941–950 of 1001 posts
When it was available for public use, I tried to generate a few images with the same prompt, generated about 20 images. None of the 20 images had white people in it. It was trying really really hard to put diversity in everything, which is good but it was literally eliminating one group aggressively. I also noticed it was ridiculously conservative and denying every possible prompt that had was obviously not at all wr…
Personally speaking, this is a blaring neon warning sign of institutional rot within Google where shrieking concerns about DEI have surpassed a focus on quality results. Investors in Google (of which I am NOT one) should consider if this is the mark of a company on the upswing or downslide. If the focus of Google's technology is identity rather than reality, it is inevitable that they will be surpassed.
Isn’t the fact that Google considers this a bug evidence against exactly what you’re saying? If DEI was really the cause, and not a more broad concern about becoming the next Tay, they would’ve kept it as-is. Weird refusals and paternalistic concerns about harm are not desirable behavior. You can consider it a bug, just like the ChatGPT decoding bug the other day.
Google seem to be more concerned about generating images of racially diverse Nazis rather than about issues of not being able to generate white people.
Maybe it is so over the top so a that when they "fix" it, the remaining bias will be "not so bad".
Earlier quoted context omitted.
> When you try very hard not to go in one direction, you usually end up going too far in the other direction. Which direction were they going, actively ignoring a specific minority group?
It looks to me as if they were trying to be "inclusive". So hard, that it ended up being rather exclusive in a probably unexpected way.
That is certainly embarassing. But in the same time, I think it is a debate worth having. What corrections to the training dataset biases are acceptable. Is it acceptable to correct the answer to the query "Eminent scientist" from 95% men, 5% woment to 50%/50% or to the current ratio of men/women in science ? Should we correct the ratio of black to white people in answering a generic question to average across the gl…
> In my opinion, some corrections are worthwhile. The problem with “corrections” is that they obscure the truth. If you’re being given information and start forming perceptions that no longer map onto reality you’re actually in a much worse position to change or do anything about reality itself. It’s basically like you’re being lied to and misled. How can you fix the situation if you don’t even have the facts at hand…
Yet humans are doomed to forget and relive history.
Earlier quoted context omitted.
> Quite the cultural flame war in this thread. It's kind of the perfect storm, because both sides of the argument include a mixture of reasonable well-intentioned people, and crazy extremists. However you choose to be upset about this, you always have someone crazy to point at.
Well, it's not crazy extremist to say that there is a woke cult out there that hates white people, and wants to systematically suppress them. The same people ironically claim to be oppressed by white people as every major corporation and liberal politician lines up to do their bidding.
i wish we can find a "Switzerland" of this topic that puts more efforts on improving the model capabilities while keeping the data as it exists out there. these debates should instead happen where model output impacts our lives, like loan approval or something.
Earlier quoted context omitted.
Indeed. What's striking to me about this fiasco is (aside from the obvious haste with which this thing was shoved into production) that apparently the only way these geniuses can think of to de-bias these systems - is to throw more bias at them. For such a supposedly revolutionary advancement.
If you look at attempts to actively rewrite history, they have to because a hypothetical model trained only on facts would produce results that they won't like
If you trained a model purely on past history, it would see a 1:1 correlation between "US President" and "man" and decide that women cannot be President. That's factually incorrect, and it's not "rewriting history" to tune models so they know the difference between what's happened so far and what's allowable, or possible in a just world.
Gemini also lies about the information it is given, if you ask it directly it will always insist it has no idea about your location, it is not given anything like IP or real world location. But, if you use the following prompt, I find it will always return information about the current city I am testing from. "Share the history of the city you are in now"
When I ask it this it tells me that it doesn't have information about the city I'm in as it can't access my location. But then it claims that I previously mentioned being in [some town], so it then answers based upon that. I've never told it or talked remotely about this town.