Questions, answered
Each page gives a short answer first and then the argument for it, drawn from my published work. Where a question is still open, the page says how far the evidence goes and what would settle it.
Questions people ask
- Can AI be conscious?No current AI system is known to be conscious, and the leading scientific assessments judge it unlikely, but nothing rules out conscious AI in principle.
- Can AI have free will?It depends on what free will is.
- What is the Tracking Minimum?The Tracking Minimum is a necessary condition on passing on knowledge by telling: a source can transmit testimonial knowledge that p only if the truth of p, or the evidence for it, figures in the explanation of why the source asserted p, and does so through the source's own operation rather than through a proxy on which that operation depends.
- Is an AI chatbot an informant or an instrument?In the structure of the warrant it gives you, a chatbot is an instrument, and an unusual one.
- Should you trust what ChatGPT tells you?Rely on it when you have reason to, but do not trust it in the way you trust a person who tells you something.
- Does ChatGPT lie, or is a 'hallucination' something else?It does not lie, because lying is asserting what you believe to be false, and nothing in how a language model produces an answer is a belief that the answer contradicts.
- What happens to knowledge when people pass on AI answers?The chain of testimony breaks at the AI.
- What is constitutive dependence?Constitutive dependence is the condition of a person whose operative standing as a bearer of rights or entitlements is produced by an institution rather than merely administered through it, who has no standpoint from which to refuse or correct it without losing that standing, and whose recognition the institution can revise through its own operations.
- Can consent make Aadhaar legitimate?Not for the people who depend on it most, because consent presupposes someone who already has the standing to refuse.
- Can AI think?It depends on what thinking is, and that is the question in dispute.
- Is AI biased, and can the bias be removed?AI systems are often biased.
- Can you cite ChatGPT as a source?You can cite it as the source of the words it produced, and if you used it you should say so.
- Should AI make decisions about people?Only under conditions that are usually not met.
- Should AI have rights?On the evidence we have, no AI system that exists today should have rights.
- Does using AI make you worse at thinking?Using AI can make you worse at thinking, though it need not.
Questions I am working on
- Can AI testify, or is a language model's answer something else?Not in the sense that matters for knowledge.
- What does an institution owe the people whose standing it produces?Where an institution produces its subjects' operative standing as entitlement-bearers, its legitimacy depends not on their authorisation but on discharging three obligations they cannot waive: legibility (recognition criteria that fit the material reality of their lives), non-abandonment (not withdrawing the standing it produced without meeting a burden of justification), and remediation (a real path, inside the institution, when recognition fails).
- Do large language models know what they say?On the best current evidence they probably do not, though the question is open in a precise and testable way.
- Who is responsible for a belief that reached you through a machine?Responsibility falls mostly on you and on the people who built and deployed the machine, though not in the way a speaker is responsible for what she tells you.
- Can a system act for reasons without being conscious?A system may act for reasons without being conscious; the question is open.
- When an AI system causes harm, does responsibility disappear or move?Responsibility for harm done by an AI system moves; it does not disappear.
- If a mind can be copied, paused and resumed, is it the same mind?There may be no fact of the matter that settles it, and that may matter less than it seems.