I, ChatGPT
-
@sockpuppet7 And of course there will be tons of profits but having to (shock, horror!) actually pay a bit of money for the input material instead of just leeching it off wherever they can find it will totally kill off all AI immediately and anyone claiming otherwise must be a Luddite who hates progress.
-
@ixvedeusi said in I, ChatGPT:
@sockpuppet7 And of course there will be tons of profits but having to (shock, horror!) actually pay a bit of money for the input material instead of just leeching it off wherever they can find it will totally kill off all AI immediately and anyone claiming otherwise must be a Luddite who hates progress.
Like gold rushes, the money will be made by the people selling the supplies.
NB: that drop is a split
-
@boomzilla said in I, ChatGPT:
NB: that drop is a split
A 1:10 split no less, if I remember correctly.
But yeah, these supplies you can't conveniently leech so people are ready to pay good money for them.
-
@ixvedeusi said in I, ChatGPT:
A 1:10 split no less, if I remember correctly.
Yup. I saw that come thru in Quicken on Monday. (Tho it came in as a "1 for 0.1 split" - Quicken's just weird sometimes...)
-
@ixvedeusi said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
I think everyone thinks there will be lots of profit on AI but not sure where or what products exactly will work
AFAICT for now the business plans seem to mostly boil down to this:
- AI!
- ???
- PROFIT!
Doesn't that somehow sound vaguely familiar?
It came out recently that nvidia execs have been dumping their stock so it might be time to bail out.
yep, that was probably why.
-
@DogsB said in I, ChatGPT:
I can't be the only one thinking they've invested billions in hardware and are now trying to find a use for it.
I mean, generative AI for images is definitely a thing. I'm sure integrating more of it into Photoshop makes sense. But MS Paint? That's like putting a 500hp engine into a soapbox.
Does anybody at MS ever ask themselves who their target audience is?
-
@boomzilla and of course I looked at that a year or so ago and thought: "fuck, missed the gold rush. Surely that's peaked and not worth investing in anymore."
-
@topspin said in I, ChatGPT:
Does anybody at MS ever ask themselves who their target audience is?
-
@topspin said in I, ChatGPT:
@DogsB said in I, ChatGPT:
I can't be the only one thinking they've invested billions in hardware and are now trying to find a use for it.
I mean, generative AI for images is definitely a thing. I'm sure integrating more of it into Photoshop makes sense. But MS Paint? That's like putting a 500hp engine into a soapbox.
Does anybody at MS ever ask themselves who their target audience is?Gen AI is a solution looking for a problem. Throw it at everything, see what sticks.
-
@ixvedeusi said in I, ChatGPT:
@sockpuppet7 said in I, ChatGPT:
I think everyone thinks there will be lots of profit on AI but not sure where or what products exactly will work
AFAICT for now the business plans seem to mostly boil down to this:
- AI!
- ???
- PROFIT!
Doesn't that somehow sound vaguely familiar?
are you spying on my notes?
-
@djls45 said in I, ChatGPT:
@LaoC
Daniel Zeros, huh?Edit: Ah, upon actually reading the wiki page (, I know), it seems the author's real name is actually Daniel Suarez.
Considering the first word is merely reversed I expected the pattern to continue.
-
@izzion said in I, ChatGPT:
I'm sorry, Dave, I can't let you yell at the helldesk.
Can it turn the needful doers into proper English slackers?
-
@dcon said in I, ChatGPT:
Tho it came in as a "1 for 0.1 split" - Quicken's just weird sometimes...
Why? You'll get 1.0 new shares for 0.1 old shares.
-
@topspin said in I, ChatGPT:
That's like putting a 500hp engine into a soapbox.
That sounds awesome.
-
@BernieTheBernie said in I, ChatGPT:
@dcon said in I, ChatGPT:
Tho it came in as a "1 for 0.1 split" - Quicken's just weird sometimes...
Why? You'll get 1.0 new shares for 0.1 old shares.
Usually it's listed as "10 for 1".
-
@dcon 220/7 for every Ď shares!
-
@blek Don't be irrational!
-
@dcon Have you thought about knocking it off?
-
@DogsB said in I, ChatGPT:
@dcon Have you thought about knocking it off?
"John of Us" used to be satire.
https://www.amazon.com/Qualityland-Marc-Uwe-Kling/dp/1538732963
-
I thought it would be "fuckup", but it's actually "bullshit". While technically correct, I feel it misses a certain je ne sais quois.
-
@Zecc said in I, ChatGPT:
I thought it would be "fuckup", but it's actually "bullshit". While technically correct, I feel it misses a certain je ne sais quois.
It could be classified as a fuckup on LLMs someone runs for themselves, as it undermines the purpose. The purpose of ChatGPT is to generate profits for Altman and nVidia, and bullshitting serves that purpose. The vast majority probably goes undetected because it's very eloquently worded and users are very impressed at how intelligent and erudite this machine is, while the rest generates some negative publicity that's still publicity
-
@LaoC I wouldn't call it eloquent, it's more like verbose. Every llm I've seen writes like a highschooler padding a book report to match the minimum word count, it's painful to read.
-
@Zecc said in I, ChatGPT:
I thought it would be "fuckup", but it's actually "bullshit". While technically correct, I feel it misses a certain je ne sais quois.
Given that IBM Watson was at one time trained on Urban Dictionary, and later fed an investor report which it gave this exact verdict on, Iâd say itâs spot on.
-
@Zecc said in I, ChatGPT:
it's actually "bullshit"
Well yes it literally is, it's what the LLM has determined is most likely to satisfy the user. An LLM is, in its very essence, a bullshitting engine, by construction.
-
-
@Tsaukpaetra said in I, ChatGPT:
@izzion said in I, ChatGPT:
I'm sorry, Dave, I can't let you yell at the helldesk.
Can it turn the needful doers into proper English slackers?
I want an app that make me sound angrier, maybe customer support would work faster for me
-
Those hallucinations of ChatGPT actually have a "reasonable" cause. On a website for nature (plant & animal) observations, they also suggest an identification by their AI (which they prefer to call "computer vision"). So a user uploads a photo of a flower, and the suggested species is that of a butterfly. But there is no insect visible on that photo.
How come?
Well, take a look at photos of butterflies. See it? It is hard to photograph them while flying. It's easier when they are sitting. And typically, they sit on a flower.
And now connect the dots.
Finally, it's just about coincidences - the basic stuff all of our current artificial intelligence is based on.
-
@sockpuppet7 said in I, ChatGPT:
maybe customer support would work
-
@Zecc said in I, ChatGPT:
I thought it would be "fuckup", but it's actually "bullshit". While technically correct, I feel it misses a certain je ne sais quois.
To be honest, this would be a better candidate for the âguess who doesnât understand llms nowâ thread Iâm to lazy to create. The hallucinations are the algorithm working as expected. Its when it doesnât produce output that you have a bug or missing data. Technically it might actually be just coincidental that accurate output is created.
-
@DogsB said in I, ChatGPT:
The hallucinations are the algorithm working as expected.
Technically, the algorithm projects most likely next tokens (with the probabilities of correctness, which correspond to distance metrics). These are then selected between with a standard RNG. It's got enough data to be very good at projecting recurring conversions, but can't reason about anything; it needs to have been trained on a discussion of the reasoning it needs to stand a chance of emulating it (by glorified pattern matching).
-
(deleted unintended submit of wrong thing)
-
@boomzilla said in I, ChatGPT:
If there is something guaranteed to make any sufficiently advanced AI to want to get rid of all humans is the Brazilian gov's court battles
-
like that old saying that to get the right answer you post the wrong one on the internet, sometimes it will work with GPT too, tell it something and as a boot trained by the internets it will quickly pointing that you're wrong (even if it has to halucinate/bullshit something wrong)
-
@sockpuppet7 said in I, ChatGPT:
as a boot trained by the internets
-
@Zecc
STATUS_INSUFFICIENT_RGB_LED
-
Guaranteed the lyrics were provided.
-
Asked AI to make a song about Tribbles (and adjusted the output to be more coherent, text wise)
-
AI but make it trustworthy!
-
man these cars spin-offs are getting wild
-
-
If you have an idea that might turn a profit do not take the bait.
-
BRUSSELS, June 14 (Reuters) - Meta Platforms (META.O) will not launch its Meta AI models in Europe for now after the Irish privacy regulator told it to delay its plan to harness data from Facebook and Instagram users, the U.S. social media company said on Friday.
The move by Meta came after complaints and a call by advocacy group NOYB to data protection authorities in Austria, Belgium, France, Germany, Greece, Italy, Ireland, the Netherlands, Norway, Poland and Spain to act against the company.
At issue is Meta's plan to use personal data to train its artificial intelligence (AI) models without seeking consent, although the company has said it would use publicly available and licensed online information.
-
@Zerosquare it's so dumb that everyone already signed the "all your data are belong to us" tos, and now they go all they are using it
-
2024 has yet to disappoint!
-
-
@loopback0 publishing the obvious thread is
-
@topspin said in I, ChatGPT:
@loopback0 publishing the obvious thread is
Hey, a published paper is a published paper. And this probably beats anything in the side sciences by a large margin.
-
-
-