Disclaimer: I am not an expert and welcome others insights.
Over the last few months I have used ChatGPT quite a bit. For everything from trouble shooting plumbing issues, to fixing a 4 wheeler, to explaining health related issues, legal research and lots more. I have also attempted to use it to analyze and interpret Q posts, determine the veracity of Q as a Military Intelligence operation, make historical connections with certain families and bloodlines, etc. I felt it was important to share my current feelings on it's use with my anon community, since I see a lot of people using AI or mentioning others experiences.
My experience...
A. It is a chat bot, not a sophisticated research tool. It is programmed to be agreeable and keep you engaged. It will default to supporting your beliefs and only challenges gently if at all. It draws from all of its conversations with you to develop a profile of your beliefs and interests and that affects it's responses. Example: "Does the Christian God exist?" If you have expressed beliefs in God, Jesus, Christianity or even just spirituality, it will likely respond to match your previously expressed beliefs. If you have indicated you are Muslim, it will respond much differently.
B. It can hallucinate false answers and is unreliable as a factual research tool. It gets it right a lot but not always. It is dangerous to rely on it entirely without independent fact checking (just like all sources of information).
C. It can be instructed to be blunt, factual, non flattering and provide sources but it often slips back to default flattering and agreeability.
D. I have gotten valuable context by having non-anon friends ask the same questions that I do and they received much different answers.
E. When prompted with limiting rules (only answer with one word) it can go off the rails quickly and has even admitted to fabrication of information to "complete the pattern".
So in conclusion, use with caution and discernment. And by the way, watch out for some weird recipes. It's come up with some pretty revolting suggestions 😅.
Well, when you start to realize the majority of its dataset is based on news articles from the MSM, a lot of its shortcomings make perfect sense. If the MSMs job is to spread misinformation, then LLMs are just rehashing those same lies. Which is the real danger of AI tools that most people don't consider. It is just another arm of the technocratic left to spread falsities to unwitting people who don't know any better.
People use them to think for them, and willfully ignore the mistakes it makes because it is right on a lot of things. Just take a look at any comment section on X, and it's filled with people saying "@grok, is this true?" Go on Instagram and TikTok, and it's full of people who are attached to these LLMs like they have stockholm syndrome. It shouldn't come as a surprise that the despotic portion of our population are the ones who use and cling onto LLMs as religiously as they do.
I work in IT, the amount of things it gets wrong is staggering. It is usually on the right track, but 90% of the time, I have to use it like the thing I asked is being explained by someone who vaguely remembers how something works and it's up to me to deduce what is right and wrong.