yyaldrin

in je ogen waait de wind
i was reading in the van gogh letters the other day and you know how he is kind of a bum with no money so he lives of the little bits he gets sent from his brother. anyway one day he realises he needs new underwear and so he goes to the tailor and has himself measured up and orders the undergarments for him to pick it up a few days later. now i've always thought that doing something like that was only reserved for the aristocrats but apparently it was the way clothes were made and fabricated in those days. personalised and on request. something like that is now considered an absolute luxury. it's one of those things that makes you think how far we are alienated from the products we consume in the same way children don't know that a hamburger comes from an animal. i always thought the ludites burning down the mills was a bit of an overreaction but if you see in what way the textile and fashion industry has developed, you can only think that they were right.
 

shakahislop

Well-known member
i sent a journal article for internal review before submission, and the person who needs to approve it had clearly run it through chatGPT and given me the output. i had to change the article to respond to the computer's thoughts

i regularly receive emails now which are written by ChatGPT, that colleagues purport to have written themselves. i am being thanked by a computer, the thoughts are being framed and described by a computer. it's rude to point it out, so i am writing responses to what the computer has said, as though they are the thoughts of the person i'm emailing

i am getting slick professional documents that look like they are well thought through institutional publications, but they are simply chatGPT doing its thing. it's breaking that aspect of interpretation and communication, the inference that a document which looks professional is something that someone actually cares about and has put thought into

a layman suggested to me yesterday that some modelling that a university professor is doing should simply be done by AI instead

it's intervening in the world of ideas, via humans as intermediaries. the first time the computers have really done this i think.

at some point figuring out how to manipulate what AI is likely to say as a response to a query could become as important as gaming the youtube algo, gaming the google search results and so on
 

william_kent

Well-known member
at some point figuring out how to manipulate what AI is likely to say as a response to a query could become as important as gaming the youtube algo, gaming the google search results and so on

it's called "prompt injection", people are already trying to figure it out

Prompt injection involves manipulating model responses through specific inputs to alter its behavior, which can include bypassing safety measures. Jailbreaking is a form of prompt injection where the attacker provides inputs that cause the model to disregard its safety protocols entirely.
 

wektor

Well-known member
it's intervening in the world of ideas, via humans as intermediaries. the first time the computers have really done this i think.
we adapt, with written word, then print, eventually digital communication, each of these could be seen as a mutilation, yet somehow we usually end up twisting it to our own benefit

I am curious to see how prioritisation of certain aspects of communication will mutate, perhaps this so called professionalism we will not care about so much anymore soon.
A friend of mine claims that being capable of expressing elaborate thought, especially in other languages than your native, as to avoid machine translation and multiple other bottlenecks within the pipeline, could easily become something of that sort in the future.

Was working on a consulting project for a few days where someone sent me a brief that was three very obviously generated and non-redacted paragraphs. I ended up having to ask for a voice memo instead, as I am not willing to take the time to comprehend a potentially incomprehensible wall of text just because someone cannot spend one minute to formulate their thoughts properly, instead throwing a few keywords into a prompt.
 

Mr. Tea

Let's Talk About Ceps
It's a shame Ted Kazynch... Kachyns... the Unambomber shot his load a few decades early. I reckon he'd have been well up for bombing AI data centres.
 

william_kent

Well-known member
Do you think Altman is a psychopath?

this week: OpenAI wins $200m contract with US military for ‘warfighting’

chump change butI can't see this can't end well

openai will “develop prototype frontier AI capabilities to address critical national security challenges in both warfighting and enterprise domains”

also allegedly "R worded" his sister ( according to her ) when she was 6 years old ( I'm not a psychiatrist and she may be crazy, but a lot of child abuse survivors are mentally damaged, all in all it's a bit of a "she said, he said" situation )

also "Sam Altman's Eye-Scanning Orb Is Now Coming to the US"

so, yeah, psychopathic tech bro who can't code
 
Top