violet wrote: ↑Sat Feb 04, 2023 3:19 pm

Jerry Atrick wrote: ↑Sat Feb 04, 2023 3:09 pm

Doc67 wrote: ↑Sat Feb 04, 2023 2:31 pm

Jerry Atrick wrote: ↑Sat Feb 04, 2023 1:56 pm
upload

It's merely the sum of its inputs

Can you ask for a poem praising our very own Prime Minister HE?

You might get it published in the Khmer Times.

Chat GPT is not available in Cambodia

I'll check it when on pc later as my pc is ostensibly not in Cambodia thanks to VPN

So, your earlier posts were from the internet and not generated by an interaction with chatGPT. I don’t know if you thought you were being funny, or if you succumbed to your bias.

It's not just the VPN. They also reject Cambodian phone numbers (or at least they did mine)

There are other ways to use it, but that's convenient for them

Jerry Atrick wrote: ↑Sat Feb 04, 2023 3:20 pm

violet wrote: ↑Sat Feb 04, 2023 3:19 pm
So, your earlier posts were from the internet and not generated by an interaction with chatGPT. I don’t know if you thought you were being funny, or if you succumbed to your bias.

What??

I believe she is implying that you reposted a screenshot from a dubious source and that it was probably faked to sway public opinion

(even tho there are too many examples to ignore, many of which can be proven to still work, and many that did work at one stage, but no longer do - they are constantly refining the algorithm and releasing new features, including a new paid tier as recently as two days ago)

violet wrote: ↑Sat Feb 04, 2023 2:36 pm

Jerry Atrick wrote: ↑Sat Feb 04, 2023 1:56 pm
upload

It's merely the sum of its inputs

Did you find those images online or generate those responses from chatGPT yourself?

I don’t get that response

@Jerry Atrick

violet wrote: ↑Sat Feb 04, 2023 2:36 pm

I'd just like to reiterate that the fact it doesn't work now does not reflect the response you might have gotten yesterday (or prior to the most recent version released) and given how many examples I've seen of this kind of thing out of ChatGPT it wouldn't at all surprise me if Jerry Atrick's screenshots were once a legitimate response from ChatGPT.

What this does mean tho is that despite it's bias, there's at least one person on the OpenAI team who has some common sense, and this is a good thing.

Go them for being better than the people who produced the content that trained their AI. Let's hope this continues.

I did consider that it would continually refine, but suspect this bias was either never there or it was there for all and the screenshots were taken at different points in time after adjustment (but keeping the original for latinos and black people). I’m not saying JA did this.

We all have a responsibility to fact check and not spread stuff that creates divide and reinforces negative bias.
As said, it will refine itself. So why put the stuff on the internet to be repeated to suit people’s agendas? Just give the feedback like it asks for so it can improve

violet wrote: ↑Sat Feb 04, 2023 3:17 pm Is chatGPT politically neutral?

I’m not trolling, trying to be witty, attempting to be a snappy headline writer - I just want accuracy that helps people new to the technology use the right bloody name.

Haha...I just got it. It is obvious now!

When I posed the question I was hesitant as tech is not my strong suit and thought I might get wafted aside by the more savvy posters for posting a dumb question. I should of at least got the bloody name right.

Yes, mods, save my further blushes and do the clicky changy thing.

violet wrote: ↑Sat Feb 04, 2023 3:40 pm I did consider that it would continually refine, but suspect this bias was either never there or it was there for all and the screenshots were taken at different points in time after adjustment (but keeping the original for latinos and black people). I’m not saying JA did this.

We all have a responsibility to fact check and not spread stuff that creates divide and reinforces negative bias.
As said, it will refine itself. So why put the stuff on the internet to be repeated to suit people’s agendas? Just give the feedback like it asks for so it can improve

The bias is there, check some of the links above about the dataset it was trained on.

Ask it some specific questions you know to be politically sensitive. There were some hilarious Biden and Trump examples, if I can find them in my history I'll retest some more prompts myself.

I'm sure they've addressed most of them BECAUSE of the screenshots, they certainly did not spend all that time and money training an AI for a few practical jokes.

And I've linked in this very thread to something about fact checking...