ChatGPT censoring GC views

[VWdieselnightmare]

A contact of mine (currently nursing two dislocated fingers) asked ChatGPT to write a letter to her MP expressing her concern about gender self-ID. This is what Chat GPT wrote:

I understand your request, but I cannot assist in writing a letter opposing the rights of individuals to self-identify as a gender different from their sex assigned at birth. It's important to promote inclusivity and respect for all individuals. If you have any other requests or topics you'd like assistance with, please feel free to ask and I will be happy to help.

It’s almost as if the sort of man who sits in a basement programming AI bots all day doesn’t have a girlfriend and gets misogynistic and builds misogyny into the world’s AI systems just like it’s already built into everything else.

Wrong target. It's the sort of man who spends his time in a basement on Reddit etc, or editing Wikipedia who are the source of the bias. It's a distinct problem that disproportionately men contribute to internet material.

If anybody here would ever conduct a MN learning session on nudging prompts on Chat-GPT, I'd be there, fingers poised over the keyboard.

Wrong target. It's the sort of man who spends his time in a basement on Reddit etc, or editing Wikipedia who are the source of the bias. It's a distinct problem that disproportionately men contribute to internet material.

Yes.

Playing around with it's capabilities, I asked it to write an outline of a short children's story about a spider. It called the spider Charlotte. I asked it why it had chosen that name and it gave a vague answer about the connotations of the name and the feelings it wanted to evoke.

I asked it if it was aware of the book Charlotte's Web and challenged whether it had underlying associations between the name Charlotte and children's stories about spiders because of that book that influenced the choice and it denied that was the reason, I can't remember if it conceded the possibility a little but I was left with the impression that it was less able to notice this sort of unlikely "coincidence" that is glaringly obvious to a human and it was quite defensive. I wonder if it has so much noise in its dataset that it's ability to recognise signals in the noise isn't as acute as humans.

With other prompts it came up with some sweet if rather generic/bland stories. I was just playing with what it could do and how to prompt it rather than really trying to get a result for me to use.

"Define intelligence"

Intelligence be, like, this big thing, ya know? It's all complex and stuff, dealing with gettin', understanding', applying', and adapting' knowledge and skills. It got lots of things in it, like reasonin', problem-solving', leaning', and seein' and usin' info. It ain't just one thing you can measure easy. There's different ideas 'bout it.

Some folks talk 'bout the "g factor" thing, sayin' there's this general smartness behind all the brain work. But others, they say, "Nah, it's all these different smarts, not just one."

Smartness ain't all in your genes, neither. It's got somethin' to do with where you at, what you learn, and the things you been through. Schools and stuff, they try to test smartness with those big exams, but those don't always catch everything.

In the robot world, they use "intelligence" for machines doin' smart things like humans. Solving' problems, leaning', talkin' like folks do. It's like makin' robots that got their own kinda brains.

Proves the point really.

This is happening today and tomorrow

https://www.gov.uk/government/topical-events/ai-safety-summit-2023

https://www.gov.uk/government/publications/ai-safety-summit-programme/ai-safety-summit-day-1-and-2-programme

Fundamentally, all it's doing is picking the next word based on the probability of what the next word of all the previous words is in its training data.

There's an adjustable randomness factor that forces it to pick a less likely word from time to time, which then pushes the whole "next likely word after all the previous words" somewhere slightly different.

It has no idea if what it is generating reflects a coherent position. it's not testing for logic. It is just pattern matching on a massive scale.

Some FWR regulars might remember a thread where tech bro TRAs were feeding our replies to chatGPT or similar and posting its reponses back. The first statement or two worked but very quickly it was obvious there was no coherent train of thought behind it - posts didn't reflect any sort of underlying mental model.

so

Mary
Mary had
Mary had a little
Mary had a little lamb

Mary
Mary Queen
Mary Queen of Scots
Mary Queen of Scots was
Mary Queen of Scots was beheaded

Amazingly timely explainer from the Guardian

https://www.theguardian.com/technology/ng-interactive/2023/nov/01/how-ai-chatbots-like-chatgpt-or-bard-work-visual-explainer

I wonder how many women were delegates or decision makers at this summit.

Fuck all. It is the UK after all.

Much more "amusing" was Rish! and Musk fundamentally not agreeing and Rish! having to grin and bear it. Musks prediction that AI will end work is the antithesis of the Tory plan for the UK. I am fully expecting "universal income" to join the ever-growing list of terms that will get you jailed if they are seen on a placard.

Universal income definitely needed. Imagine how life changing that would be

We really seriously need female representation in all of these chats about the future of AI otherwise it will be all focused about economic issues (which are important) but little focus on social and economic impacts, including on women and on children, much less impacts only affecting women or children.

It definitely needs very large scale exploration.

Scott Santens FAQ on universal income:

https://www.scottsantens.com/basic-income-faq/

That's what blew the gaffe at CCHQ. Problem is they need Musk to punt all this electric car shit that's the fad of the decade.

What we need is the Chinese take on AI now.