The ChatGPT Thread

[trophyshy 8997]

3 hours ago, strawb said:

Why are red peppers so much nicer than green?

Because FUCK GREEN. That’s why.  

[strawb 5407]

18 minutes ago, trophyshy said:

Because FUCK GREEN. That’s why.  

Good point 

[Andrew 7459]

4 hours ago, Monkeys Fist said:

CT finding out that green, red and yellow peppers are just the same plant at different stages of ripeness 

CT has evolved beyond the need for "food"

 

[RobinRobin 14490]

Just now, Andrew said:

CT has evolved beyond the need for "food"

 

[Howmanheyman 43837]

6 hours ago, RobinRobin said:

"Come here! ..... Come here and listen! .... I'm not even hungry, man!!"

 

 

FYP

[Meenzer 17990]

Bloody hell  

 

https://interestingengineering.com/culture/2wai-digital-holoavatar-app

 

 

[Meenzer 17990]

 

[MrBass 4182]

On 03/11/2025 at 10:45, Monkeys Fist said:

That’s a link to Vogue, a magazine about knickers and haircuts, written for splitarses by splitarses. 
Their relationship to food is combatant/foe. 
 

 

Show me an Italian chefs recipe with peppers in his risotto. 

 

Gonna have to ask ChatGPT what a 'splitarse' is 

[Monkeys Fist 50629]

1 hour ago, MrBass said:

 

Gonna have to ask ChatGPT what a 'splitarse' is 

It’s a perfectly cromulent term for one of those woman types. 

[MrBass 4182]

[Meenzer 17990]

https://www.bbc.co.uk/news/articles/c8drzv37z4jo

 

 

[Rayvin 8315]

For a contract I'm working at the moment I conducted 20 hours of stakeholder interviews for a discovery phase to understand their system setup. I inputted all of the transcripts into ChatGPT, fed in workflow documents, and then started interrogating it about what it had 'learned'. I have to say, it was immense. It's probably saved me 30 hours of work.

 

The idea that it's useless simply isn't true. I basically turned it into an encyclopedia on this company. Yes, you have to sense check it - and now and then it did get something confused or make an association that it shouldn't have, but honestly, it's only occasional. I think using them well is probably a bigger issue than their reliability.

 

Just being able to ask it "Where did this idea come from and what do we mean by it" and for it to go "It came from our conversation with xxxx specifically in reference to yyyy and is reinforced by another conversation with zzzz" is incredibly helpful.

[ewerk 39152]

Yeah but can AI pair my entire family’s socks? Until then it can fuck off.

[ewerk 39152]

FFS.

 

 

[Meenzer 17990]

2 hours ago, Rayvin said:

Yes, you have to sense check it - and now and then it did get something confused or make an association that it shouldn't have, but honestly, it's only occasional. I think using them well is probably a bigger issue than their reliability.

 

This is where I object to what it's done to translation, i.e. AI techbros coming in and claiming it can outright replace language professionals. Well, no, it can't in high-end contexts, because you can't and won't ever be able to fully rely on its output. A human in the loop will always be needed, and good clients are already becoming aware of that. Far fewer humans than before, working at lower rates, though? Absolutely - and the low-hanging fruit is stuff it definitely can hoover up with a decent success rate that doesn't require too much human intervention to ensure plausibility. So I suppose that's the equivalent of what you're working with here - it essentially replicates a junior assistant who makes mistakes but is generally a lot of help. That's great, totally agreed. Although it's a bit frightening to think about what that means in terms of entry-level roles and career progression for younger people. 😬

[Christmas Tree 6688]

Police forces already in the UK trialling it as part of body cam to record witness statements ready for prosecution.

[Gemmill 63006]

They should try using the record function on their body cams. 

[Rayvin 8315]

1 hour ago, Meenzer said:

 

This is where I object to what it's done to translation, i.e. AI techbros coming in and claiming it can outright replace language professionals. Well, no, it can't in high-end contexts, because you can't and won't ever be able to fully rely on its output. A human in the loop will always be needed, and good clients are already becoming aware of that. Far fewer humans than before, working at lower rates, though? Absolutely - and the low-hanging fruit is stuff it definitely can hoover up with a decent success rate that doesn't require too much human intervention to ensure plausibility. So I suppose that's the equivalent of what you're working with here - it essentially replicates a junior assistant who makes mistakes but is generally a lot of help. That's great, totally agreed. Although it's a bit frightening to think about what that means in terms of entry-level roles and career progression for younger people. 😬

 

Yes I should be clear - it's not good for society at all that I was able to use it this successfully. It is absolutely replacing a secretary or some manner of assistant. And for $20/month.

 

You're right about translation mind, one of my good friends is a translator, as is his partner, and they're frequently expressing frustrations on LinkedIn around AI and people's desperation to use it. I agree completely it's not ready for such complex tasks. In fact, even for him I basically had to tell it to consistently review the transcripts whenever I wanted an answer, rather than review it's knowledge about the transcripts. Diverting it back to the source every single time is the only safe way of doing it IMO.

[Christmas Tree 6688]

Don’t be remember all this concern when driverless taxis reared its head, ya cunts!

[Renton 27742]

AI is very useful now. Like Rayvin, I've recently discovered it's power in taking transcripts which I can then interrogate and extract exactly who said what with a verbatim quote or summary format. Far better than me going from memory or scribbled notes.

 

For me it's got applications in helping me technically and for ideation too. But crucially, I think to use it safely and effectively you really need to know this stuff already so you know when it's hallucinating or just generally bull shitting to gain your favour. It can speed up medical writing. It can summarise and partially appraise studies. It can write me code. It can draw CT with a silly panama hat.

 

I think the social aspects of it and it's impact on junior staff in particular might be very serious. Coupled with remote working it's scary. It's deskilling potential is also frightening. That will affect all sorts of professionals. It steals IP which I don't have issue with unless this is used for commercial gain which of course it is.

 

I understand why people hate it. But it's here now, and is only going to grow in future years. We're all going to have to adapt to it.

 

Edited November 19 by Renton

[Meenzer 17990]

21 minutes ago, Christmas Tree said:

Don’t be remember all this concern when driverless taxis reared its head, ya cunts!

 

Taxi drivers are about the only thing Grok is less racist than.

[Andrew 7459]

2 hours ago, Meenzer said:

 

This is where I object to what it's done to translation, i.e. AI techbros coming in and claiming it can outright replace language professionals. Well, no, it can't in high-end contexts, because you can't and won't ever be able to fully rely on its output. A human in the loop will always be needed, and good clients are already becoming aware of that. Far fewer humans than before, working at lower rates, though? Absolutely - and the low-hanging fruit is stuff it definitely can hoover up with a decent success rate that doesn't require too much human intervention to ensure plausibility. So I suppose that's the equivalent of what you're working with here - it essentially replicates a junior assistant who makes mistakes but is generally a lot of help. That's great, totally agreed. Although it's a bit frightening to think about what that means in terms of entry-level roles and career progression for younger people. 😬

This is a problem I have with it in my industry, the uses of it is replacing both training (which is an incredibly good way to make yourself better at what you do by actually teaching others) but its also producing developers who are reliant on it in a way that isn't sustainable but also means they're 'writing' stuff they don't fully understand. 

 

There is, at some point, going to be an absolute black hole of mid-experience developers around imo, there'll just be the ones that were around before AI who may well still use it for stuff and the ones who were around since it started who's deeper understanding of their craft is practically non-existent.
 

Then that first lot will retire and we'll all be absolutely fucked if something breaks and the second group are the only ones around to fix it.

In its current form at least, as a tool in the hands of experts it is useful (but not truly necessary) and as a toy in the hands of the general public it is a nightmare.

[Renton 27742]

6 minutes ago, Andrew said:

This is a problem I have with it in my industry, the uses of it is replacing both training (which is an incredibly good way to make yourself better at what you do by actually teaching others) but its also producing developers who are reliant on it in a way that isn't sustainable but also means they're 'writing' stuff they don't fully understand. 

 

There is, at some point, going to be an absolute black hole of mid-experience developers around imo, there'll just be the ones that were around before AI who may well still use it for stuff and the ones who were around since it started who's deeper understanding of their craft is practically non-existent.
 

Then that first lot will retire and we'll all be absolutely fucked if something breaks and the second group are the only ones around to fix it.

In its current form at least, as a tool in the hands of experts it is useful (but not truly necessary) and as a toy in the hands of the general public it is a nightmare.

 

Yeah, I'd agree with this 100%.

In my field, when I graduated, it was pre internet and writing a page of a journal reuured a days physical browsing and cross referencing in a library. The internet was transformative, but you still had to understand your stuff. This feels different, like AI can be used to mask ignorance in any field. Its effect on education is going to be profound and ultimately very negative I feel.

[Rayvin 8315]

I'd agree with that too. Frankly I feel like it could replace me at this point, but then I suppose what it can't quite do is really appreciate context or nuance unless it is spoonfed, and knowing what to spoonfeed it is half the trick. Also the world isn't ready to trust it blindly yet.

 

But in the end, it's going to hammer consultants. Just hope I've moved on by then.

[Rayvin 8315]

Should also mention I'm using it to eliminate various parts of the academic publishing process in my little publishing company so it's also going to devastate copyediting, typesetting, all roles that are devoted to skilled book production. It's still reliant on human intervention but it eliminates the skill gap between me and a professional who's been at this for 20 years.