I, ChatGPT
-
-
The "Things that remind you of WTDWTF members" thread is
-
@Zerosquare said in I, ChatGPT:
The "Things that remind you of WTDWTF members" thread is
Absolutely.
"To test how rich the variety of ChatGPT’s jokes is, we asked it to tell a joke a thousand times," they write. "All responses were grammatically correct. Almost all outputs contained exactly one joke.
.
-
@Zerosquare said in I, ChatGPT:
The "Things that remind you of WTDWTF members" thread is
-
@TimeBandit makes perfect sense for what it is.
-
@Zecc said in I, ChatGPT:
"To test how rich the variety of ChatGPT’s jokes is, we asked it to tell a joke a thousand times," they write. "All responses were grammatically correct. Almost all outputs contained exactly one joke.
.
"Insanity is doing the same thing over and over and expecting different results."
-
This post is deleted!
-
@remi Strange. It works with dices. Or with lottery. Or with roulette.
-
@BernieTheBernie The insanity in that case is expecting to win.
-
Being productive means you're doing more hard work because the easy work is automated you nimrods!
-
@DogsB said in I, ChatGPT:
Being productive means you're doing more hard work because the easy work is automated you nimrods!
Maybe but the initial problem in TFA isn't that, but that people are apparently surprised that using AI to create stuff is increasing quantity over quality.
TFA also ends up discussing solving problems with AI that are caused by AI, and I think using a problem to "solve" itself seems like new territory of stupidity.
-
@loopback0 said in I, ChatGPT:
TFA also ends up discussing solving problems with AI that are caused by AI, and I think using a problem to "solve" itself seems like new territory of stupidity.
Really though, that's just bog standard hammer-nail principle.
-
@loopback0 said in I, ChatGPT:
I think using a problem to "solve" itself seems like new territory of stupidity.
Unfortunately no. This is a very old human tradition.
-
-
-
@GOG said in I, ChatGPT:
@DogsB said in I, ChatGPT:
That was totally unexpected.
-
-
-
@DogsB ahh, good, so it is now at an aggregated 100% right. Awesome!
-
@DogsB said in I, ChatGPT:
Over the course of the study researchers found that in March GPT-4 was able to correctly identify that the number 17077 is a prime number 97.6% of the times it was asked. But just three months later, its accuracy plummeted a lowly 2.4%. Meanwhile, the GPT-3.5 model had virtually the opposite trajectory. The March version got the answer to the same question right just 7.4% of the time—while the June version was consistently right, answering correctly 86.8% of the time.
What an idiotic test. So you ask the thing the same yes-or-no question many times and it gives inconsistent answers, because it's basically
bluffingrepeating hearsay, and when that number changes you conclude it's got worse at doing math. But it never did any math.
-
@topspin Followed to the actual "paper". They appear to really have asked the same question. With the same number.
Filed under: This paper is a waste of both the reader's and the author's time.
-
@cvi Not waste ofthe author's time: he got a new item to put on his list of publications, and gets citations, too.
-
@BernieTheBernie said in I, ChatGPT:
@cvi Not waste ofthe author's time: he got a new item to put on his list of publications, and gets citations, too.
Plus now he’s an AI Researcher!
-
-
When you want to have a LLM tell you how to build a bomb, you just need to know a couple of special characters and how to place them into your prompt.
-
@BernieTheBernie And thus we, once again, discover that the Halting Problem exists.
-
@GOG It's also a concrete example of Asimov's stories (e.g., bypassing the three laws).
-
@BernieTheBernie said in I, ChatGPT:
Strange. It works with dices.
I just tried a quick sample, but only got 6 different results.
-
@dkf get better dices?
-
-
@BernieTheBernie said in I, ChatGPT:
It works with dices.
It slices! It dices! It makes julienne fries! But wait, there's more!
-
@HardwareGeek but is it a combination hookah and coffee maker?
-
@Arantor
reminds me of this dubbel use of the dubbel deep fryer
or if anyone speaks Dutch and likes 90s skids
https://www.youtube.com/watch?v=9awfPymLI4U&ab_channel=MrRamazanLale2Definitely 90ties since they also promoted Windows 98 with L&H Speech control but since that was a Flemish company located in the far west side of the country it could only be used you spoke the right dialect.
https://www.youtube.com/watch?v=sx-tTMIByb8&ab_channel=MVDSome serious going on there with these big ass CRTs ...
-
@Luhmann said in I, ChatGPT:
Some serious going on there with these big ass CRTs ...
Still have one (17") in my garage...
-
@dcon said in I, ChatGPT:
Still have one (17")
-
-
I'd say good, but they'll just get bought out for pennies by Microsoft or Google or similar and get integrated but otherwise carry on.
-
@loopback0 said in I, ChatGPT:
I'd say good, but they'll just get bought out for pennies by Microsoft or Google or similar and get integrated but otherwise carry on.
How can they “edge closer to bankruptcy”?
Microsoft has already poured billions into it.
-
@topspin said in I, ChatGPT:
How can they “edge closer to bankruptcy”?
Microsoft has already poured billions into it.It's hosted on Azure, so they already paid back the billions to Microsoft for hosting
-
@topspin said in I, ChatGPT:
@loopback0 said in I, ChatGPT:
I'd say good, but they'll just get bought out for pennies by Microsoft or Google or similar and get integrated but otherwise carry on.
How can they “edge closer to bankruptcy”?
Microsoft has already poured billions into it.TFA says ChatGPT costs $700k to run. Per day.
-
@loopback0 said in I, ChatGPT:
@topspin said in I, ChatGPT:
@loopback0 said in I, ChatGPT:
I'd say good, but they'll just get bought out for pennies by Microsoft or Google or similar and get integrated but otherwise carry on.
How can they “edge closer to bankruptcy”?
Microsoft has already poured billions into it.TFA says ChatGPT costs $700k to run. Per day.
That’s only a quarter billion per year. Not billions.
You couldn’t even 3d model Zuck’s legs for that .
-
@topspin the later Microsoft investment was over many years. They won't have just chucked all of those billions in already. Now they won't need to when OpenAI goes bankrupt and they scoop up the remaining 51%.
-
@loopback0 said in I, ChatGPT:
@topspin the later Microsoft investment was over many years. They won't have just chucked all of those billions in already. Now they won't need to when OpenAI goes bankrupt and they scoop up the remaining 51%.
OTOH, what's stopping them from flipping a switch and making access paid-only? Yeah, they'd lose a lot of "customers", but I doubt it be all of them and they'd still have the money from Microsoft et al..
-
@cvi said in I, ChatGPT:
@loopback0 said in I, ChatGPT:
@topspin the later Microsoft investment was over many years. They won't have just chucked all of those billions in already. Now they won't need to when OpenAI goes bankrupt and they scoop up the remaining 51%.
OTOH, what's stopping them from flipping a switch and making access paid-only? Yeah, they'd lose a lot of "customers", but I doubt it be all of them and they'd still have the money from Microsoft et al..
But their engagement and exposure!
-
@loopback0 said in I, ChatGPT:
@topspin said in I, ChatGPT:
@loopback0 said in I, ChatGPT:
I'd say good, but they'll just get bought out for pennies by Microsoft or Google or similar and get integrated but otherwise carry on.
How can they “edge closer to bankruptcy”?
Microsoft has already poured billions into it.TFA says ChatGPT costs $700k to run. Per day.
To run.
Note that training is far more ressource intensive. By orders of magnitude.
-
-
Has the register taken leave of their senses with that headline? Will!
-
@DogsB said in I, ChatGPT:
Has the register taken leave of their senses with that headline? Will!
That's what the actual policy says. Will.
But also this:Extracting Data. Unless explicitly permitted, you may not use web scraping, web harvesting, or web data extraction methods to extract data from the AI services.
You fucking horsepissful moron clowns made this entire thing by doing just that!
-
@Applied-Mediocrity quod licet Iovi, non licet bovi.
-
@topspin said in I, ChatGPT:
@loopback0 said in I, ChatGPT:
@topspin said in I, ChatGPT:
@loopback0 said in I, ChatGPT:
I'd say good, but they'll just get bought out for pennies by Microsoft or Google or similar and get integrated but otherwise carry on.
How can they “edge closer to bankruptcy”?
Microsoft has already poured billions into it.TFA says ChatGPT costs $700k to run. Per day.
That’s only a quarter billion per year. Not billions.
You couldn’t even 3d model Zuck’s legs for that .That's just operating it. The real cost seems to be their training methods. I'm sure they're scrambling to figure out how to recover from all the damage they've done to its "accuracy."
Not to mention the next bigger version.