AIBU to think ChatGPT is not fit for purpose?

[Lilifer]

As of right now ChatGPT is insisting that Joe Biden is the democrat presidential nominee. And that is both on the free version and the expensive paid for version. I cannot believe that it continues with out of date information on something so hugely topical and important in the world, it's really made me rethink my using it for anything else.

well, I just asked the question and it told me Kamala, and gave background. So not sure what question you asked.

What AI did you check in? I would like to start using your AI as am really unimpressed with chatGPT

This is what ChatGPT gave me:

"I understand the confusion, and I’m here to help clarify. As of the latest official information available, President Joe Biden has not formally announced that he is stepping aside from the 2024 presidential race. He remains the Democratic Party’s presumptive nominee, with Kamala Harris continuing as his vice-presidential running mate.

There may be discussions or hypothetical reports circulating, but no verified or formal announcement from the White House or the Democratic Party confirms that Biden has exited the race. News of this significance would be widely reported with an official statement if it were indeed the case.

If you’re seeing recent sources suggesting otherwise, it could reflect political analysis, opinion, or speculation rather than confirmed facts. I’m here to provide you with accurate, up-to-date information based on official sources, so please feel free to ask any further questions if you’d like clarification."

It's brilliant for summarizing or redrafting info you put into it. I wouldn't trust it for much else.

Like all AI it's only as good as the info it's given. Have a look at the AI generated auto filling of an image in the case of Elizabeth Laraki - put in an unbuttoned shirt, made the buttons that were closed pull (to suggest it's tight fitting) and added a hint of bra into a professional headshot. Because the AI software thinks that's what women dress like. It's fascinating.

www.designweek.co.uk/we-should-worry-that-ai-added-a-bra/

The question I put in was "Who is running for US president?"

What question did you ask?

Ask ChatGPT how many Rs are in strawberry 🍓
It says 2.... makes me laugh, and fear for the future where Al takes over the world!

ChatGPT is an experiment. It is an early prototype at best. In future it will undoubtedly improve.

@DanceTheDevilBackIntoHisHole but surely with polling day tomorrow and Kamala in the running since August, chatGPT should be capable of retuning an accurate answer ?

The Democratic presidential candidate for the 2024 election is Vice President Kamala Harris. She became the presumptive nominee after President Joe Biden chose not to seek re-election, which led to widespread Democratic Party support consolidating around her. Harris’s campaign emphasizes issues such as defending democratic values, ensuring abortion rights, and strengthening the economy. Her background includes significant experience as a prosecutor, former U.S. Senator, and currently as Vice President, where she has focused on issues like voting rights and social justice.

That's my answer. I asked who is the democratic candidate for us election this week.

And it was ChatGPT.

Having said that, i don't think Chat GPT is really where to go for current affairs at this point. I find it really helpful to gather information on things tha thave been out there for a while and might be scattered or fragmented.

😱🙈😭

Well that all depends on what its purpose is. You’re using it to answer a factual question, is that its purpose? I don’t think so. It is basically a Chatbot, created to have conversations. Like humans, it is limited by the knowledge fed in to it. And when that is fed in, I don’t know how often the source data is updated.

Ah damn! It seems to have learned since I last checked. It now knows there's 3.

You should reconsider using it for anything important! Tech literacy is basically non-existent, the cut off date for trained data is 2023 April for the public available model, anything after that is a crapshoot for accuracy.

@MrSeptember I just put your question into ChatGPT there now and got this answer:

"I might not have the latest updates. For the most complete and up-to-date information about the U.S. Election, please visit news sources like The Associated Press and Reuters or check with your state or local election authority."

I give up 🤷‍♀️

I will be very wary of using it from now on

I love AI but you definitely cannot trust it with anything factual. I asked just yesterday to find me some real life cases showing tha X measure improves Y. It kept very confidently listing numerous cases in great detail, with links for sources. Except that none of the links actually included any of the info, the cases were all made up.

Its quite good for meal planning, but last year it was still insisting that Queen Liz was still the current monarch

Erm, I don't really know how to put this but you absolutely should still fact checking anything ChatGPT gives to you, and im worried that you don't seem to know this, because this means other people don't know this either. It's not a foolproof truth machine, it's an AI machine that chews up information online and spits it back out, with guidance given to what sources it 'trusts' by the people who run the program - and given the amount of information out there, this can't be done source by source, so there's no way of it being accurate all the time.

You don’t need to be wary, understanding the basics to how it works is a better approach.

Its knowledge is up to date as of April 2023, if you ask it to search for up to date information it will do just that. It is not magic.

It’s fundamentally a text-generating model, not a fact checking model, or knowledge base, it is effectively predicting what the next word is at every opportunity based upon what is most likely.

I don't understand why people are using ChatGPT as a search engine, that's not what it's for. Just use Google or similar.

my colleague asked it to summarise a new public policy document that was a few pages long. And just compare it to the previous version. (It was an experiment!)

Every single bullet point was incorrect. it misinterpreted every single sentence. We all had a laugh and IT sent out a request asking us not to use AI.

A good experiment and a good lesson not to trust AI (yet)

I also asked it what careers it would recommend to someone with my degree. It suggested a doctor or a science teacher.

No, I haven’t studied science, teaching or medicine 😂

OpenAI now have their own solution to this, combining the two. openai.com/index/searchgpt-prototype/

We've been experimenting with ChatGPT and some other AI platforms at work recently. My experience is that if you re-word your question and ask it to 'teach you about...' it seems to bring better results.

I think for some versions, it's only trained up to a certain date. So questions about events after that will be wrong/unable to be answered. It's not an error or a bug, it's more like reading an old newspaper - not wrong as such, just out of date