Google explains Gemini’s “embarrassing” AI pictures of diverse Nazis
Google explains Gemini’s “embarrassing” AI pictures of diverse Nazis
Gemini will undergo testing before it can start generating people again.
It's a pretty extreme example but I can definitely see how an AI would fumble need to know historical context.
Calling it AI at all was a mistake, it has no thought capability, it just spits out what it was asked for like the world's dumbest genie.
Yeah. it's called Generative AI because all it does is generate random crap based on ingested material and questions. It's a fancy autocomplete and people are relying on it.
The term "AI" has a much broader meaning and use than the sci-fi "thinking machine" that people are interpeting it as. The term has been in use by scientists for many decades already and these generative image programs and LLMs definitely fit within it.
You are likely thinking of AGI, or artificial general intelligence. We don't have those yet, but these things aren't intended to be AGI so that's to be expected.
It’s probably not the base model. It’s probably the hidden prompt augmentation rules that everyone has been scrambling to add to these models.
Gemini, like Chat GPT, it it likely appending simple prompts with more detail behind the scenes so results come out more varied. For example, since the models regurgitate the content people promote on the internet, if you searched for “attractive person” last year, you’d probably get a white person 95% of the time. When something is over represented in mainstream media, gen AI will reflect that back.
Now, before something goes into the gen AI black box, it is secretly given racial and photo style modifiers so users get more diverse people and image styles unless the user manually specifies what they want to see.
Generally this works well, but it can backfire hilariously if you ask for it to draw a group of people who are famous to having a certain ethnicity.
The best part is that this was added in response to people complaining about it not generating diverse enough people, so they made it randomly add different races to person requests that don't specify race.
Why would anyone do that? Like just tell the AI you want diverse people yourself, making it the default just adds more moving parts that could fail(and you end up with diverse nazis or ryan gosling as black panther)
It's done because the underlying training data is heavily biased to begin with. It's been a known issue for along time with AI/ML, for example racist cameras have been an issue for decades https://petapixel.com/2010/01/22/racist-camera-phenomenon-explained-almost/.
So they do this to try to correct for biases in their training data. It's a terrible idea, and shows the rocky path forward for GenAI, but it's easier than actually fixing the problem ¯\_(ツ)_/¯
The funniest thing about this is that its like a parable for the modern paradox of tolerance.
It reminds me of that cartoon on the front page the other day where the Palestinians being bombed were talking about being a part of history because it was a brown woman who vetoed the ceasefire.
Wdym?
OpenAI was literally adding the word diverse into their images last year as well.
I remember getting an image once where it even wrote the word “diverse” into image, and AI only add random words pulled from the prompt.
Dunno why Google doing it now is suddenly big news?
New event happens, people react.
It’s gone woke! Abandon ship!
If I can't get my AI Nazified image fix, then what even is the point? /s
Tay did nothing wrong
Why doesn't the screenshot show the prompt but only its title? Unnecessarirly suspicious since it wouldn't even be hard to fake it.