is Chat-GPT refusing this image?
is Chat-GPT refusing this image?
I asked Chat-GPT aboout Elon's salute, just for fun, and it replied it could not analyse pictures and I needed to describe it to him. I had pictures analysed before, even traslating and explaining a cartoon joke. I tought it could be caused by the .webp format. So I retried the joke image and it worked fine. I uploaded a png of Elon and it went wrong. Can others repeat the experience or was it just a fluke? I am not angry just let down a little.
suggested it was fake, in part because it depicts a "controversial handgesture". asked to clarify, it names the roman salute.
to be honest elon using the nazi salute was probably not in the training data or high up on the probability chart.
As far as I know Elon although has beef with OpenAi.
to be honest elon using the nazi salute was probably not in the training data
That's not how it works. If the training data had to include specific people doing specific things it'd be useless. Instead, the training let's it infer individual image elements from different sources. It never had to be trained on orange pentagons to be able identify "orange" and "pentagon" from an image of an orange pentagon.
didnt want to suggest this was a response purpose-made for this image. still i find it worrying that instead of admitting it doesnt know this image/ cant tell if its fake or real, it instead directly jumps to "fake"
or high up on the probability chart.
Well well well turns out artificial “intelligence” isn’t so intelligent after all
lol thats actually very high up in the probability chart
Open AI is in it for the money. Musk has money.
I explained it for everyone. You're welcome.
Musk is on the board as I recall.
I fucking hate the term '''''roman''''' salute soooooooo much.
fwiw i tried the same thing a few hours after the photo came out and it told me the photo must be fake and i need to be careful about misinformation
To be fair....based on it knowing who it is, and what he's doing, if you hadn't seen it with your own eyes it doesn't sound plausible. Like if you heard that there was nazi salutes being done at the official presidential inauguration, you'd think "ah, this guys exaggerating. He probably made a small gesture that's being taken out of context"
Except no. He full on did a nazi salute. Enthusiastically at that. And then he thought "Oh, my bad, the people in the back probably couldn't see that. LET ME DO ANOTHER NAZI SALUTE FOR THE PEOPLE IN THE BACK BEHIND ME!"
And he did it A SECOND time!
It's not chatGPTs fault it's unable to discern absurdity from reality at this point. You can't blame the program for what life has become. I fully expect The Onion to publish an article over the next 4 years that just says "We give up. Fuck it. We can't make up crazier shit than NPR at this rate."
I am so glad I watched the video. Seeing his arm in motion really drives it home. (Skip ahead about 17 seconds to spare yourself the agony of hearing his voice.)
Talking about his “heart” was obviously only included as an incredibly weak attempt at an excuse for the salute. I’m sure there are apologists out there who have lapped it up, though.
He didn't just do another Nazi salute, he did it at the fucking flag.
I was this guy. People on here were calling the party of Trump nazis. (I'm not US and keep forgetting the names of you parties) I thought they were being stupidly accelerating. Now I'm convinced. The nazis themselves convinced me. Problem is that them publicly doing nazi stuff will not change anything. They could hang brown people in the capitol and it would change nothing.
Nothing and Noone has ever gaslit me more than AI's. Seriously.
The real mildly infuriating is wasting resources on countless Lemmy servers with drivel from chatgpt.
I'll give you that, and will keep it in mind.
This is very likely a discussion of censorship.
Edit: to be clear, chatgpt's censorship of musk talk. A lot of people get information from chatgpt now the same way they get information from any other internet source. I think it's an appropriate discussion.
Y'all need better AI :) The one I "use" gives a good argument in favor of the musk is a fascist interpretation
We would need to see the input too. You can coach an AI response easily.
First question:
And what you are replying to is the output from my second question:
What AI is this?
I use the assistant on kagi, model is Claude 3.5 Sonnet
Wow, that's actually a really impressive analysis
Why is use in quotes?
Probably because they don't use it seriously, just for crap like this.
Musk is a capitalist. OpenAI is owned by capitalists. Draw your conclusions from that.
🤯
maybe they have an elon filter
Not mildly.
the picture no I admit it can be very infuriating, Personally it got me cautiously worried. Chat GPT getting crashed when asking about the picture however is mildly. As its not an essential service and I can make my own mind, that was curiosity.
I imagine there's something in there working to prevent deepfakes.
ChatGPT is already speaking highly of the EO policy statements being put into place, starting on inauguration day. It can no longer be trusted.
It could never be trusted to start with. There is no intelligence to it, its a model based on society's media soup. Its very easy to manipulate by users depending on how you present your query, and yes it is also tweaked by open AI political views. I expect it to enshitify even faster than google but for now it can be useful in many ways as long as you are aware of whet you are interacting with.
Some of the models have access to the Internet, and it can process information very fast. As long as it is neutral, it can be "trusted" as far as, it is trying. So you can send it off on a journey to scrape web pages for you and speed a search 30fold. If you've used it long enough to know the difference between the hallucinations and mistakes, it can accelerate your work, your learning, your research. You know it's just regurgitating the Internet and you lens appropriately. You can ask about disparate products and have a table generated to fix the disparate values in their specs. You can have it search large datasets and find errors without having to scroll for 30 minutes. It can regurgitate general learning (that you cross-reference to be sure) and accelerate said learning.
Yes, there is no intelligence to it (neural nets are digital neurons, per se, but that doesn't mean intelligence), but it is a query-response system that has advantages versus using the current enshittified Internet directly.
Now that it is obviously biased and tainted, no form of it can be trusted. That's the difference.
probably either a network issue or content filter (bc Nazi salute).
AI was supposed to save us
No, it is supposed to be a weapon in the class war.. https://logicmag.io/failure/the-automation-charade/
Good read. Thanks.
Since when
There is no fun in nazi pictures, you are normalizing fascism, blocked and reported you
Normalizing fascism? This picture is at the very top of lemmy.world posted on midly infuriating... What are you reporting? oh you blocked me.. I wont know I guess.
If you can still see that comment you're not blocked. I think they deleted their account.Esit: im wrong about the blocking. Nothing changes for you when someone blocks you. It's not like reddit. It's more an ignore/mute button than a block. (Or a personal shadowban)
They definitely deleted thier account though lol
What you should have said was nothing.
This is not “normalizing fascism.” You have made the worst interpretation possible.
