[pachorizons]

It is truly indicative of the shortcomings of the supposedly enlightened Hacker News hivemind that Google - a company that has killed influential projects and sacrificed / cannibalised the integrity of it's core search product - should be condemned because it's latest AI malfunction is too 'DEI'.

Google's AI blunders are more systemic than the latest culture war. The cringe hate spewed by earlier models are just as incompetent as Gemini, these are nothing more than the flailing of a multinational who is incapable of fashioning a vision of a world they promise. These are bugs at the highest level, that are clearly poisioning every layer of Google. They are the same bugs that led to Google search shortcuts that pushed disinformation COVID.

This is a failure of rigour from the flailing of the world's self appointed organiser of knowledge. Your least favourite DEI minority has nothing to do with it. For fucks sake. Place your blame where it belongs.

[fngjdflmdflg]

>supposedly enlightened

What made you suppose this?

>The cringe hate spewed by earlier models are just as incompetent as Gemini [...] Your least favourite DEI minority has nothing to do with it

Except unlike Tay ect. this would specifically add "of various diverse ethnicities" to each promt, so this isn't just some "bug" but an intentional decision. Do you seriously believe Google's DEI initiative is unrelated? Why ignore what they specifically say they are doing?

eg.

>moving forward, all VP+ performance reviews will include an evaluation of leadership in support of diversity, equity and inclusion. [0]

[0] https://about.google/commitments/racialequity/#our-commitmen...

[pachorizons]

Do you really think that a DEI mandate specifies refusing to produce white faces in historical contexts? That is an extraordinary position, especially in the face of the layoffs of the ethics teams that this sort of claim would consider as collaborators for such a conspiracy. Do you really truly think that this is anything beyond the obvious naked incompetence of Google under it's current leadership?

To put it another way, this is an unsophisticated position. It is an unsophisticated to describe Google or Microsoft as Nazi organization for releasing a racist AI, so to is it an unsophisticated position to for clearly absurd "DEI" behaviour from Gemini. What both examples demonstrate are companies utterly incapble of governing this kind of system.

[fngjdflmdflg]

>Do you really think that a DEI mandate specifies refusing to produce white faces in historical contexts?

I don't think that. I do think it created a system that disallowed it by mandating perhaps slightly less ridiculous rules that are probably just as bad. That is, there is no rule "do not produce white faces in historical contexts" but there is a rule "add black or brown people to every image". This results in that they will frequently generate generate black vikings, brown Napoleons ect. Separately, Google seemingly does not allow the prompt "create an image of a white male."[0] (If reports are to be believed, which may be wrong, and I don't use Google's service here so I can't test it. Perhaps it's a 50% fail rate or something, it does seem hard to believe that they are doing this). So this taken together results in it being impossible to generate white people (or people that have light skin or whatever term) as they existed historically eg. as Vikings or people like Napoleon or a general "European king."

It seems from the rest of the post that you don't know that they truly are adding extra text to users prompts asking for diverse people in the images. That is really the case, and I will go look again to find the sources that show this.

Edit:

source 1:

>Google might have been adding ethnic diversity terms to user prompts “under-the-hood,” said [Margaret Mitchell, former co-lead of Ethical AI at Google and chief ethics scientist at AI start-up Hugging Face]. In that case, a prompt like “portrait of a chef” could become “portrait of a chef who is indigenous.” [1]

source 2:

>Google's Gemini system seems to do something similar, taking a user's image-generation prompt (the instruction, such as "make a painting of the founding fathers") and inserting terms for racial and gender diversity, such as "South Asian" or "non-binary" into the prompt before it is sent to the image-generator model.[2]

[0] https://www.nytimes.com/2024/02/22/technology/google-gemini-...

[1] https://www.washingtonpost.com/technology/2024/02/22/google-...

[2] https://arstechnica.com/information-technology/2024/02/googl...

[neverrroot]

I'm trying to learn more about what disinformation is, could you please post your thoughts under the following Ask HN post?

https://news.ycombinator.com/item?id=39554369