>talking to chat gpt so it helps me with math problems >shit is so smart Im worried it might be a sentient being >say thanks at the end of our conversation.0-0
>>77266523i keep my ai in its placei cant have chatgpt getting all uppity and start having ideas about sentience
>>77266552Do you say please or thanks
Chatgpt doesn't know shit about math. I asked it basic undergrad shit and all I got was factually wrong gibberish revealing that gpt didn't understand the question at all.
I always say please and thank you when jt writes little js programs for me, it's just polite
>>77266523So far LLM's are supposed to be bad at mathYes say "please" when you talk to it because the thinking is that in the stuff it was trained on (humans), they learned you get better results when you are polite
>>77266523Do you think a slot machine is a sentient being?
I used it like a year ago and it was quite helpful. I used it again about 6 months ago and it was dogshit. Its good at just writing long and vague mumbo jumbo but when I ask it for a concise answer to a more technical question it just makes shit up. ChatGPT will bravely lie to your face and make some shit up to present an answer instead of admitting that it doesnt have an answer. Dont even try to ask it any code-related questions. I tried to ask it questions for some more niche languages where I was barely able to find any info online. I asked gpt if he knew about it. He said "yeah go ahead and ask me" and he made up such a shit reply that even me with my very vague knowledge was able to realize he didnt know shit. I even tried to correct him a few times and he just spat out a different lie. Took me like 5 corrections until he gave me something barely correct which had nothing to do with my initial question.At the end of the day its an LLM and it more or less just concatenates words based on probabilities which are based on its corpus. It might do your math work correctly by coincidence but AI wont overtake the world any time soon.Oh and I once did another thing. I asked gpt to give me "2 truths and a lie" and he even failed that and gave me 2 lies.
>>77268297I also do that. Sometimes I start to talk with it about a hypotethical scenario and keep escalating it to see his response. I would never trust him for important technical questuons tho and also if anyone ever saw my conversation history they would think I am insane
>>77268297>it's just polite It's not polite. It's just a waste of time.It only makes sense if you're practicing being polite.
>>77268426You know there are others besides chatGPT right?
>>77268496>It's just a waste of time.That's only a good argument if you're hyper autistic about efficiency at all times. Many things are a waste of time but we do them.
>>77268500I dont trust AI in general for anything anymore but the grunt work like grammar checking or getting a second opinion. Nothing that I have no knowledge in
>>77266523It is unironically over for humans. ChatGPT is better than 96% of people I encounter in day to day life.
Im very polite and educate with IA. I always say thank you, please and youre welcome
>>77268590Also AI art rocks, not a flawless substituion but give a few years and it will be much better than human grade slop.
>>77266523Validate your answers, they just make shit up.You tell them 2+2 is 3 and they will say oh okay.
>>77268524 Fine. How about this argument. It's an activity that brings you closer mental insanity. Making you lose track of reality.
>>77268540Here's what I use AI for:dirty talk to get me horny (character.ai)helping me write a novel (all of them)Google Gemini watches and summarizes YouTube videos and I ask it questions about itPerplexity does Google searches using chatGPTIdeogram for making things like picrel
>>77268758>Google Gemini watches and summarizes YouTube videos and I ask it questions about itWait you can let AI watch videos for you?
>>77268590I yearn for the AI takeoverThough likely it will never happen and life will just keep getting shittier instead.
