How intelligent is AI compared to life — and could it be conscious?
Explore the scientific landscape of intelligence and consciousness, from microscopic organisms to the world's most advanced artificial intelligence.
In this atlas, AI refers primarily to contemporary large language models, multimodal foundation models, and reasoning systems. Their abilities are assessed through documented performance and behavioural evaluations, not assumed subjective experience.
Two different questions, kept apart
Intelligence is not one number
We track ten dimensions, from learning to autonomy. A bee, a crow and a language model each show a different profile, and collapsing them into an IQ-style score would hide what matters.
Consciousness is not measured here
No accepted method measures consciousness across species and machines. The atlas grades the strength of evidence, and keeps consciousness (experience) separate from self-consciousness (a model of oneself).
Absence of evidence is labelled
Where a capacity has not been tested, the atlas says "insufficient evidence". It does not fill gaps with guesses, and every grade links to its sources.
Where AI sits among living things
Swipe the chart sideways to see the full scale.
AI vs organism
This comparison answers three separate questions. A strong answer to one says nothing about the others.
Intelligence: what can each one do?
Measured through behaviour: experiments for animals, benchmarks and evaluations for AI. Bars show how strong the evidence is that each capacity is present.
Evidence profile
Each spoke is one dimension; the further out, the stronger the evidence. The two shapes show different kinds of mind, not a ranking.
Verified benchmark results
Benchmarks have no organism equivalent. They place a model among AI systems and, where available, against human performance.
Consciousness: is there experience?
No instrument measures experience directly. We compare the strength of evidence for indicators such as perceptual awareness and the capacity to feel pain or pleasure.
Self-consciousness: does it know itself?
Self-recognition and monitoring of one's own knowledge. These can be shown in functional form, which does not by itself mean they are experienced.
Intelligence, consciousness and self-consciousness, explained
Three different questions
What a system can do, whether it experiences anything, and whether it knows itself. Each needs its own kind of evidence.
The three can come apart
Why a single intelligence score would mislead
Each spoke is one dimension; distance from the centre is the strength of evidence. Choose up to five profiles: similar totals can hide completely different shapes.
How scientists test intelligence
Five classic methods, from benchmarks for AI to memory tests for birds. For each method: why it is done, what is measured and what was found, then a figure of how it works.
The parts of consciousness
Consciousness and self-consciousness overlap but are not the same. Select any part to see what it means and how it is studied.
How scientists look for consciousness
Six methods. For each method: why it is done, what is measured and what was found, then a figure of how it works.
Competing theories, compared
None is established. A preregistered 2025 test of the two leading theories supported parts of both and challenged each.
Where the distinction changes decisions
How every grade was made, and on what sources
Evidence grades
Grades summarise how much published evidence currently bears on whether a capacity is present. They describe the state of the literature, not the value of any study, and they are not probabilities of consciousness.