ChatGPT can get in the fucking bin. Fuming.

[noblegiraffe]

There was a thread on twitter asking people to see what ChatGPT said about them, so I tried it with my MN username. After correctly identifying that I was a well-respected member of the MN community I decided to ask it some more direct questions.

Anyway, ChatGPT confidently told me that I was an openly Conservative voter who regularly expresses right wing views and then that I voted for Nigel fucking Farage in 2015.

What. The. Fuck.

I have to say that any remaining faith I had in AI has been shattered by this.

Most people (one would hope) are saying what they believe to be true, even if it is actual bollocks. Obviously trolls and blaggers exist. But ChatGPT doesn't have that as a concern. It isn't that it wants to lie, its just that it doesn't know what truth and lieing is. (If you asked it to define truth and lies it could write a definition of each. But it doesn't know what it is in the same way a human does).
So, if asked a question, it tries to answer it in a way that sounds convincing and in a manner that sounds like a real person. All good fun. Absolutely ludicrous that people would rely on it to actually give accurate information. It doesn't have the ability to know what that is.

Oooh! You can ask it "Which poster on Mumsnet is an expert on <enter whatever subject you want>?"

I asked it who the sex expert was. 2 names came up. 😂

There is no "real" poster for that story. Its summarised some stereotypical mumsnet stories (troublesome mother in law, mumsnet thread becoming a classic) and written its own version based on that. In some ways its very creative. Its why you can get interesting results if you ask it to write a story based on X and Y (it looks for lots of different similar stories and borrows elements from each). But in this case, its like you asked it to "describe a typical mumsnet post" or "imagine a story where a poster n mumsnet causes furore" and run with it.

Its interesting because it is already interpreting what people are asking for in creative ways - but it doesn't understand the difference between creative writing and accurate reporting.

The the same could be said "Absolutely ludicrous that people would rely on it to actually give accurate information" with the general population, as i think books and textbooks or experts are the more reliable sources

Understanding the Nature of AI Responses

1. Statistical Prediction vs. Knowledge:
2. ChatGPT, and similar models, operate on the basis of statistical correlations between words and phrases learned from vast amounts of text data. When asked a question, the model predicts what the most likely sequence of words would be, given the input, without understanding the underlying reality.
3. The model doesn’t “know” things in the way a human does. It doesn’t have beliefs, consciousness, or an understanding of truth. Its responses are generated based on patterns in the data it was trained on.
4. Truth and Lies:
5. Humans have an intrinsic sense of truth based on perception, memory, and reasoning. We understand truth as a correspondence to reality and lies as intentional deviations from that reality.
6. ChatGPT can define these concepts because it’s been trained on text where these terms are used, but it does not have the subjective experience or cognitive framework to “know” what truth or lying is. It simply outputs text that matches patterns found in its training data.
7. Accuracy and Reliability:
8. Since the model does not comprehend truth, it can sometimes generate information that is inaccurate, outdated, or misleading, even if the response sounds convincing. This happens because the model lacks the ability to verify facts or understand the context deeply enough to ensure accuracy.
9. While the model can be a useful tool for generating ideas, drafting text, or exploring topics, it should not be relied upon for definitive or critical information without further verification from reliable sources.

Implications for Users Given this, it’s important for users to approach AI-generated content with a critical eye. Recognizing that these tools are not a substitute for human judgment, expert analysis, or rigorous fact-checking is crucial. AI can assist in many areas, but it is not infallible and should be used as one of many resources rather than the sole source of information.

My ChatGPT told me GC people might not believe I'm a women now I've had hysterectomy. Apparently individual's GC views significantly on this.
So there's that

Yes, but people seem to rely on it more if anything because there is an idea it is compiling lots of information together - e.g. it can search all the information on the internet about boat engines to give the most accurate information about boat engines. It knows everything but it also knows nothing.

There are definitely people out there using ChatGPT like google. Or asking it for medical advice!

It has its place. It's quite useful for suggesting an itinerary at a holiday destination. It provides a useful starting point. I've also put in "If I like X destination, where else would I like?" It's thrown up some interesting ideas that I wouldn't have thought of otherwise.

You might find yourself trying to go to places that don't exist!

Hilarious 🤣

After correctly identifying that I was a well-respected member of the MN community...

🤣🤣🤣

Its not designed to give true or accurate information. Its basically trained to sound convincing.

It's definitely male.

ChatGPT says:

"Chersfrozenface is a well-known and active user on the parenting forum Mumsnet. The name is a playful reference to Cher's famously youthful appearance, suggesting a sense of humor and a nod to popular culture. On Mumsnet, Chersfrozenface is recognized for their distinctive posting style, often contributing to discussions with wit, sarcasm, and strong opinions. They are regarded as one of the site's memorable personalities and have developed a reputation among other users for their blunt and straightforward manner of expressing their views."

Well, that's not libellous, anyway.

Oh, and autocarrot wanted to turn ChatGPT into 'catgut'.

I admit it has caught me out twice (or at least the co-pilot thing has) with the sources present when it has incorrectly summarised or made a false claim.
Luckily, it was just on here not for school but it has taught me - like we teach the kids - to check and check again all sources and references to said sources!!

What was fascinating was using it as a therapist. I was curious as to AI psychotherapy/friendship bots so had a conversation with the Pilot and I can see how that might work in the future. I thought it was going to be as annoying as the chat help you have on sites like British Gas, Insurance etc but it has been tweaked a little, to show more empathy!

I've just chat GPT'd me.

It says "The name seems to be a playful or humorous moniker rather than a reference to any particular public figure or well-known individual"

Just proves it knows nothing.

Omg
I'm shocked at my choices.
I will indeed have a word with myself.

In rl I've cut people off for less. 🤣

Seemed fairly accurate about me.

"CormorantStrikesBack" is an active user on forums like Mumsnet, where they often engage in discussions related to social and cultural topics. The user is known for expressing strong opinions, particularly in debates around gender and identity. While the exact identity of CormorantStrikesBack isn't publicly known, the user has built a reputation for participating in conversations that can be polarizing, particularly around issues like misgendering and the inclusion of transgender individuals in certain roles. Their comments often reflect a critical stance on certain social trends, which can lead to intense discussions on these platforms

Well nothing about you as a user on here 😂

I just need to say that while generative AI is fun to play around with, it is also utterly appalling for the environment and labour rights. One hour training the data uses as much energy as 80 households in a year.

I take it back. Just asked it what my political views are and it said conservative. Which is rubbish. Define seems to equate gender critical with being right wing.

It didn’t know what I do for a job or if I have children even though I have mentioned my job on here and talk about Dd a lot. Maybe it doesn’t know what Dd means!!

Oh it does know what Dd means, just asked.

In fairness chatgpt correctly “diagnosed” dh with an unusual medical condition when the GP had been stumped for months.

I then asked it about my collection of symptoms which I’d been struggling with for years and the drs had been useless and it suggested something and that’s now been confirmed by a consultant.

A lot of the stuff it said about you is quite general. "engages in discussions related to social topics" "is known for her strong opinions". It is using the same stock phrase to describe multiple people, and tweaking them slightly. Its basically "describe a typical mumsnet user", "include quirky details about a story she might have told once". it is using peoples posts - the fact that someone mentions they are a teacher definately made it in, as does the fact you post a lot on FWR but its just writing a "convincing" description based on that.