For sale: what-is.si — this domain is available for purchase via Sedo.com

Make an offer ↗
A field guide · No. 01

What is SI?

SI stands for Superintelligence — an intellect that would far surpass the best human minds in practically every field. Here is what that means, how it differs from today's AI, and why people take it seriously.

su·per·in·tel·li·gence  /ˌsuːpərɪnˈtɛlɪdʒəns/  noun
§ 1 — Definitions

Two letters, two very different ideas.

AI

Artificial Intelligence

Computer systems that perform tasks which normally require human intelligence — recognising speech, translating languages, spotting patterns, writing, planning and making decisions.

Almost all AI in use today is narrow: it can be extraordinarily good at specific things, like playing chess, reading X-rays or generating text, while lacking the broad, flexible understanding a person brings to everyday life.

Modern AI is mostly built with machine learning. Instead of being programmed rule by rule, systems learn from enormous amounts of data and improve through training.

  • Voice assistants, translation and search
  • Recommendation feeds and spam filters
  • Large language models and image generators
  • Medical imaging, fraud detection, self-driving research

SI

Superintelligence

A hypothetical intelligence that greatly exceeds the cognitive performance of humans in virtually all domains of interest — science, strategy, creativity, social skill and general wisdom.

Superintelligence does not exist yet. It is a concept researchers use to think about where AI could ultimately lead: not a better tool for one task, but a mind that outperforms the whole of humanity's best experts at almost everything.

It could be reached gradually, or very quickly if an advanced system becomes capable of improving its own design — an idea often called an "intelligence explosion".

  • Speed — thinking like a human, but millions of times faster
  • Collective — vast numbers of minds working in concert
  • Quality — reasoning that is fundamentally deeper than ours
§ 2 — The ladder

From AI to SI, in three steps.

People usually describe the path as a sequence of capability levels. The middle rung — general intelligence — is the bridge between the tools we use today and a true superintelligence.

Level 1 · ANI

Narrow AI

Artificial Narrow Intelligence

Excellent at a specific task or a family of tasks, but brittle outside them. A chess engine cannot drive a car.

Status: here today, everywhere

Level 2 · AGI

General AI

Artificial General Intelligence

A system that can learn and perform any intellectual task a human can, transferring knowledge flexibly between domains.

Status: actively pursued, timelines debated

Level 3 · ASI / SI

Superintelligence

Artificial Superintelligence

An intellect that vastly outperforms the best humans in nearly every field — from scientific discovery to long-term planning.

Its arrival would likely be the most consequential event in human history.

Status: hypothetical

§ 3 — Key ideas

Six basic statements about Superintelligence.

Intelligence is not magic — it is a capability.

If intelligence comes from information processing, there is no known law that says human brains are the upper limit. Machines may one day go far beyond it.

Today's AI is not superintelligent.

Current systems are impressive and improving fast, but they still make basic mistakes, lack reliable judgement and depend on human direction.

The step after AGI could be short.

A system that matches human researchers could help build its own successor, compressing decades of progress into far less time.

Its potential upside is enormous.

Superintelligence could accelerate cures for disease, clean energy, scientific breakthroughs and abundance on a scale that is hard to imagine.

Alignment is the central challenge.

A mind more capable than ours must reliably pursue goals compatible with human values. Getting this right before SI exists is the work of AI safety research.

It is a question for everyone.

How SI is built, governed and shared is not only a technical matter — it is a choice for societies, governments and every person it will affect.

§ 4 — Questions

Frequently asked.

What is the difference between AI and SI?

AI is the broad field of building machines that perform intelligent tasks — and it already exists all around us. SI, or Superintelligence, is a specific, hypothetical outcome of that field: a system smarter than the brightest humans at nearly everything.

Is ChatGPT (or any current chatbot) a superintelligence?

No. Large language models can write, summarise and reason across many topics, and sometimes outperform people on particular tests. But they remain inconsistent, can be confidently wrong, and do not surpass expert humans across the board.

When might Superintelligence arrive?

Nobody knows. Estimates from researchers range from within a decade to many decades — or never. Rapid progress in recent years has caused many experts to shorten their forecasts, but significant uncertainty remains.

Is Superintelligence dangerous?

It could be. The concern is not that machines become "evil", but that a highly capable system pursuing poorly specified goals could cause serious harm, and be difficult to correct. That is why alignment and governance are taken seriously by labs, researchers and policymakers.

Who coined the term "superintelligence"?

The idea goes back to mathematician I. J. Good, who in 1965 described an "ultraintelligent machine" and a possible "intelligence explosion". Philosopher Nick Bostrom popularised the term with his 2014 book Superintelligence: Paths, Dangers, Strategies.

Why the domain "what-is.si"?

.si is the country-code domain of Slovenia — and a perfect fit for the abbreviation of Superintelligence. "what-is.si" reads naturally as the question millions of people are starting to ask. The domain is for sale.

§ 5 — Glossary

A small dictionary of AI & SI.

The words you keep hearing, from agent to zero-shot — each in a sentence or two.

A

what-is Agent:
An AI system that takes actions toward a goal on its own — browsing, writing code, using tools — rather than only answering one question at a time.
what-is AGI:
Artificial General Intelligence: a system able to learn and perform any intellectual task a human can, moving flexibly between domains.
what-is AI:
Artificial Intelligence: computer systems that do tasks normally requiring human intelligence, such as understanding language, recognising images or making decisions.
what-is AI safety:
The research field focused on making AI systems reliable, controllable and beneficial, and on preventing accidents or misuse.
what-is Algorithm:
A precise, step-by-step set of instructions a computer follows to solve a problem or complete a task.
what-is Alignment:
The challenge of making an AI system's goals and behaviour match what its designers and humanity actually intend.
what-is ANI:
Artificial Narrow Intelligence: AI that excels at one specific task or family of tasks but cannot generalise beyond them. All AI today is narrow.
what-is ASI:
Artificial Superintelligence: an intellect that vastly outperforms the best human minds in nearly every field. Often shortened to SI.
what-is Attention:
A mechanism that lets a model weigh which parts of its input matter most for each step — the key idea behind transformers.

B

what-is Backpropagation:
The method neural networks use to learn: errors are passed backwards through the network to work out how each weight should change.
what-is Benchmark:
A standardised test used to measure and compare the abilities of AI systems.
what-is Bias:
Systematic unfairness in an AI system's outputs, usually inherited from skewed or unrepresentative training data.

C

what-is Chain of thought:
Having a model reason step by step before giving its final answer, which often improves accuracy on hard problems.
what-is Chatbot:
A program that holds a conversation with people in natural language, today usually powered by a large language model.
what-is Compute:
The raw processing power — chips, time and energy — used to train and run AI models. A key driver of AI progress.
what-is Computer vision:
The branch of AI that teaches machines to interpret images and video: recognising faces, objects, scenes and motion.
what-is Context window:
The amount of text (measured in tokens) a language model can take into account at once.
what-is Corrigibility:
The property of an AI system that allows humans to correct, adjust or shut it down without it resisting.

D

what-is Dataset:
A structured collection of examples — text, images, numbers — used to train or evaluate a model.
what-is Deep learning:
Machine learning with neural networks of many layers, able to learn complex patterns directly from raw data.
what-is Diffusion model:
A generative model that creates images, audio or video by gradually turning random noise into a coherent result.

E

what-is Embedding:
A list of numbers that represents the meaning of a word, image or document, so that similar things sit close together.
what-is Emergent ability:
A skill that appears in larger models but is absent in smaller ones, without being explicitly trained for.
what-is Evals:
Structured evaluations that probe what a model can do — including dangerous capabilities — before and after release.
what-is Existential risk:
A risk that could cause human extinction or permanently curtail humanity's future. Often shortened to x-risk.

F

what-is Few-shot learning:
Getting a model to perform a new task by showing it only a handful of examples in the prompt.
what-is Fine-tuning:
Further training a pre-trained model on a smaller, specific dataset to adapt it to a particular task or style.
what-is Foundation model:
A large model trained on broad data that can be adapted to many downstream tasks.
what-is Frontier model:
One of the most capable AI models in existence at a given time, at the leading edge of the field.

G

what-is Generative AI:
AI that creates new content — text, images, music, code or video — rather than only classifying or predicting.
what-is GPT:
Generative Pre-trained Transformer: a family of language models, and the architecture pattern behind many modern chatbots.
what-is GPU:
Graphics Processing Unit: a chip built for massively parallel calculation, now the workhorse of AI training.
what-is Gradient descent:
The optimisation method that nudges a model's parameters step by step in the direction that reduces its error.

H

what-is Hallucination:
When an AI model produces confident-sounding information that is false or made up.
what-is Hard takeoff:
A scenario in which AI goes from roughly human-level to superintelligent very quickly — in days, weeks or months.

I

what-is Inference:
Running a trained model to produce an output, as opposed to training it.
what-is Instrumental convergence:
The idea that almost any goal leads an agent to pursue the same sub-goals, such as acquiring resources and avoiding being switched off.
what-is Intelligence explosion:
A runaway cycle in which an AI improves its own design, each version building a smarter successor faster than the last.
what-is Interpretability:
Research into understanding what is actually happening inside a model — which features it represents and how it reaches decisions.

J

what-is Jailbreak:
A prompt crafted to trick a model into ignoring its safety rules or producing content it was designed to refuse.

L

what-is Large language model:
A neural network trained on vast amounts of text to predict and generate language. Abbreviated LLM.

M

what-is Machine learning:
An approach to AI in which systems learn patterns from data instead of being programmed with explicit rules.
what-is Model:
The trained mathematical system that turns inputs into outputs — the 'brain' that results from training.
what-is Multimodal:
Able to understand or generate several kinds of data, such as text, images, audio and video, in one model.

N

what-is Natural language processing:
The field of AI concerned with understanding and producing human language. Abbreviated NLP.
what-is Neural network:
A computing system loosely inspired by the brain: layers of connected units whose connection strengths are learned from data.

O

what-is Open weights:
A model whose trained parameters are publicly released, so anyone can run, study or modify it.
what-is Orthogonality thesis:
The idea that intelligence and goals are independent: a highly intelligent system could pursue almost any objective.
what-is Overfitting:
When a model memorises its training data so closely that it performs poorly on new, unseen examples.

P

what-is Paperclip maximizer:
A thought experiment: an AI told only to make paperclips could convert everything, including us, into paperclips. It illustrates misaligned goals.
what-is Parameter:
One of the adjustable numbers inside a model that is tuned during training. Large models have billions of them.
what-is Pre-training:
The first, large-scale phase of training in which a model learns general patterns from a huge dataset.
what-is Prompt:
The instruction or input text given to an AI model to tell it what to do.

R

what-is Recursive self-improvement:
An AI repeatedly improving its own capabilities, with each improvement making the next one easier.
what-is Red teaming:
Deliberately attacking or stress-testing an AI system to find its flaws, weaknesses and dangerous behaviours.
what-is Reinforcement learning:
Learning by trial and error: an agent takes actions and is rewarded or penalised according to the results.
what-is Reward hacking:
When an AI finds an unintended shortcut to maximise its reward without actually doing what was wanted.
what-is RLHF:
Reinforcement Learning from Human Feedback: training a model using people's ratings of its answers to make it more helpful and safe.

S

what-is Scaling laws:
Observed patterns showing that model performance improves predictably with more data, parameters and compute.
what-is Singularity:
A hypothetical point at which technological growth, driven by superintelligence, becomes so fast that the future is impossible to predict.
what-is Soft takeoff:
A scenario in which the transition to superintelligence unfolds gradually over years or decades, leaving time to adapt.
what-is Superintelligence:
An intellect that greatly exceeds human cognitive performance in virtually all domains. The SI in what-is.si.
what-is Supervised learning:
Training a model on labelled examples, where each input comes with the correct answer.
what-is Synthetic data:
Training data generated by AI or simulation rather than collected from the real world.

T

what-is Token:
A chunk of text — a word, part of a word or a symbol — that a language model reads and writes.
what-is Training:
The process of feeding data to a model and adjusting its parameters so it gets better at a task.
what-is Transformer:
The neural network architecture, introduced in 2017, that underlies most modern language and multimodal models.
what-is Turing test:
A test proposed by Alan Turing in 1950: can a machine converse so well that a person cannot tell it apart from a human?

U

what-is Unsupervised learning:
Training a model to find structure and patterns in data that has no labels.

W

what-is Weights:
The learned numerical values inside a neural network that determine how it transforms inputs into outputs.

X

what-is X-risk:
Short for existential risk: a threat to humanity's survival or long-term potential.

Z

what-is Zero-shot learning:
A model performing a task it was never shown an example of, using only an instruction.
§ 6 — Notice

This domain is for sale.

Own the question everyone is about to ask. what-is.si is a short, memorable, brandable name for an AI publication, research lab, educational project or product in the Superintelligence space.

The sale is handled securely through Sedo, the world's leading domain marketplace.

Available domain

what-is.si

Marketplace
Sedo.com
Transfer
Secure escrow & transfer via Sedo
Ideal for
AI media, education, research, startups
Buy or make an offer on Sedo Opens sedo.com in a new tab.