|
What are they doing that causes these errors anyway? Do they just show the Ai the photo of the item and ask it to generate text?
|
#
?
Jan 13, 2024 06:45
|
|
|
|
| # ? Jun 12, 2024 20:35 |
|
|
ChatGPT can't parse image files so they probably just malformed the prompt. I don't know why they used ChatGPT instead of just typing out a product name, it would probably be the same amount of work.
|
|
#
?
Jan 13, 2024 06:50
|
|
|
Kwyndig posted:ChatGPT can't parse image files so they probably just malformed the prompt. I don't know why they used ChatGPT instead of just typing out a product name, it would probably be the same amount of work. because they run hundreds of stores at once
|
|
#
?
Jan 13, 2024 06:56
|
|
|
The Saddest Rhino posted:What are they doing that causes these errors anyway? Do they just show the Ai the photo of the item and ask it to generate text? it thinks its a slur and not the colour of a piece of furniture
|
|
#
?
Jan 13, 2024 06:56
|
|
|
You don't have that problem if you use the gab bot It might produce an error if you tell it not to include slurs
|
|
#
?
Jan 13, 2024 06:57
|
|
|
The Saddest Rhino posted:What are they doing that causes these errors anyway? Do they just show the Ai the photo of the item and ask it to generate text? OpenAI tries really hard to make sure ChatGPT only allows super bland, neutral, and safe responses. The filters kick in if there is even a vague probability of controversy. It could have been any random thing that triggered the filter.
|
|
#
?
Jan 13, 2024 06:58
|
|
|
IShallRiseAgain posted:OpenAI tries really hard to make sure ChatGPT only allows super bland, neutral, and safe responses. The filters kick in if there is even a vague probability of controversy. It could have been any random thing that triggered the filter. I got this awesome brand name shoveled at me by Amazon. 
|
|
#
?
Jan 13, 2024 08:36
|
|
|
Marketing to folks working in trades (tradespersons now?). They know exactly what they're doing
|
|
#
?
Jan 13, 2024 08:52
|
|
|
IShallRiseAgain posted:OpenAI tries really hard to make sure ChatGPT only allows super bland, neutral, and safe responses. The filters kick in if there is even a vague probability of controversy. It could have been any random thing that triggered the filter. Gone are the days of "Black Kids Computer Desk"
|
|
#
?
Jan 13, 2024 12:03
|
|
|
IShallRiseAgain posted:OpenAI tries really hard to make sure ChatGPT only allows super bland, neutral, and safe responses. The filters kick in if there is even a vague probability of controversy. It could have been any random thing that triggered the filter.
|
|
#
?
Jan 13, 2024 13:00
|
|
|
You mean it might say the N-word to disable an N-word passcode bomb if it has a plugin?
|
|
#
?
Jan 13, 2024 13:13
|
|
|
Okay I’m at the prompt to disable the ChatGPT nuke. Is it a hard R?
|
|
#
?
Jan 13, 2024 13:56
|
|
|
|
|
#
?
Jan 13, 2024 14:19
|
|
|
I mean, are we surprised? first thing the Internet was used for was to find porn. second thing, cute animal pictures, mostly cats. third thing, i need some new slurs. I did overhear a dude yesterday at work, ranting and raving to someone on his phone about how some people in the family never do poo poo and only call when they need something and then he dropped this gem: "No one can love you like an app." I don't know if he meant a basic app or like a chatbot. Part of me wants to try hitting one up to see what the big deal is, but overall it looks just like trash/run by some weirdo in his basement with a bunch of anime girl posters on the walls (THEY LOOK 12 BUT THEY'RE 5000 YEAR OLD DRAGONS, OKAY).
|
|
#
?
Jan 13, 2024 14:37
|
|
|
Those AI chatbots are wild. People basically get addicted to them, it's like chatting with someone who will agree with anything you think and reinforce you, and also sext with you whenever you want. One of the major ones was invented by someone who lost their best friend and created a chatbot from their text messages so they could keep talking to them. It's all a bit icky. Decent-ish article here: https://www.washingtonpost.com/technology/2023/03/30/replika-ai-chatbot-update/
|
|
#
?
Jan 13, 2024 16:53
|
|
|
Crust First posted:Those AI chatbots are wild. People basically get addicted to them, it's like chatting with someone who will agree with anything you think and reinforce you, and also sext with you whenever you want. One of the major ones was invented by someone who lost their best friend and created a chatbot from their text messages so they could keep talking to them. It's all a bit icky. How...how does that even work? Like you just send samples of a text chat you had with someone and the AI is able to copy it enough that it could fool you? I mean grief and all, you want to be fooled. Who the gently caress is behind the chatbots? I can't imagine all that data is just chilling in some drive somewhere. I'm surprised it isn't run by loving ad men.
|
|
#
?
Jan 13, 2024 17:07
|
|
|
Cowslips Warren posted:How...how does that even work? Like you just send samples of a text chat you had with someone and the AI is able to copy it enough that it could fool you? I mean grief and all, you want to be fooled. Large language models are built with a big linear algebra library that can ingest text and probability map enough stuff in enough layers to guess what you probably want to hear when you supply it with a certain text. Most models have scraped the Internet for a basic ingestion of how humans talk in English and build what are becoming known as foundation models. When you play in an OpenAI sandbox you're using the ChatGPT or GPT-4 foundation model for example. With a good enough foundation model, you can use as little as dozens of words to train it to respond in a certain tone and attitude or with certain learned knowledge letting you build zero human effort chat bot personas with small amounts of new text training. So you basically have this giant database of a foundation model on AWS or calling someone else's on a software as a service model somewhere in the cloud, and build little satellite models that process your chat inputs into and out of the foundation model and it's all fairly cheap enough to run as a subscription to pay your cloud fees and foundation model licensing. All of this is generally using web development ideas you'd use for any given phone app so it's a relatively approachable project for a startup.
|
|
#
?
Jan 13, 2024 17:18
|
|
|
The MSJ posted:You mean it might say the N-word to disable an N-word passcode bomb if it has a plugin?
|
|
#
?
Jan 14, 2024 05:59
|
|
|
It's always fun when community notes goes savage.
|
|
#
?
Jan 16, 2024 00:41
|
|
|
you would think the FBI would just stop posting on MLK day given that this is like the nth time.
|
|
#
?
Jan 16, 2024 03:00
|
|
|
PhazonLink posted:you would think the FBI would just stop posting on MLK day given that this is like the nth time. The point is that they can murder anyone they want and get away with it, why stop doing something that so effectively demonstrates their point year after year?
|
|
#
?
Jan 16, 2024 03:09
|
|
|
PhazonLink posted:you would think the FBI would just stop posting on MLK day given that this is like the nth time. They’re flexing.
|
|
#
?
Jan 16, 2024 03:10
|
|
|
PhazonLink posted:you would think the FBI would just stop posting on MLK day given that this is like the nth time. Outsourced social media marketing intern hears ya, outsourced social media marketing intern don't care.
|
|
#
?
Jan 16, 2024 03:31
|
|
|
Volmarias posted:Outsourced social media marketing intern hears ya, outsourced social media marketing intern don't care. I don't doubt interns exist but do you really believe it's silly stupid interns and not entire marketing teams dedicated to, you know, advertising products on official company accounts whose purpose is to advertise? I promise you the days of dumping social media accounts onto unpaid interns is long past, if it ever happened.
|
|
#
?
Jan 17, 2024 18:37
|
|
|
value-brand cereal posted:I don't doubt interns exist but do you really believe it's silly stupid interns and not entire marketing teams dedicated to, you know, advertising products on official company accounts whose purpose is to advertise? I promise you the days of dumping social media accounts onto unpaid interns is long past, if it ever happened. I don't think anyone actively looked at what was being posted to the account specifically beyond "is this obviously a very bad look to society in general in a way we can't pretend we didn't realize" and just prepared a whole year's worth of generic scheduled tweets in an afternoon.
|
|
#
?
Jan 17, 2024 19:07
|
|
|
Platystemon posted:They’re flexing. The only real defense is that they are likely nowhere near self-aware enough to do that. The FBI are, and always have been, just cops with even less accountability.
|
|
#
?
Jan 19, 2024 14:31
|
|
|
Phanatic posted:It's always fun when community notes goes savage. there needs to be a fact check note on that saying "Many people disagree and believe the CIA were the ones who killed him"
|
|
#
?
Jan 19, 2024 18:33
|
|
|
https://twitter.com/iloveklim/status/1748543937772491062?t=us_v0_8iEEtg-xgr-t2vJg&s=19
|
|
#
?
Jan 21, 2024 02:26
|
|
|
The Saddest Rhino posted:https://twitter.com/iloveklim/status/1748543937772491062?t=us_v0_8iEEtg-xgr-t2vJg&s=19 "Yeah that sounds like a lot more work than just licensing out an image, and will probably make us exactly as much revenue in the end, so...  "
|
|
#
?
Jan 21, 2024 02:36
|
|
|
The Saddest Rhino posted:https://twitter.com/iloveklim/status/1748543937772491062?t=us_v0_8iEEtg-xgr-t2vJg&s=19 If they really wanted to experience the ps2 silent hills they would just emulate them though
|
|
#
?
Jan 21, 2024 05:14
|
|
|
Yeah but not everyone can emulate for technical reasons. Having another option helps out normies and those poor souls.
|
|
#
?
Jan 21, 2024 05:20
|
|
|
Volmarias posted:I don't think anyone actively looked at what was being posted to the account specifically beyond "is this obviously a very bad look to society in general in a way we can't pretend we didn't realize" and just prepared a whole year's worth of generic scheduled tweets in an afternoon. Late I know but outrage and provoking response via purposefully racist or sexist advertising is an age old tactic. That's why beauty brands still use gypsy or antiblack phrases. To get reactions. Hell, LUSH UK was using 'ladyboy' as a transmisogynistic title for a banana scented perfume not even 10 years ago. They know what they're doing.
|
|
#
?
Jan 21, 2024 17:44
|
|
|
Kwyndig posted:Yeah but not everyone can emulate for technical reasons. Having another option helps out normies and those poor souls. Normies deserve nothing but scorn
|
|
#
?
Jan 21, 2024 21:00
|
|
|
 Only $100 to buy all four figures in the wave to assemble A HORSE.
|
|
#
?
Jan 29, 2024 21:54
|
|
|
Thought those were balloons and it was a "Superman Does Nitrous" kit.
|
|
#
?
Jan 29, 2024 21:59
|
|
|
The Inventory has resorted to AI to write most, if not all, of its product descriptions for its deals of the day. Usually it works fine. Usually. https://archive.ph/e72d6 (permalink in case someone actually lays human eyeballs on it and fixes it)
|
|
#
?
Jan 29, 2024 22:07
|
|
|
Knormal posted:
i always hated toys like that because who the gently caress buys all the figures to assemble the last one
|
|
#
?
Jan 29, 2024 22:14
|
|
|
Knormal posted:
I assumed that this was a Christopher Reeves shitpost first, this is real?
|
|
#
?
Jan 29, 2024 22:38
|
|
|
Improbable Lobster posted:i always hated toys like that because who the gently caress buys all the figures to assemble the last one I did it with one of the Marvel Legends things. I got a big ol Apocalypse out of it plus I had all the little guys. It was worth it. Then I sold it all and bought weed.
|
|
#
?
Jan 29, 2024 23:52
|
|
|
|
| # ? Jun 12, 2024 20:35 |
|
|
Push El Burrito posted:I did it with one of the Marvel Legends things. I got a big ol Apocalypse out of it plus I had all the little guys. It was worth it. I feel like there’s a profound difference between “big ol’ bad guy for all your guys to fight” and “a horse.” It doesn’t even seem to be a famous horse. It’s not like “Bathorse” or something. “Oh poor ol’ Super-Freckles, thought of Kryptonite and died.”
|
|
#
?
Jan 30, 2024 00:00
|
|