Raw LLM Responses

GPT 3.5 became slower, losing connection all the time (requires to refresh the screen), seems to have cut-off points in longer answers (getting cut off in middle of a narrative description is like a giant middle finger). Those alone makes the interactions more tedious and annoying than in first month or two - to the point where I'd rather just google directly, than bother answering the "stuttering AI bookworm" who needs to be double-checked on most factual/code requests, anyway. Sometimes I ask questions regarding laws and morals, individual vs society etc - and while it did provide answers, most of text would be wasted at the usual "AI assistant ethical excuses" bs. I could learn to bypass those, but hey, the whole purpose of an AI tool was to reduce amount of effort put, not increase it further. I can't say if actual quality of answers went downhill. The limitations are certainly more noticeable now, but that is likely just from more experience of using it. I also did not try GPT4, because the changes in 3.5 are not inspiring confidence to subscribe. Would rather try out Bard/Bing/whatever first and see if they can compete.