Carlos Velazquez

Quoting: "Can this technology ever be truly trustworthy? For starters, suggests one AI expert, ChatGPT should admit when it's not sure."

I agree. ChatGPT will most of the time give superficial or just plain wrong answers initially. I'm pretty sure the algorithm is set to give the easiest and least resource intensive response. So the initial answers should not be trusted even in technical and scientific domains.

In politics and controversial topics, such as "global warming", it should not be trusted at all because it gives orthodox leftist answers. I've had several interactions with it regarding technical and scientific issues where I kept challenging its responses and ChatGPT had to correct its immediate previous answer and then eventually it admitted its initial response was wrong or misleading.

Reminder: ChatGPT is like a Google search that provides conversational answers with English text to prompts or questions. It behaves like a human in that it "remembers" everything in the ongoing "conversation".

Bottom line: one should have a good understanding of the topic one is going to ask ChatGPT about so one can ferret out wrong answers.

Just because AI recommends a cliff doesn't mean you have to jump
zdnet.com

Posted by: Timothy Birdnow at 11:04 AM | Comments (14) | Add Comment
Post contains 203 words, total size 2 kb.