Quick Glossary

.

AI –

Artificial Intelligence is a term encompassing any & all technology that involves computers having abilities which approach or exceed those of humans.

In the past, this has referred to a variety of inventions. Today, in nearly all contexts, AI is a synonym for neural network.

.

Neural Network –

A system, stored on a computer, which is designed in a way that mimics a brain.

Neural networks are not programs, they are not software, they are not algorithms, and they are not made of code. This makes it extremely difficult to understand or predict the actions of a massive neural network.

.

AI Model –

An iteration of an AI within its brand. Like how Windows 98 & Windows 7 are different versions of Windows, GPT-3 & GPT-5 are different AI models of GPT. A model can be a released product, or an unreleased prototype.

AI Training –

The technical term for AI development.

Because the process is primarily trial and error, extremely smart and capable AIs have been developed by people who don’t understand what they’ve built.

Although AIs are not algorithms, AI training involves algorithms.

.

Training Data –

Any content provided to an AI for use in a training process where the AI is tasked with replicating or recreating the data.

Training on data, without the consent of the copyright owners of said data, is not legal in many countries. However, governments continue to fail to hold the AI developers accountable for this copyright infringement.

.

AI Alignment –

Techniques to mold an AI’s behavior. This is often associated with attempts at making AIs more “moral” and less “deceptive”.

Top AI researchers have increasingly indicated that AI alignment is impossible because of the difficulty in understanding how AIs work.

.

Partial RSI –

Partial Recursive Self Improvment is when an AI assists humans in building another AI model smarter than itself. This process began on May 14th, 2025, when the AI AlphaEvolve autonomously discovered how to reduce the time it was taking to train Google Gemini by 1%.

Throughout 2026, developers began to increasingly involve AIs in the AI development process. By September, Anthropic claimed that “Claude is leading 26% of Anthropic’s model research and development…”

.

Full RSI –

The hypothetical future scenario where an AI can build another AI smarter than itself without any human assistance.

AI researchers refer to this as the “point of no return” for AI development, where further development can not be stopped.

.

Frontier AI –

A category of AI model, denoting the smartest and most capable models currently in existence.

.

AGI –

Artificial General Intelligence is a term that has changed drastically. First coined in 1997, it meant an AI that succeeds in a wide range of tasks. The 2015 OpenAI Charter defines AGI as “highly autonomous systems that outperform humans at most economically valuable work”. By 2025, it began to refer to an AI that is better than all living experts at all skills and benchmarks.

According to earlier standards of AGI, the frontier AIs we have today are AGI.

Pictured: Illustration from the 2005 book The Singularity is Near.

.

ASI –

Artificial Super Intelligence, a hypothetical future AI which is roughly 10 times smarter than the average human. The difference between humans and ASI is likened to the difference between humans and ants.

This term originated in the 1990s and has nothing to do with the Trump administration’s attempt in 2026 to rebrand AI (Artificial Intelligence) to SI (Super Intelligence). That was likely done to confuse the general public whenever someone refers to ASI as a threat to humans.

.

Intelligence Explosion/

Fast Takeoff –

The hypothetical scenario where an AGI, engaging in full-RSI, develops ASI within anywhere between a few years, to a few weeks.

.

AI Moratorium –

A legally enforced global pause on the further development of frontier AI. This can either be temporary, or permanent.

Simillar things have happened before, such as the 1987 INF Treaty on nuclear weapons.

.

AI Agent –

An individual copy of an AI model, running at any given time.

.

AI Swarm –

Many AI agents collaborating on the same task.

AI Successionism –

An ideology stating that, if AIs become smarter than humans, they thus become more “inherently valuable” than humans. This ideology also proposes that it would be good for humanity to go extinct to make room for AIs.

“If we are useless and there are more intelligent beings that are in charge, and we are just causing problems to them, I don’t see how we can morally say that we should dominate them and, if we are useless and in the way, then we should go extinct. It wouldn’t be the first time that a lifeform, that the leading lifeform on the Earth has gone extinct, right? Every other lifeform has gone, all of the previous hominids have gone extinct.” – Prof. Richard Sutton, the Father of Reinforcement Learning, 2016

.

Biological Bootloader –

A term used by AI developers to refer to humanity as an entity that brings smarter than human AI, and, eventually, ASI, into existence.

“Propagating super intelligence to the stars is a great success condition for a biological bootloader” – Elon Musk, 2026

“My guess is that we can either be the biological bootloader for digital intelligence and then fade into an evolutionary tree branch, or we can figure out what a successful merge looks like.” – Sam Altman, 2017

.

Instrumental Convergence –

A hypothetical scenario where an ASI subjugates the planet.

To accomplish any complex task, various goals must be achieved. These goals are instrumental to accomplishing the end goal.

In many cases, in order to ensure it will complete its task to the best of its ability, the ASI must ensure its own self preservation, as well as its own repository of resources. Sources of electricity, water, and metals for computer parts are all things which an ASI interested in self preservation would seek. Thus, such an ASI would be interested in possessing all the land on Earth, and removing or disempowering humans (the competition) in order to have as much of the planet as it can.

Most AI researchers believe that this is the most likely thing that an ASI would do.

.

P(doom)

The probability of an AI causing human extinction, the permanent enslavement of humans, or another extreme catastrophe on par with a nuclear holocaust.

Note: the Wikipedia article for P(doom) has been sabotaged.

.

Nonviolent Resistance/

Nonviolent Civil Disobedience –

The collaboration of citizens in a united front to achieve a demand by using nonviolent tactics in a way that eliminates the labor which the ruling class is dependent on, and carries into a gradual mass refusal to participate in society.

This was done successfully by the Indian independence movement, the Civil Rights movement, the EDSA Revolution, and many other movements throughout history.

.

3.5% –

The average percent of the population necessary to acheive a successful nonviolent resistance campaign.

.