The Machine Question is an independent research project exploring artificial intelligence through philosophy, technology and human cognition. Essays on intelligence, creativity and the evolving relationship between humans and machines.
Intelligence
What makes a system intelligent?
Artificial intelligence does more than extend the range of things machines can do. It forces us to reconsider the criteria by which intelligence itself is recognised, measured and distinguished from computation.
The Question
Intelligence was never as easy to define as it was to recognise.
For most of human history, the problem could remain partially hidden. Intelligence was associated with beings that reasoned, learned, spoke, adapted and acted within the world. Artificial intelligence separates those capabilities from the organism that once appeared to unite them.
?
A category under pressure
We commonly recognise intelligence through its manifestations: the ability to solve problems, acquire knowledge, adapt to new conditions, use language, infer relationships or produce appropriate responses in unfamiliar situations. Yet none of these abilities, taken individually, provides an uncontested definition of intelligence.
That ambiguity mattered less while sophisticated cognitive performance appeared inseparable from biological minds. Machines change the structure of the problem. Systems can now perform tasks once regarded as evidence of intelligence without necessarily possessing the other properties that humans have historically associated with an intelligent mind.
The resulting difficulty is not simply whether machines have crossed some invisible threshold. It is that the threshold itself becomes difficult to locate.
The Conceptual Turn
The emergence of machine intelligence therefore creates a reversal. Instead of asking only whether machines satisfy our definition of intelligence, we must ask whether our definition was ever sufficiently precise.
Three Distinctions
The question changes depending on what we believe intelligence requires.
Performance
Is intelligence defined by what a system can successfully do, regardless of the mechanism by which the result is produced?
Process
Does intelligence require particular forms of reasoning, representation or internal organisation rather than successful behaviour alone?
Experience
Must an intelligent system understand, experience or possess awareness of what it is doing, or are these separate questions?
The Inquiry
The Machine Question begins from a different premise: machine intelligence may tell us as much about the instability of our categories as it does about the capabilities of machines.
What We Examine
Intelligence is not a single capability.
The field is approached through a set of related problems. Each isolates a different dimension of intelligence and tests whether familiar human categories remain adequate when similar capabilities appear in machines.
From answer to inference.
What distinguishes a system that produces a correct answer from one that can genuinely be said to reason? The inquiry examines inference, problem-solving, abstraction and the difference between successful output and an intelligible process of arriving at it.
Adaptation without experience.
Machine learning transforms performance through exposure to data, optimisation and feedback. The question is how this form of adaptation relates to human learning, and whether the shared vocabulary conceals fundamentally different processes.
Performance and meaning.
A system may manipulate language, solve problems and respond appropriately without resolving the philosophical question of whether it understands what it is doing. This field examines the boundary between functional competence and semantic understanding.
Beyond the task.
Intelligence has often been associated with the ability to transfer knowledge across situations, adapt to novelty and operate beyond narrowly defined tasks. Machine systems make it possible to ask which forms of generalisation matter and how they should be recognised.
How intelligence becomes visible.
Every claim about intelligence depends on a test, benchmark, observation or criterion. The inquiry therefore examines how intelligence is measured, what those measurements capture and which assumptions are embedded in the tests themselves.
Does intelligence require a world?
Human intelligence develops through bodies, environments, social relationships and lived experience. Machine intelligence raises the question of whether intelligence can be separated from embodiment, or whether some forms of understanding depend upon being situated in a world.
The Central Distinction
The Machine Question does not assume that human and machine intelligence must be identical in order to be comparable. The inquiry asks instead which properties are essential to the category — and which may have been mistaken for essential only because human intelligence was once the only example available.
One field, several levels of explanation.
These questions cannot be settled at the level of technical performance alone. They involve cognitive science, philosophy, language, measurement, computation and the historical concepts through which intelligence has been recognised.
The purpose of the field is therefore not to produce a single definition as quickly as possible, but to identify which distinction is doing the work whenever a system is described as intelligent.
Key Questions
Questions that remain open.
The field of intelligence is defined less by a settled answer than by a set of distinctions that remain unresolved. These questions organise the inquiry and provide recurring points of return across essays, concepts and research notes.
Is successful performance sufficient for intelligence?
If a system consistently solves problems, adapts to new tasks and produces appropriate responses, what additional property would be required before we describe it as intelligent?
Can reasoning exist without understanding?
Machine systems can produce sequences that resemble inference. The unresolved question is whether reasoning should be defined by the structure of the process, the quality of the result, or the presence of understanding.
Does intelligence require embodiment?
Human cognition develops through bodies, environments and social interaction. Machine intelligence tests whether those conditions are essential to intelligence or specific to one form of it.
Is intelligence one thing or a family of capabilities?
The category may conceal multiple properties — reasoning, learning, adaptation, abstraction, memory, language and planning — that need not emerge together in every system.
Can intelligence be measured independently of the tasks we choose?
Benchmarks make intelligence visible through selected forms of performance. Every measure therefore raises a second question: whether the test captures intelligence or merely the ability to succeed under its own assumptions.
Can fundamentally different systems be intelligent in fundamentally different ways?
If intelligence can emerge from architectures unlike the human brain, comparison may require abandoning the assumption that machine intelligence must resemble human intelligence in order to count as intelligence at all.
Open by Design
These questions are not placeholders awaiting a final answer. They define a research field in which new capabilities may alter the meaning of the terms used to describe them.
Essays
Arguments developed through the field.
Essays develop individual questions beyond the field map, connecting evidence, concepts and interpretation into a sustained argument. This section grows as the Intelligence corpus develops.
Featured Essay
Why Artificial Intelligence Changes the Meaning of Intelligence
What happens to the concept of intelligence when capabilities once used to identify it can also be reproduced by machines?
The essay examines a reversal produced by artificial intelligence: instead of asking only whether machines can become intelligent, their capabilities force us to reconsider the definition against which the comparison is being made. Reasoning, language, generalisation and understanding become not simply machine capabilities to be measured, but categories whose boundaries must themselves be examined.
Read the essay
Developing Research
Future essays will extend this field into questions that cannot be resolved within a single text. These are active lines of inquiry rather than placeholder publications.
Reasoning without understanding
An investigation into whether successful inference, problem-solving and explanation require semantic understanding, or whether these capacities can meaningfully diverge.
The measurement of machine intelligence
A study of benchmarks, tests and criteria used to make intelligence visible — and of the assumptions embedded in the act of measuring it.
The Wider Corpus
Intelligence also intersects with essays classified under Language, Knowledge, Creativity, Power and Humanity.
Explore all essays
Concepts & Frameworks
The ideas and lenses behind the field.
Intelligence is examined through concepts that define the problem and frameworks that determine the level from which the problem is viewed. The distinction matters: concepts identify what is at stake; frameworks organise how it is interpreted.
Concepts
What must be defined.
Concepts isolate the categories whose meaning becomes unstable when capabilities associated with intelligence appear in non-human systems.
01
Reasoning
Inference, problem-solving and the distinction between producing an answer and arriving at one.
02
Understanding
The relationship between functional competence, semantic content and genuine comprehension.
03
Knowledge
Whether information processing, representation and justified belief belong to the same category.
04
Agency
Whether intelligent behaviour requires intention, goal formation or independent capacity to act.
Frameworks
How the problem is examined.
Frameworks provide complementary levels of explanation. No single one is assumed to account for intelligence in all of its dimensions.
01
Cognitive
Examines reasoning, learning, memory and behaviour in relation to human and machine cognition.
02
Language
Tests how linguistic performance relates to representation, meaning and understanding.
03
Human
Asks which properties used to define intelligence depend specifically on human embodiment and experience.
04
Institutional
Examines how definitions of intelligence become embedded in testing, governance and decision systems.
Research Architecture
These are selected connections rather than an exhaustive map. The wider Concepts and Frameworks sections allow the Intelligence field to be read in relation to the full research architecture of The Machine Question.