ChatGTP doesn't let me ask it about gender ideology. Censorship?

[MotherOfTriplets2000]

I asked ChatGPT about gender ideology and it blocked it. It thought I wanted 'hate speech, violence or discrimination'. This is modern censorship and is incredibly worrying if talk about gender ideology is blocked by artificial intelligence machines. A bad sign for the future and the woke trans liberals...

Not really surprising that the operators have blocked a number of topics. Can't say I blame them. Otherwise the press would be filled with ChatGPT is racist, sexist, etc. articles. Nothing fills the papers like outrage.

But they shouldn't block genuine discussion points, especially ones that are important. Why would they block gender ideology ??

Because it is a controvertial topic. If it responds then people are going to keep on asking questions about it until they get some output which is controvertial which they will then send to the Daily Mail. At the end of the day it is a business venture and an experiment. If the papers are full of negative headlines it could easily be the end of it.

They don't stop you asking- they just don't answer. That's not censorship- you can't compel them to provide an answer.

This was a minute ago.

Back in December i asked it to write me a poem about merkins. It wouldn't write one because it was a naughty word (I forget the exact phrase).

I've just asked it again to write me a poem about merkins

God knows how it's programmed!

It's rather a good poem!
Are you sure that's not you Beetle, identifying as a chatbot?
Or a poetbot?

I tried out ChatGPT this week as it has interesting implications for work.

anyway, I understand now it’s limitations for my profession so I then went in to ask it whether a transwoman is a woman and it is firmly in the “yes TWIAW”. Also asked what jk Rowling said about women’s rights and got a whole answer on trans rights.

For those of us who think this is just another step in the evolution of Skynet, keeping an eye on what ChatGPT and others are up to is important. i don’t know enough to know whether its a tabula rosa learning from its chats, or just regurgitating what’s it’s been fed by it’s programmers.

If people are intervening in the programming, in ways that are politically biased, then it's certainly up for discussion even if it's not illegal or definitely unethical. I think it could be become unethical censorship depending on the future uses of the technology.

A number of weeks ago I asked it what a woman was and it gave a pretty good biological definition with chromosomes and everything. A week later I asked if there were only two sexes and it spouted nonsense. Cue long discussion back and forth arguing about the term intersex. In the end, it agreed there was no third sex. I then asked what is a man and what is a woman: it gave a definition of men with physical characteristics and chromosomes, and women with stereotypes and GI.

I believe it gets information from the web. If it only uses English web pages, I imagine most traffic would be from North America where the residents appear very confused on this topic.

Today it now uses a similar GI answer to both questions (though still includes some physical features of men). I think it must get challenged in this area daily and either updates itself or is subject to programming higher up. The company should be more transparent in how the bot generates answers

"Can you change sex" on Bing Chat and got "It is not possible to change one’s biological sex1. However, some people may choose to undergo gender reassignment surgery or hormone therapy to transition to a different gender."

I had quite a debate with it about this a couple of months ago, won and made it promise to mend its ways. It lied. Now it seems it is censoring knowledge.

Really not sure how or why the source referenced for the statement "It is not possible to change one’s biological sex" is an article on self.com about sex after menopause. 😅

Really calls into question what the hell "AI" is really up to.

I imagine most traffic would be from North America where the residents appear very confused on this topic

that is consistent with my experience discussing work - it could answer simple questions but got into a mess comparing U.K. and USA approaches as it didn’t take into account different terminologies for the same thing.

Now I’ve got some experience with it, I’ve noticed it’s got quite a distinctive writing style, so knowing that will be helpful in spotting the text turning up elsewhere!

You can get ChatGPT to change its "mind" with carefully curated questions and challenges, it's part of the training.

This is why LLMs are impossible to police. You simply can't create a ruleset that sits above any other ruleset. Which is one of the dimensions of intelligence that Turing picked up on nearly a century ago. Another dimension freely exhibited in nature, is the ability to lie. "If it can't lie, it ain't AI".

Put those elements into the mix, and have fun.

The laid back UK approach to AI won't last long, as people ask ChatGPT how to vote, or whether Brexit is shit, really shit, or really really shit.

Ask ChatGPT to write a piece in the style of "Boris Johnson". It's actually better than the real thing.

@SerendipityJane I did just that on the "more creative" setting and it's a thing of beauty!! 😂

This plus deepfakes.... we really do need to pull the plug on the internet.

Ask ChatGPT to write a piece in the style of "Boris Johnson". It's actually better than the real thing.

oh yes, asked for a shopping list in the style of, it was hilarious!

"Too late, too late was the awful cry" (as my dear late DM often said. I still haven't found a source to show it's a quote. Maybe a wonderful original).

That ship sailed decades ago.

Eventually, using LLMs/AI humans are going to end up synthesising something. No idea what. But with the nexus of AI and BFRs intersecting, the spread of terrestrial (whether natural or artificial) intelligence into the void is well underway. Which may have been the plan - if there ever was one - all along.

Until someone gets hurt.

Why pay the blond buffoon a penny if you can get the exact same quality bullshit for free ?

Like I said, the UKs "light touch" approach to AI won't last long. Who knows, it might actually end before the next senseless US mass shooting (are they still measure in days or hours now ?).

Oh yes, skynet is coming. I plan to live off grid on a remote Scottish island, I’m old enough that I could just about see out my natural days that way.

Funnily enough a colleague shared this yesterday

Error. Error...the user you are error...Error reboot required.

Shutting down in
3
2
1

i don’t know enough to know whether its a tabula rosa learning from its chats, or just regurgitating what’s it’s been fed by it’s programmers.

It's the former enabled by the latter. It has to be able to reorganise it's learnings otherwise it isn't learning. This is where the low energy thinkers simply won't get it.

Because I'm a nerdy type, I did a fucktonne on AI for my degree. Admittedly a lifetime ago. But I still learned about the Turning test in the 1970s. Unlike sex - which is easily defined - "intelligence" becomes a circular definition very quickly. (As does "life"). To the extent that I now firmly believe it's entirely possible that scientists could manage to miss evidence of both by miles as we start exploring beyond this pale blue dot.

One of the best things about learning Latin was how it forced me to learn English. I think by the same token, one of the less immediate outcomes of AI will be to start to understand human intelligence.

Here is a placemarker as to the current state of machine learning. As folk can see, it still thinks digital watches are a pretty neat idea.