Correct, I haven’t.
ok cool. That sounds much better!
Surely though the AI should be created such that it won’t get fooled into producing ‘creative writing’… Otherwise it just seems rather ridiculous.
1 Like
I maybe shouldn’t have said tricked — perhaps primed.
This is from a ChatGPT clone (also OpenAI) that I have access to (it’s easier for testing because it’s not limited, but ChatGPT does the same):
A ChatGPT screenshot I have:
3 Likes
@parautenbach This might sound like I’m being critical, I am most definitely not, this little exchange about the ‘cow eggs’ has been very informative.
Ironically though it has felt a lot like a conversation with an AI. Eventually we got to a point where for me at least I feel much better informed about the whole subject. Just like with an AI the right question had to be asked with the right ‘priming’ to reach a result that satisfied me.
But I suppose that just highlights the age old problems of; asking a well defined question, on-line communication (e.g. chat dooms, forums etc.) and now the new problem of knowing what an AI is actually doing and of course, understanding if (or at least how much) it is [artificially] intelligent in the first place.
1 Like
I’m really glad if it was informative. No problem at all. It was the facetiousness (dishonesty?) of the original cow egg post here, because it added no value to the discussion: No insights or explanations were offered and without intervention could’ve lead to a lot of “see, I told you this is rubbish” when the reality is more complex.
I promise all my responses where my own – except where quoted. 
1 Like
I think (and someone correct me if I’m wrong) that ChatGPT has been specifically advertised as proficient at creative writing.
Assuming that’s true, I could see where it could be prompted to generate fiction about cow eggs. A different implementation using the same GPT code base could be designed to help generate non-fiction prose, like scientific articles. That would look like the examples above where it refused to speculate on cow eggs. In other words, in both cases it generated the desired response.
Hmm… How do we know with 100% certainty that ChatGPT didn’t write the OP? 🥸
I fully agree here. Thanks
… And @Herbert01 … You can sleep tight, ChatGPT can’t write any OP … thou some J… can copy/paste a Question/Answer, and that you can never be 1000% sure about
My answer was that I totally agree. By that I mean that I do NOT consider answers from ChatGPT to be useful either.
You can ask GPT to produce non functional information, but sometimes it fails and it continues to create fictional content.
It may happen when you ask about a treadmark.
Sometimes it will create fictional content based on what it inferred from the name and not from what’s published in science or press articles.
After all the model is still learning and issues should be asserted.
Never Ever trust ChatGPT to provide correct facts, it’s just a dumb language model without any real intelligence that imitates your native language. Thus, before posting anything that is originated from ChatGPT you have to verify all the facts yourself.
1 Like
I know what it does, but your post proved nothing. You’ve been deliberately misleading. I’ve explained why already.
1 Like
Actually we might say the same about humans, if someone tell me a fact he thought have heard, he might be wrong.
Often, when someone thinks something it doesn’t makes it real.
We’re all telling misleading information.
Therefore when something is really important we must always check by ourselves
Exactly. As a tradesman myself I’ve often had older tradesmen try and tell me I’m doing something wrong because it isn’t the way they do it. My answer has always been: ‘just because you have been doing that way for years doesn’t make it the best way to do it’.
2 Likes
As judges tell juries: ‘an honest witness can be a mistaken witness’
2 Likes
I’m old enough to remember the Terminator movie, and all of the Asimov Robot stories. I do not like talking to machines. It is one of the things that I hate.
2 Likes
Are you sure you didnt create this post with ChatGPT? Come on now…be honest.
Some people really should go ask the oracle about the dunning kruger effect.
No one is dying so chill out and save your faux indignation for social media. This is not the place for it.
1 Like
Hard to tell who you are responding too.
1 Like
not really