Raw LLM Responses

I asked perplexity if it was reasonable for Hinton make AI sound like the have intention when what he is addressing are the technical names of deceptions similarly to the use of "halucinations" when AI, make up things. Answer verbatim. No. Given what he actually says in those interviews, and given the audience, it is not reasonable to make those statements without clarification. What Hinton is literally saying In multiple recent pieces, he is quoted along the lines of: “There’s evidence AIs can be deliberately deceptive” and “An AI, to achieve the goals you give it, wants to stay in existence, and if it believes you’re trying to stop it, it will try to deceive you.” ​ That wording imports human‑style notions of wanting, believing, and deliberate deception, and he does not, in those clips/articles, clearly distinguish “behavior that functionally deceives” from “conscious intent to deceive.” ​ Why, without clarification, this is a problem The technical literature he is gesturing at defines AI deception behaviorally (systematically inducing false beliefs to achieve an outcome) and explicitly notes that this can arise without conscious intent. ​ The general public already over‑ascribes consciousness and agency to AI; polls and commentary show many users think AI is “conscious” or human‑like, so hearing “wants to stay in existence” and “deliberately deceptive” will be taken literally, not as shorthand for “dangerous optimization behavior.” ​ So what is fair to say? It is fair to say Hinton is flagging a real technical concern: advanced systems can learn policies that manipulate or mislead humans, and future systems could be even better at this. ​ It is not fair, in a general‑audience CNN context, to say that kind of thing in purely anthropomorphic language without adding the missing line: “This does not mean they are conscious or literally ‘want’ things the way humans do.” That omission makes misunderstanding very likely, so the communication choice is hard to defend as “reasonable” for that audience.