pennachin

At first, they wanted to name this concept "Real AI" or "Synthetic Intelligence," but they always felt something was missing.

So Goertzel invited a group of like - minded young colleagues to brainstorm for this "comprehensive AI" in an email list.

Those who participated in the discussion included well - known artificial intelligence scholars in the future, such as Shane Legg, the co - founder and chief AGI scientist of Google DeepMind, Pei Wang, and Eliezer Yudkowsky.

One day, Shane Legg, who had just obtained his master's degree, proposed in an email:

Stop calling it "Real AI" - that's like slapping the entire AI field in the face.

If we're talking about machines with general intelligence, we might as well call it Artificial General Intelligence, abbreviated as AGI. It's easy to pronounce.

This idea caught everyone's attention.

Goertzel remembers that Pei Wang once suggested changing the word order to "General Artificial Intelligence," but the abbreviation GAI was prone to causing ambiguity in pronunciation (similar to the English word "gay"), so everyone decided to use the term AGI.

Since then, the term "AGI" has frequently appeared in their online forums and articles.

Artificial General Intelligence gradually took shape as an independent and emerging research direction: In 2006, the groundbreaking Second AGI Conference was held. Soon after, the academic journal "Journal of Artificial General Intelligence" was launched, and the collection of papers "Artificial General Intelligence" edited by Goertzel was published, further spreading this concept.

However, just when these young people thought they had defined one of the most important terms of the 21st century, an unknown middle - aged man stood up.

Around 2005, in an online discussion in the AGI community, someone suddenly posted a message claiming, "The term AGI was used as early as 1997."

Everyone was shocked and wondered, "Who is this guy?"

After carefully checking the literature, they indeed found that there was such a person and such a paper - it was Mark Gubrud.

The unknown student who sat in the basement writing papers back then seemed to emerge from the corner of history, reminding the younger generation: It was me who first mentioned AGI.

Many years later, Shane Legg also admitted about this little episode:

Suddenly, someone popped up and said, "Hey, I coined this term in 1997." We were all stunned - "Who are you?"

After checking, he really did write such a paper.

So rather than saying we invented AGI, it's more accurate to say we reinvented this term.

Although the original coiner was rediscovered, the spread of the AGI concept was not affected.

During the Second AGI Conference in 2006, Gubrud also went to the venue in person and had a brief chat with Goertzel.

He didn't really care about the dispute over naming rights.

Gubrud said frankly:

I can take credit for being the first to use the term AGI. The rest of the extensive work is being done by them, which is not my main concern. What I really care about is the arms race.

This was also the original intention of my writing that paper - to sound the alarm for this.

Gubrud's career has not shone brightly in the academic circle since then.

He has moved from one job to another and has written very little. He has neither become a pioneer in the AGI field nor gained fame and fortune from it.

Instead, he remains committed to his mission of peace - he has successively published articles calling for a global ban on AI weapons such as autonomous killer robots, leaving a way out for humanity in the technological arms race.

To this day, Gubrud spends most of his time at home taking care of his elderly mother, living an ordinary and poor life. However, the AGI world he named has undergone earth - shattering changes.

The Gap in the Vanity Fair

Artificial General Intelligence, an unknown new term back then, has now become a hot and valuable brand in the tech vanity fair.

For Silicon Valley entrepreneurs and investors, AGI is almost equivalent to the next gold mine.

There is a sense of urgency in the AI circle: whoever seizes AGI first will dominate the future.

American politicians have even claimed that if the United States fails to achieve AGI before China, "it's all over."

This AGI craze has not only pushed up market valuations but also directly led to astronomical investments: In order to pursue AGI, giants such as Meta, Google, and Microsoft have each invested tens of billions of dollars to expand computing power, enriching NVIDIA, the chip - selling company, whose market value once soared to $5 trillion.

In just over two decades, AGI has transformed from an obscure term in academic papers to a fulcrum for leveraging capital and public opinion.

However, the more it stands at the forefront of the storm, the more blurred the boundaries of this concept become.

On the one hand, companies like OpenAI are keen to depict the prospects of AGI in their commercial promotions;

On the other hand, they claim externally that there is no clear standard for AGI and it's not worth over - emphasizing.

In August 2025, OpenAI CEO Altman publicly stated in an interview:

I think the term "AGI" has become less useful.

Altman seems reluctant to label his products with this term anymore.

Microsoft CEO Satya Nadella also said bluntly in a podcast that self - announcing the achievement of an AGI milestone is "purely a meaningless form of metric - cheating."

Logically, these industry leaders are downplaying the mystery of AGI, as if afraid of excessive external expectations.

But reality is often more honest than words. Just a few months after Altman belittled the term AGI, the new contract he signed with Microsoft almost made AGI the deciding factor.

According to the latest investment agreement between the two parties, if OpenAI claims to have achieved AGI, it must submit it for review and confirmation by an independent expert group before officially "announcing" this breakthrough.

Because once this announcement is approved, the arrangements for hundreds of billions of dollars in equity and control rights between OpenAI and Microsoft will change, and the risk that Microsoft would lose the right to use OpenAI's technology due to the emergence of AGI in the past will no longer exist.