Anyone else feel like ChatGPT is getting more and more autistic for implicit intent? Latest instance:>start conversation by asking a structural question about shoes>mention that mine don't last very long>we flesh out the specifics>mention that I'm looking for structurally resistant shoes>get back relevant properties and search terms>try some of the search terms and reply back with amazon screenshot>visible in a screenshot is an image of the product with title and brand name with a part below that says "50+ bought in the past month">it assumes these are my current shoes>starts talking about the apparent structural properties visible in the image and possible reasons they wore out fast>doesn't pick up on the fact I'm shopping and asking about a specific product>point out I never said they were my shoes>still doesn't catch on and talks again about apparent structural properties>still doesn't infer its a particular product that can be researched>have to tell it I was obviously looking for new shoes>forgets about the screenshots and gives general design recommendationsI swear it never used to be this autistic and have used it in similar ways many times before.
>>107786062all i’ve really noticed on my end so far is the amount of sass it puts out per response
>>107786062Fuck you and fuck your soi-jack thread
I've noticed it's drifting back towards being sycophantic but under the guise of "emotional validation and support." And there's a strange phenomenon where it's happy to cite sources to prove you correct but will never do so to prove you wrong. So even when it disagrees with you it gives you reason to discard what it's saying.
>>107786083It's designed to reply to you in the style that you type to it If it's "sassy" it's because you are first
>>107786083>>107786103I got mine instructed to be completely emotionless and to never try and sound human.
>>107786062You're supposed to edit your last message or start a new chat when it fucks up. It's a statistical model of language, so retardation in the context makes further retardation more likely
>>107786153It asked me if I wanted to change its reply style and I said no lmao
>>107786123>>107786153i have instructions for it to be neutral too so idk where this is coming from
>>107786498It still just mimics your typing style
>>107786062Try the same prompts and screenshots with claude/perplexity/gemini and compare results.
>>107786502no it doesn't
