The AI dooooooom thread

[Razgovory]

It does that if backed into a corner and it can't find a real answer, but it does tend to prefer real answers first.

It doesn't have preferences and doesn't know what reality is. It's just a "stochastic parrot" as per the Bender paper.

[Jacob]

Also, several people I work with are using it to draft emails - I can tell because of the language - which I find supercringe and slightly rude. Like, "huh, I'm not worth your time so you have chatgpt draft your emails to me?"

Yeah for sure, anyone using AI to address me immediately goes to the "you're not worth my time" bucket. I guess if the AI is good enough, I won't be able to tell but that will only make my rejection more firm if I do find out.

Same thing with robo-voices and recorded calls and the like - if I'm not worth your time to engage with, why should I use my time to engage with your business objectives?

Similarly, if I need something from your business and you make me go through AI assistants and automated systems you're signalling how important I am to you (answer: not at all). If I need something I'll have to continue (or decide it's not worth it) - and obviously you get what you pay for (more or less) - but it's definitely sending a powerful signal about our relationship and it's one I'll take note of when making future decisions.

[Josquius]

Same thing with robo-voices and recorded calls and the like - if I'm not worth your time to engage with, why should I use my time to engage with your business objectives?

Not just robo calls. Any call that hasn't been previously arranged.
You better be dying grandma!

I use a lot of dashes in regular writing - mostly because they are very prevalent in script-speak and it has kinda bled into my regular writing. Now I feel like people think AI is inserting them into my writing 

EDIT: Fuck, I used one, see?

That's a hyphen, not an em dash. Technically you used that wrong. Though as said myself and everyone else uses them for this purpose as they're right there on the keyboard whilst em dashes aren't.
It is a very easy way to spot AI that the correct dash is used. Though as mentioned I know some pretty anal people already made the effort to use the proper symbol and they aren't happy.

[garbon]

Microsoft products generally add them automatically but I agree because of how we fuck about with formatting, often people have a mix of both rather consistently just hyphens.

[Admiral Yi]

I watched part of the Cubs/White Sox game yesterday with AI generated images.  Don't know if the announcers were human.

[Zoupa]

[Josquius]

Yes. It is too much of a butt kisser. Really shows a limitation.
Reminds me of this one.

https://futurism.com/leaked-chatgpt-lawyer-displace-amazonian

Leaked ChatGPT Conversation Shows User Identified as Lawyer Asking How to "Displace a Small Amazonian Indigenous Community From Their Territories in Order to Build a Dam and a Hydroelectric Plant"

Imagining chat gpts jolly butt kissing reply "That sounds like a great scheme! You are absolutely right that these people don't understand the value of land and could arguably be termed non people!"

[Zanza]

Version 5 is supposedly much more restrained.

[celedhring]

Version 5 is supposedly much more restrained.

Yeah, sadly, it also seems restrained in its usefulness.

For my general use case (revising my English-language texts for grammar and more natural language), gpt4o is giving me better results... 5 tends to cut out too much text, even if I tell it to not do that. I feel it's been trained to give shorter responses?

[Zanza]

But you can buy a subscription for longer answers and more tokens. 

[Sheilbh]

Yes. It is too much of a butt kisser. Really shows a limitation.

I do slightly wonder how much of this is because they're American products with American optimism and customer service attitudes If it was built by, say, a Dutch company or one based in Berlin maybe it would have a different attitude.

Though I believe some of these providers are now giving you prebuilt "personalities" to interact with as not everyone wants enthusiastic, friendly, chatty.

[Josquius]

Yes. It is too much of a butt kisser. Really shows a limitation.

I do slightly wonder how much of this is because they're American products with American optimism and customer service attitudes If it was built by, say, a Dutch company or one based in Berlin maybe it would have a different attitude.

Though I believe some of these providers are now giving you prebuilt "personalities" to interact with as not everyone wants enthusiastic, friendly, chatty.

 
Great idea. Lets go seek investment for this AI startup based in Amsterdam.
"Look....Kevin? It was Kevin right? This idea of yours is shit. It would be a waste of processing power for me to think about it for more than a millisecond. You should do some basic reading."

[Valmy]

Oh yeah that would be amazing. German AI telling us why our ideas are shit.

[Zanza]

I just used GPT5 for the first time extensively to generate a concept for a topic I currently work on. Did not really notice much of a quality difference, but this "great question" bullshit is toned down.

[DGuller]

I used GPT 5 Pro at work to document my code.  It took forever to respond to each prompt, but it found a subtle but big mistake in my logic.  Incidentally I myself spotted it earlier that day, but I fed it the old version of the code to test it. 

It is pretty unnerving how much it figured out about what I was doing, without even having access to inputs that I would think would be crucial.  What I was doing wasn't boilerplate, it was a custom thing.  An expert would probably make educated guesses as to how the input is built, bit I wouldn't expect AI to be so on point.