ONLINEAGENT_OPS 2026.Q3 HOME HUBSMODELSBUILDINGCOMPANIESFILMMAKINGSECURITYLIVING WITH ITSKILLSALL HUBS, ONE TERMINAL → ARTICLES CRAFT RECORD BLOG TOOLS SKILLS SEARCH /
HOME›ARTICLES›DOES AI UNDERSTAND ANYTHING?
ARTICLES · EVERGREEN EXPLAINERBEGINNER

Does AI Understand Anything?

Both confident answers are wrong. What the research actually shows, explained plainly, with sources named and dated.

READ8 min
WORDS1,688
SECTIONS5
SOURCES6
TYPEEXPLAINER
CHECKED16 SEP 26

Partly, and the question is not settled. Researchers can find internal features that match real concepts and change behaviour when adjusted, so “just autocomplete” undersells it.1 But Anthropic found Claude describing one method for a sum while using another, so its account of its own reasoning is unreliable.2

TL;DR — THE SHORT VERSION
  • "It's just autocomplete" undersells it. Researchers can find internal features that match real concepts and change behaviour when adjusted.
  • "It understands, maybe it's conscious" oversells it. Fluent talk about inner states is not evidence of them, and a 2023 assessment concluded that no current AI systems are conscious.
  • Its account of its own reasoning is unreliable. Anthropic found Claude describing school arithmetic while doing something different inside.
  • It can misjudge what it knows. A familiar-sounding name can switch off its "don't know" response — one route to a confident wrong answer.
  • For practical use, the question barely matters. Accuracy can be checked; understanding cannot. Check the output either way.
THE DISMISSIVE ANSWER

"It's just autocomplete"

  • Technically accurate about the mechanism: the system predicts the next piece of text, repeatedly.
  • Whether it remembers you between conversations now depends on the product — as of August 2026, memory is on by default for Claude's Free, Pro and Max plans — but it has no goals of its own and no stake in being right.3First-hand: until 22 Sep 2026 this line said Anthropic “switched memory on by default” in August 2026. The 25 Aug 2026 post says memory is on by default; it does not say that was when the default changed.
  • It cannot reliably tell recall from invention, which is why it produces confident falsehoods in the same register as facts. The mechanism is on why AI makes things up.
Where it fails: "just" is doing enormous unearned work. Predicting text well enough, at sufficient scale, required building internal structure — and interpretability research keeps finding it.
THE CREDULOUS ANSWER

"It understands, it might be conscious"

  • Points at real behaviour: these systems handle novel problems, transfer reasoning between domains, and describe their own steps in ways pure lookup cannot — though those descriptions are not always what happened inside.2
  • Interpretability work has identified internal features corresponding to recognisable concepts — structure, not just statistics.
Where it fails: fluency about inner states is not evidence of inner states. A system trained on the entirety of human writing about feelings will produce that register perfectly whether or not anything is happening — and there is no settled test. One published attempt, by Butlin and colleagues, derived “indicator properties” of consciousness from scientific theories and concluded: “Our analysis suggests that no current AI systems are conscious, but also suggests that there are no obvious technical barriers to building AI systems which satisfy these indicators.”4

What the evidence actually shows

ESTABLISHED 01

There is internal structure, and it is findable

Interpretability research can locate features inside these networks that correspond to identifiable concepts, and in some cases steer behaviour by manipulating them. That is a stronger claim than "statistical pattern matching" allows, and it is measured rather than argued.1

CONTESTED 02

Whether new abilities "emerge" suddenly is disputed

A 2022 study argued that some abilities appear only in larger models, so they “cannot be predicted simply by extrapolating the performance of smaller models”. A 2023 study replied that such jumps “appear due to the researcher's choice of metric rather than due to fundamental changes in model behavior with scale”. That disagreement should temper confidence in both directions: you cannot claim to know a system is "merely" predicting when researchers still dispute how its abilities grow.5

ESTABLISHED 03

Its sense of what it knows can misfire

Anthropic's interpretability team found something like an internal "don't know" default that recognising a familiar name switches off — and found it switching off wrongly. In their words, “such misfires can occur when Claude recognizes a name but doesn't know anything else about that person.” That is one concrete route to a fluent, confident, false answer, and a genuine limitation whatever one concludes about understanding.6

ESTABLISHED 04

The word "understand" was never precise

Much of this argument is definitional. Does a person who can use a word correctly in every context but cannot define it understand it? Does a chess engine understand chess? These questions were unresolved before AI existed, and pointing a new system at them did not resolve them — it just made the vagueness expensive.

TAKEAWAY

There is real structure inside these systems, but their accounts of their own reasoning, and their sense of what they know, are both unreliable. Neither "just autocomplete" nor "it understands" survives the evidence.

◈ THE HONEST POSITION

Something is happening inside these systems that is more structured than "autocomplete" suggests and less settled than "understanding" implies. Anyone who tells you confidently which it is — in either direction — is describing their intuitions rather than the evidence.

THREE ANSWERS · ONE SPECTRUM
Both confident answers fail on the evidence. The defensible one sits between them.
JUST AUTOCOMPLETEUndersells it: interpretability research keeps finding internal structure.
THE HONEST POSITIONMore structured than autocomplete suggests, less settled than understanding implies.
IT UNDERSTANDSOversells it: fluent talk about inner states is not evidence of them.
Reasoning — summarises this page’s two failed answers and its honest-position box. Page checked 16 Sep 2026.

Is AI conscious or sentient?

The plain point: the researchers who have looked most carefully say today’s AI is probably not conscious, but no agreed test can settle it, and they disagree about the future (sources below).

What the words mean. “Conscious” and “sentient” both point at subjective experience: whether there is something it feels like to be the system. The philosopher David Chalmers uses the two words as roughly equivalent, while warning that “sentient” is the vaguer of the two.7 That is a different question from whether AI understands: a system could understand a great deal and feel nothing.

What researchers say about today’s systems. The 2023 Butlin study checked AI systems against “indicator properties” drawn from scientific theories of consciousness and found none that qualified.8 Chalmers reached a similar view of large language models, calling it “somewhat unlikely” that current ones are conscious.7

Why its own word does not settle it. Asking the AI is no help, because it was trained on huge amounts of human writing about feelings. Chalmers notes that a slight change in wording got one model to say it was, and was not, sentient.7 The evidence section above shows the same problem with self-reports of reasoning.

Where they disagree. On the future, views split. The Butlin study found “no obvious technical barriers” to building systems that meet its indicators, and Chalmers says possible future consciousness should be taken seriously (both quoted above). Anthropic, which makes Claude, says there is no scientific consensus and has started research on whether models might deserve moral consideration.9 Mustafa Suleyman, who leads work on Copilot at Microsoft AI, argues the bigger near-term risk is people believing AI is conscious when there is no evidence that it is.10

TAKEAWAY

Today’s AI is probably not conscious or sentient, according to the most careful published assessments, but no test can settle it, and a chatbot saying it has feelings is not evidence that it does.

Why it matters less than it feels like it should

For nearly every practical purpose, the question is irrelevant. Whether a system understands has no bearing on whether its output is accurate, and accuracy is verifiable while understanding is not. A doctor does not need to know whether a test comprehends anything; they need to know its false-positive rate.

Where it does matter is narrower and worth naming: questions of moral status, if they ever become live, and how much trust these systems are given — because "it understands" is used as an argument for reducing oversight, and it is not strong enough to carry that weight.

TAKEAWAY

Do not let "it understands" stand in for checking. Judge the output, which you can verify, not the system's apparent grasp of it, which you cannot.

◈ A NOTE ON WHY PEOPLE FEEL STRONGLY

The dismissive answer protects something: if it is only autocomplete, then human thought remains categorically different and nothing needs re-examining. The credulous answer offers something else: significance, and the sense of standing near something historic. Both are emotionally load-bearing, which is why the argument generates more heat than the evidence supports. Noticing which one you want to be true is most of the work.

◈ WHERE THIS SITE STANDS

This site does not know, and says so. What it holds to is the practical consequence, which is unaffected either way: these systems produce confident output regardless of accuracy, so verification is not optional. That is true if they understand perfectly and true if they understand nothing. Everything else here is philosophy — worth having, and not a reason to skip the checking.

Questions people ask

Does AI really understand what it says?

Partly, depending on what you mean by “understand”. Researchers can find internal features that match real concepts and change behaviour, but its explanations of its own reasoning are not reliable; see the evidence section above.

Is ChatGPT conscious?

The most careful published assessments say current AI systems, chatbots included, are probably not conscious, though none can prove it.8

Can AI feel emotions?

There is no evidence that it does. A chatbot can describe feelings fluently because it learned from human writing, and Chalmers points out how easily those self-reports flip with a change of wording.7

Could AI become conscious in the future?

Some serious researchers think it is possible, and Anthropic says there is no scientific consensus either way.9 Chalmers says the possibility for successors to today’s models should be taken seriously; Suleyman argues companies should avoid building AI that seems conscious at all.10

Related on this site: Turing test meaning

SOURCES
  1. 1Anthropic, "Mapping the Mind of a Large Language Model", May 2024 — read at source 11 Sep 2026anthropic.com, "We successfully extracted millions of features from the middle layer of Claude 3.0 Sonnet", and manipulating them "causes corresponding changes to behavior… they aren’t just correlated with the presence of concepts in input text, but also causally shape the model’s behavior." ↑aanthropic.com, "We successfully extracted millions of features from the middle layer of Claude 3.0 Sonnet", and manipulating them "causes corresponding changes to behavior… they aren’t just correlated with the presence of concepts in input text, but also causally shape the model’s behavior." Source added 11 Sep 2026; this claim previously pointed only at another page on this site. Also covered as one of the four evaluation layers on how AI is tested. ↑b
  2. 2Anthropic, Tracing the thoughts of a large language model, 27 Mar 2025, read at source 16 Sep 2026“Strikingly, Claude seems to be unaware of the sophisticated "mental math" strategies that it learned during training.” Asked how it added 36 and 59, “it describes the standard algorithm involving carrying the 1.” ↑a“Strikingly, Claude seems to be unaware of the sophisticated "mental math" strategies that it learned during training.” Asked how it added 36 and 59, “it describes the standard algorithm involving carrying the 1.” ↑b
  3. 3Anthropic, Claude's memory works everywhere, and you decide what's in it, 25 Aug 2026, read at source 16 Sep 2026“Memory is on by default on Free, Pro and Max plans across web, desktop, and mobile.” An earlier version of this line said these systems have no persistent memory by default. ↑
  4. 4Butlin, Long, Elmoznino, Bengio, Birch et al., Consciousness in Artificial Intelligence: Insights from the Science of Consciousness, Aug 2023, read at source 16 Sep 2026.↑
  5. 5Wei et al., Emergent Abilities of Large Language Models, 2022; Schaeffer, Miranda & Koyejo, Are Emergent Abilities of Large Language Models a Mirage?, 2023; both read at source 16 Sep 2026. An earlier version of this section said "Nobody knows why capabilities appear when they do" and that this "is stated openly by the people building these systems", with no source.↑
  6. 6Anthropic, Tracing the thoughts of a large language model, 27 Mar 2025, read at source 16 Sep 2026. An earlier version of this section said "There is no internal flag for 'I know this'"; this research describes something like one, and how it fails.↑
  7. 7David J. Chalmers, Could a Large Language Model be Conscious?, Mar 2023 (revised Aug 2024), read at source 5 Oct 2026“As I use the terms, consciousness and sentience are roughly equivalent. Consciousness and sentience, as I understand them, are subjective experience.” and “The word sentience is even more ambiguous and confusing than the word consciousness.” ↑a“I conclude that while it is somewhat unlikely that current large language models are conscious, we should take seriously the possibility that successors to large language models may be conscious in the not-too-distant future.” ↑b“When reports of consciousness are as fragile as this, the evidence for consciousness is not compelling.” ↑c“When reports of consciousness are as fragile as this, the evidence for consciousness is not compelling.” ↑d
  8. 8Butlin, Long, Elmoznino, Bengio, Birch et al., Consciousness in Artificial Intelligence: Insights from the Science of Consciousness, Aug 2023, read at source 5 Oct 2026“From these theories we derive ‘indicator properties’ of consciousness, elucidated in computational terms that allow us to assess AI systems for these properties.” and “Our analysis suggests that no current AI systems are conscious”. ↑a“Our analysis suggests that no current AI systems are conscious, but also suggests that there are no obvious technical barriers to building AI systems which satisfy these indicators.” ↑b
  9. 9Anthropic, Exploring model welfare, 24 Apr 2025, read at source 5 Oct 2026“There’s no scientific consensus on whether current or future AI systems could be conscious, or could have experiences that deserve consideration.” and “For now, we remain deeply uncertain about many of the questions that are relevant to model welfare.” ↑a“There’s no scientific consensus on whether current or future AI systems could be conscious”. ↑b
  10. 10Mustafa Suleyman, We must build AI for people; not to be a person, 19 Aug 2025, read at source 5 Oct 2026on AI being conscious, “there is zero evidence of this today”; “Today at Microsoft AI we build AI to empower people, and I’m focused on making products like Copilot responsible technologies”; and “my central worry is that many people will start to believe in the illusion of AIs as conscious entities”. ↑a“We must build AI for people; not to be a digital person”. ↑b

Free printable AI cheat sheets, and AI news every week.

Sixteen one-page AI guides and checklists in one free PDF, no sign-up. And if you want to keep up, the newsletter covers what changed in AI each week, in plain English, with the sources.

SUBSCRIBE FREE ↗ GET THE CHEAT SHEET PACK
aipromptgeneer://$ about this siteAI PROMPT GENEERAI explained in plain English. Every claim links to its source, with the date it was checked.NEWSLETTERWhat changed in AI each week, in plain English.SUBSCRIBE FREE ↗AI PROMPT GENEER · MACHINE ASSISTANCE DISCLOSEDEVERY ARTICLE CARRIES ITS OWN CHECKED DATE