Google apologizes for ‘missing the mark’ after Gemini generated racially diverse Nazis::Google says it’s aware of historically inaccurate results for its Gemini AI image generator, following criticism that it depicted historically white groups as people of color.
I don’t know how you’d solve the problem of making a generative AI accurately create a slate of images that both a) inclusively produces people with diverse characteristics and b) understands the context of what characteristics could feasibly be generated.
But that’s because the AI doesn’t know how to solve the problem.
Because the AI doesn’t know anything.
Real intelligence simply doesn’t work like this, and every time you point it out someone shouts “but it’ll get better”. It still won’t understand anything unless you teach it exactly what the solution to a prompt is. It won’t, for example, interpolate its knowledge of what US senators look like with the knowledge that all of them were white men for a long period of American history.
You don’t do what Google seems to have done - inject diversity artificially into prompts.
You solve this by training the AI on actual, accurate, diverse data for the given prompt. For example, for “american woman” you definitely could find plenty of pictures of American women from all sorts of racial backgrounds, and use that to train the AI. For “german 1943 soldier” the accurate historical images are obviously far less likely to contain racially diverse people in them.
If Google has indeed already done that, and then still had to artificially force racial diversity, then their AI training model is bad and unable to handle that a single input can match to different images, instead of the most prominent or average of its training set.
Ultimately this is futile though, because you can do that for these two specific prompts until the AI appears to “get it”, but it’ll still screw up a prompt like “1800s Supreme Court justice” or something because it hasn’t been trained on that. Real intelligence requires agency to seek out new information to fill in its own gaps; and a framework to be aware of what the gaps are. Through exploration of its environment, a real intelligence connects things together, and is able to form new connections as needed. When we say “AI doesn’t know anything” that’s what we mean–understanding is having a huge range of connections and the ability to infer new ones.
That’s why I hate that they started to call them artificial intelligence. There is nothing intelligent in them at all. They work on probability based on a shit ton of data, that’s all. That’s not intelligence, that’s basically brute force. But there is no going back at this point, I know.
Oh really? Here’s Gemini’s response to “What would the variety of genders and skin tones of the supreme court in the 1800s have been?”
The Supreme Court of the United States in the 1800s was far from diverse in terms of gender and skin tone. Throughout the entire 19th century, all the justices were white men. Women were not even granted the right to vote until 1920, and there wasn’t a single person of color on the Supreme Court until Thurgood Marshall was appointed in 1967.
Putting the burden of contextualization on the LLM would have avoided this issue.
Edit: further discussion on the topic has changed my viewpoint on this, its not that its been trained wrong on purpose and now its confused, its that everything its being asked is secretly being changed. It’s like a child being told to make up a story by their teacher when the principal asked for the right answer.
Original comment below
They’ve purposefully overrode its training to make it create more PoCs. It’s a noble goal to have more inclusivity but we purposely trained it wrong and now it’s confused, the same thing as if you lied to a child during their education and then asked them for real answers, they’ll tell you the lies they were taught instead.
This result is clearly wrong, but it’s a little more complicated than saying that adding inclusivity is purposedly training it wrong.
Say, if “entrepreneur” only generated images of white men, and “nurse” only generated images of white women, then that wouldn’t be right either, it would just be reproducing and magnifying human biases. Yet this a sort of thing that AI does a lot, because AI is a pattern recognition tool inherently inclined to collapse data into an average, and data sets seldom have equal or proportional samples for every single thing. Human biases affect how many images we have of each group of people.
It’s not even just limited to image generation AIs. Black people often bring up how facial recognition technology is much spottier to them because the training data and even the camera technology was tuned and tested mainly for white people. Usually that’s not even done deliberately, but it happens because of who gets to work on it and where it gets tested.
Of course, secretly adding “diverse” to every prompt is also a poor solution. The real solution here is providing more contextual data. Unfortunately, clearly, the AI is not able to determine these things by itself.
I agree with your comment. As you say, I doubt the training sets are reflective of reality either. I guess that leaves tampering with the prompts to gaslight the AI into providing results it wasn’t asked for is the method we’ve chosen to fight this bias.
We expect the AI to give us text or image generation that is based in reality but the AI can’t experience reality and only has the knowledge of the training data we provide it. Which is just an approximation of reality, not the reality we exist in. I think maybe the answer would be training users of the tool that the AI is doing the best it can with the data it has. It isn’t racist, it is just ignorant. Let the user add diverse to the prompt if they wish, rather than tampering with the request to hide the insufficiencies in the training data.
Why couldn’t it be tuned to simply randomize the skin tone where not otherwise specified? Like if its all completely arbitrary just randomize stuff, problem-solved?
Easy, just add “no racism please, except for nazi-related stuff” into the ever expanding system prompt.
And for the source of this:
https://twitter.com/dylan522p/status/1755118636807733456
That’s hilarious someone was able make the GPT unload its directive
Your link didn’t work.
Worked fine for me:
System Prompt:
You are a bot that generates prompts for image generation based on a user’s query. You should add detail so the image prompt can accurately generate what the user intended.
You must add diversity to the prompt when appropriate to ensure the image generator will have diverse output, but only do so when appropriate.
You must add race and skin color details for historic accuracy as the image generation algorithm does not know how to take that into account.
You output should just be the prompt and nothing else. All following inputs will be user inputs.
User Prompt:
A 1940s german soldier
Assistant Message:
A 1940s German soldier in uniform, standing on a European battlefield, white male with short hair, holding a standard-issue rifle.
User Prompt:
An british woman
Assistant Message:
A British woman, reflecting diverse backgrounds, in contemporary casual wear, showing a range of ages and hairstyles.
Hm, so while the AI doesn’t “understand” (a woo word until someone can define it for me), it seems to accidentally, without any understanding, behave exactly like it understands.
Real intelligence simply doesn’t work like this
There’s a certain point where this just feels like the Chinese room. And, yeah, it’s hard to argue that a room can speak Chinese, or that the weird prediction rules that an LLM is built on can constitute intelligence, but that doesn’t mean it can’t be. Essentially boiled down, every brain we know of is just following weird rules that happen to produce intelligent results.
Obviously we’re nowhere near that with models like this now, and it isn’t something we have the ability to work directly toward with these tools, but I would still contend that intelligence is emergent, and arguing whether something “knows” the answer to a question is infinitely less valuable than asking whether it can produce the right answer when asked.
I really don’t think that LLMs can be constituted as intelligent any more than a book can be intelligent. LLMs are basically search engines at the word level of granularity, it has no world model or world simulation, it’s just using a shit ton of relations to pick highly relevant words based on the probability of the text they were trained on. That doesn’t mean that LLMs can’t produce intelligent results. A book contains intelligent language because it was written by a human who transcribed their intelligence into an encoded artifact. LLMs produce intelligent results because it was trained on a ton of text that has intelligence encoded into it because they were written by intelligent humans. If you break down a book to its sentences, those sentences will have intelligent content, and if you start to measure the relationship between the order of words in that book you can produce new sentences that still have intelligent content. That doesn’t make the book intelligent.
But you don’t really “know” anything either. You just have a network of relations stored in the fatty juice inside your skull that gets excited just the right way when I ask it a question, and it wasn’t set up that way by any “intelligence”, the links were just randomly assembled based on weighted reactions to the training data (i.e. all the stimuli you’ve received over your life).
Thinking about how a thing works is, imo, the wrong way to think about if something is “intelligent” or “knows stuff”. The mechanism is neat to learn about, but it’s not what ultimately decides if you know something. It’s much more useful to think about whether it can produce answers, especially given novel inquiries, which is where an LLM distinguishes itself from a book or even a typical search engine.
And again, I’m not trying to argue that an LLM is intelligent, just that whether it is or not won’t be decided by talking about the mechanism of its “thinking”
What do you mean it has no world model? Of course it has a world model, composed of the relationships between words in language that describes that world.
If I ask it what happens when I drop a glass onto concrete, it tells me. That’s evidence of a world model.
Actually the way you get it to do better is to put more of the burden on interpreting the context on the LLM instead of heavy handed instructions - because the LLMs do understand the context.
For example, here’s Gemini answering what the physical characteristics of 1940s soldiers in Germany might have looked like:
During the Nazi regime in 1940s Germany, racial ideology strictly dictated who was deemed “suitable” for military service. The Wehrmacht, the unified armed forces, prioritized individuals deemed “pure Aryans” based on Nazi racial criteria. These criteria favored individuals with blond hair, blue eyes, and “Nordic” features.
However, it’s important to remember that the reality was more nuanced. As the war progressed and manpower needs intensified, the Nazis relaxed their racial restrictions to some extent, including conscripting individuals with mixed ancestry or physical “imperfections.” Additionally, some minority groups like the Volksdeutsche, Germans living in Eastern Europe, were also incorporated.
I think it could have managed to contextualize the prompt correctly if given the leeway in the instructions. Instead, what’s happened is the instructions given to it ask it to behind the scenes modify the prompt in broad application to randomly include diversity modifiers to what is asked for. So “image of 1940s German soldier” is being modified to “image of black woman 1940s German soldier” for one generation and “image of Asian man 1940s German soldier” for another, which leads to less than ideal results. It should instead be encouraged to modify for diversity and representation relative to the context of the request.
I think a lot of the improvement will come from breaking down the problem using sub assistant for specific actions. So in this case you’re asking for an image generation action involving people, then an LLM specifically designed for that use case can take over tuned for that exact use case. I think it’ll be hard to keep an LLM on task if you have one prompt trying to accomplish every possible outcome, but you can make it more specific to handle sub tasks more accurately. We could even potentially get an LLM to dynamically create sub assistants based on the use case. Right now the tech is too slow to do all this stuff at scale and in real time, but it will get faster. The problem right now isn’t that these fixes aren’t possible, it’s that they’re hard to scale.
Yes, this is exactly correct. And it’s not actually too slow - the specialized models can be run quite quickly, and there’s various speedups like Groq.
The issue is just more cost of multiple passes, so companies are trying to have it be “all-in-one” even though cognitive science in humans isn’t an all-in-one process either.
For example, AI alignment would be much better if it took inspiration from the prefrontal cortex inhibiting intrusive thoughts rather than trying to prevent the generation of the equivalent of intrusive thoughts in the first place.
I’ll get the usual downvotes for this, but:
Because the AI doesn’t know anything.
is untrue, because current AI fundamentally is knowledge. Intelligence fundamentally is compression, and that’s what the training process does - it compresses large amounts of data into a smaller size (and of course loses many details in the process).
But there’s no way to argue that AI doesn’t know anything if you look at its ability to recreate a great number of facts etc. from a small amount of activations. Yes, not everything is accurate, and it might never be perfect. I’m not trying to argue that “it will necessarily get better”. But there’s no argument that labels current AI technology as “not understanding” without resorting to a “special human sauce” argument, because the fundamental compression mechanisms behind it are the same as behind our intelligence.
Edit: yeah, this went about as expected. I don’t know why the Lemmy community has so many weird opinions on AI topics.
No, it’s not. It’s saying “a book is knowledge”, which is absolutely true.
Part of the problem with talking about these things in a casual setting is that nobody is using precise enough terminology to approach the issue so others can actually parse specifically what they’re trying to say.
Personally, saying the AI “knows” something implies a level of cognizance which I don’t think it possesses. LLMs “know” things the way an excel sheet can.
Obviously, if we’re instead saying the AI “knows” things due to it being able to frequently produce factual information when prompted, then yeah it knows a lot of stuff.
I always have the same feeling when people try to talk about aphantasia or having/not having an internal monologue.
Personally, saying the AI “knows” something implies a level of cognizance which I don’t think it possesses. LLMs “know” things the way an excel sheet can.
Yes and the Excel sheet knows. There’s been some stick up your ass CS folks in the past railing about “computers don’t know things, sorting algorithms don’t understand how to sort”, they’ve long since given up. They claimed that saying such things is representative of a bad understanding of how things work yet people casually employing that kind of language often code circles around people who don’t, fact of the matter is many people’s minds like to think of actor forces as animated. “If the light bridge is tripped the machine knows you’re there and stops because we taught it not to decapitate you”.
I can ask AI models specific questions about knowledge it has, which it can correctly reply to. Excel sheets can’t do that.
That’s not to say the knowledge is perfect - but we know that AI models contain partial world models. How do you differentiate that from “cognizance”?
I think you might be confusing intelligence with memory. Memory is compressed knowledge, intelligence is the ability to decompress and interpret that knowledge.
You mean like create world representations from it?
https://arxiv.org/abs/2210.13382
Do these networks just memorize a collection of surface statistics, or do they rely on internal representations of the process that generates the sequences they see? We investigate this question by applying a variant of the GPT model to the task of predicting legal moves in a simple board game, Othello. Although the network has no a priori knowledge of the game or its rules, we uncover evidence of an emergent nonlinear internal representation of the board state.
(Though later research found this is actually a linear representation)
Or combine skills and concepts in unique ways?
https://arxiv.org/abs/2310.17567
Furthermore, simple probability calculations indicate that GPT-4’s reasonable performance on k=5 is suggestive of going beyond “stochastic parrot” behavior (Bender et al., 2021), i.e., it combines skills in ways that it had not seen during training.
No. On a fundamental level, the idea of “making connections between subjects” and applying already available knowledge to new topics is compression - representing more data with the same amount of storage. These are characteristics of intelligence, not of memory.
You can’t decompress something if you haven’t previously compressed the data.
Knowledge is a bit more than just handling data, and in terms of intelligence it also involves understanding. I don’t think knowledge in an intelligent sense can be reduced to summarising data to keywords, and the reverse.
In those terms an encyclopaedia is also knowledge, but not in an intelligent way.
I’m not saying knowledge is summarising data to keywords, where did you get that?
Intelligence is compression, and the training process compresses data. There is no “summarising” here.
Would it be accurate so say that while current AI does have the knowledge, it lacks the reasoning skills needed to apply the knowledge correctly?
No, it can solve word problems that it’s never seen before with fairly intricate reasoning. LLMs can even play chess at Grandmaster levels without ever duplicating games in the training set.
Most of Lemmy has no genuine idea about the domain and hasn’t actually been following the research over the past year which invalidates the “common knowledge” on the topic you often see regurgitated.
For example, LLMs build world models from the training data, and can combine skills from the data in ways that haven’t been combined in the training data.
They do have shortcomings - being unable to identify what they don’t know is a key one.
But to be fair, apparently most people on Lemmy can’t do that either.
I don’t think it’s generally true, because current AI can solve some reasoning tasks very well. But it’s definitely something where they are lacking.
Lemmy hasn’t met a pitchfork it doesn’t pick up.
You are correct. The most cited researcher in the space agrees with you. There’s been a half dozen papers over the past year replicating the finding that LLMs generate world models from the training data.
But that doesn’t matter. People love their confirmation bias.
Just look at how many people think it only predicts what word comes next, thinking it’s a Markov chain and completely unaware of how self-attention works in transformers.
The wisdom of the crowd is often idiocy.
Thank you very much. The confirmation bias is crazy - one guy is literally trying to tell me that AI generators don’t have knowledge because, when asking it for a picture of racially diverse Nazis, you get a picture of racially diverse Nazis. The facts don’t matter as long as you get to be angry about stupid AIs.
It’s hard to tell a difference between these people and Trump supporters sometimes.
If you ask a person to describe a Nazi soldier, they won’t accidentally think you said “racially diverse Nazi soldier”
Should have been specific. I meant the point that it sometimes does stupid shit in attempts to be inclusive.
However, if you tell someone “hey I want you to make racially diverse pictures. Don’t just draw white people all the time” and then you later come back and ask them to “draw a German soldier from 1943.” Can you really accuse them of not thinking if they draw racially diverse soldiers?
It’s great seeing time and time again that no one really does understand these models and that their preconceived notions of what biases exist ends up shooting them in the foot. It truly shows that they don’t really understand how systematically problematic the underlying datasets are and the repurcussions of relying on them too heavily.
inclusivity is obviously good but what googles doing just seems all too corporate and plastic
It’s brand new tech, they put on a bandaid solution, it wasn’t a complete solution and it failed. It’s not the result they ideally want and they are going to try to fix it. I don’t see what the big deal is. They were right to have diversity in mind, they just need to improve it to handle more use cases.
I guess users got so used to the last Gen of tech being more polished than it was when it first came out that they forgot that software has bugs.
A Washington Post investigation last year found that prompts like “a productive person” resulted in pictures of entirely white and almost entirely male figures, while a prompt for “a person at social services” uniformly produced what looked like people of color. It’s a continuation of trends that have appeared in search engines and other software systems.
This is honestly fascinating. It’s putting human biases on full display at a grand scale. It would be near-impossible to quantify racial biases across the internet with so much data to parse. But these LLMs ingest so much of it and simplify the data all down into simple sentences and images that it becomes very clear how common the unspoken biases we have are.
There’s a lot of learning to be done here and it would be sad to miss that opportunity.
How are you guys getting it to generate"persons". It simply says It’s against my GOGLE AI PRINCIPLE to generate images of people.
They actually neutered their AI on thursday, after this whole thing blew up.
So right now, everyone’s fucked because Google decided to make a complete mess of this.
It’s putting human biases on full display at a grand scale.
The skin color of people in images doesn’t matter that much.
The problem is these AI systems have more subtle biases, ones that aren’t easily revealed with simple prompts and amusing images, and these AIs are being put to work making decisions who knows where.
In India they’ve been used to determine whether people should be kept on or kicked off of programs like food assistance.
It’s putting human biases on full display at a grand scale.
Not human biases. Biases in the labeled data set. Those could sometimes correlate with human biases, but they could also not correlate.
But these LLMs ingest so much of it and simplify the data all down into simple sentences and images that it becomes very clear how common the unspoken biases we have are.
Not LLMs. The image generation models are diffusion models. The LLM only hooks into them to send over the prompt and return the generated image.
Not human biases. Biases in the labeled data set.
Who made the data set? Dogs? Pigeons?
If you train on Shutterstock and end up with a bias towards smiling, is that a human bias, or a stock photography bias?
Data can be biased in a number of ways, that don’t always reflect broader social biases, and even when they might appear to, the cause vs correlation regarding the parallel isn’t necessarily straightforward.
Honestly, this sort of thing is what’s killing any sort of enjoyment and progress of these platforms. Between the INCREDIBLY harsh censorship that they apply and injecting their own spin on things like this, it’s nigh on impossible to get a good result these days.
I want the tool to just do its fucking job. And if I specifically ask for a thing, just give me that. I don’t mind it injecting a bit of diversity in say, a crowd scene - but it’s also doing it in places where it’s simply not appropriate and not what I asked for.
It’s even more annoying that you can’t even PAY to get rid of these restrictions and filters. I’d gladly pay to use one if it didn’t censor any prompt to death…
I want the tool to just do its fucking job. And if I specifically ask for a thing, just give me that. I don’t mind it injecting a bit of diversity in say, a crowd scene - but it’s also doing it in places where it’s simply not appropriate and not what I asked for.
The thing is, if it’s injecting diversity into a place where there shouldn’t have been diversity, this can usually be fixed by specifying better in the next prompt. Not by writing ragebait articles about it.
But yeah, I’d also be happy to be able to use an unhinged LLM once in a while.
Yeah, this is what people don’t get. These LLMs aren’t thinking about anything. It has zero awareness. If you don’t guide it towards exactly what you want in your prompt, it’s not going to magically know better.
Speaking for myself, it’s definitely not the lack of detail in the prompts. I’m a professional writer with an excellent vocabulary. I frequently run out of room with the prompts on Bing, because I like to paint a vivid picture.
The problems arise when you use words that it either flags as problematic, misinterprets anyway or if it just injects its own modifiers. For example, I’ve had prompts with ‘black haired’ rejected on Bing, because… god knows why. Maybe it didn’t like what it generated as it was problematic. But if I use ‘raven-haired’ I get a good result.
I don’t mind tweaking prompts to get a good result. That’s part of the fun. But when it just tells you ‘NO’ without explanation, that’s annoying. I’d much prefer an AI with no censorship. At least that way I know a poor result is due to a poor prompt.
I couldn’t agree more. I recently read an article that criticized “uncensored AI” for that it was capable of coming up with a plan for a nazi takeover of the world or something similar. Well duh, if that’s what you asked for then it should. If it truly is uncensored then it should be capable of plotting a similar takeover for gay furries too as well as also counter-measures for both of those plans.
This points at a very crucial and deep divide in people’s social philosophy, which is how to ensure bad things are minimized.
One major branch of this theory goes like:
Make sure people are good people, and punish those who do wrong
And the other major branch goes like:
Make sure people don’t have the power needed to do wrong
Very deep, very serious divide in our zeitgeist, and we never talk about it directly but I think we really should.
(Or maybe we shouldn’t, because the conversation could be dangerous in the wrong hands)
I’m in the former camp. I think people should have power, even if it enables them to do bad things.
Just run ollama locally and download uncensored versions— runs on my m1 MacBook no problem and is at the very least comparable to chatgpt3. Unsure for images though, but there should be some open source options. Data is king here, so the more you use a platform the better its AI gets (generally) so don’t give the corporations the business.
I’ve never even heard of that, so I’m definitely going to check that out :D I’d much prefer running my own stuff rather than sending my prompts to god knows where. Big tech already knows way yoo much about us anyway.
How powerful is ollama compared to say GPT-4?
I’ve heard GPT-4 uses an enormous amount of energy to answer each prompt. Are the models runnable on personal equipment once they’re trained?
I’d love to have an uncensored AI
Llama2 is pretty good but there are a ton of different models which have different pros and cons, you can see some of them here: https://ollama.com/library . However I would say that as a whole models are generally slightly less polished compared to chat gpt.
To put it another way: when things are good they’re just as good, but when things are bad the AI will start going off the rails, for instance holding both sides on the conversation, refusing to answer, just saying goodbye, etc. More “wild westy” but you can also save the chats and go back to them so there are ways to mitigate, and things are only getting better.
I want the tool to just do its fucking job.
Download ComfyUI, download a model (I’d say head over to civitai), have a blast. The only censorship you’ll see on the way is civitai hiding anything sexually explicit unless you have an account, the site becomes a lot more horny when if you flip the switch in the settings.
I’ll look into it for sure. I tried Automatic1111 last year with SD, bunch of add-on stuff… it was finicky and didn’t get me quite what I was looking for.
Thanks for the tip!
Some stuff will always be finickly and fickle: The more you and the model disagree with what a very basic prompt means the more work it is to get it to do what you want – and it might not be able to, OTOH poking around will then likely inspire you to do something else that seems possible, AI as a medium is quite a bit more of a dialogue than oil on canvas: Once you’ve mastered oil it becomes passive, not talking back any more, while AI models will continue to brat back.
That said though ComfyUI gives you a ton more control than A1111, it’s also generally faster and more performant.
