Google says its AI image-generator would sometimes 'overcompensate' for diversity

How about Google itself allowing for diversity in the tech marketplace by willingly handing over some of its massive market share to a minority tech company?

Google's AI is not "overcompensating" it's downright racist.

It's only as intelligent as the people who program it. To be fair, it's pretty impressive that people can now generate (sometimes) useful images from a text prompt.

It’s ok to be biased about white people, if it was biased against blacks they’d be riots and a lawsuit. Such is the way the world works these days you see.

It's only as intelligent as the people who program it.

No, it's not even on the same plane of intelligence as the people who program it. AIs are still very simple, nothing even remotely close to the complexity of humans.

I think it can generate images better (and certainly faster) than most of its programmers could. AI models are good at doing what they've been trained to do, assuming the data used for training the model is good.

No, it's not even on the same plane of intelligence as the people who program it. AIs are still very simple, nothing even remotely close to the complexity of humans.

I think it can generate images better (and certainly faster) than most of its programmers could.

Can it comprehend the moral implications of the images it generates?

Because the programmers can.

It can't comprehend anything. It just processes data and generates images according to its programming and the training data it's provided.

Can it comprehend the moral implications of the images it generates?

No, it's not even on the same plane of intelligence as the people who program it. AIs are still very simple, nothing even remotely close to the complexity of humans.

If only that were true. What is causing alarm among some, like Geoffrey Hinton who created the back propagation neural net, is that it is learning better and faster than humans, not through programming but training.

He fears that algo has several advantages over human cognition.

That being said, it is still at a very early stage but the advances have been exponential.

When I first saw the images mentioned, I thought it they were comical memes, for example, the black and Asian Nazis. You'd have to be extremely ignorant or woke-radicalized not to smirk at something like that. And no, it's not "overcompensating." It's plain wrong.

How about Google itself.... willingly handing over some of its massive market share to a minority tech company?

They don't need to. Minorities already make up a majority of Alphabet's tech employees. Also, the two white founders of Google handed over the leadership of the company to a person of color, from India. Same with Microsoft. I guess that's what happens in "systemically racist America."

AI seems to be woke too.

C'mon, Google, don't go overboard like Hollywood has. Fix this.

It can't comprehend anything. It just processes data and generates images according to its programming and the training data it's provided.

Exactly. Which was my point:

it's not even on the same plane of intelligence as the people who program it. AIs are still very simple, nothing even remotely close to the complexity of humans.

No, the AI was programmed and trained to be able to “overcompensate” by woke liberal humans.

No, the AI was programmed and trained to be able to “overcompensate” by woke liberal humans

Not a surprise. Programming isn’t easy and liberals tend to be more intelligent.

If they had the capacity, can you imagine AI programmed by conspiracy theorists?

This figures. The mainstream liberal-leaning media are now rushing to Google's defense.

From the article linked below:

"With Gemini, all of the DEI initiatives that have run rampant in Big Tech for so long finally blew up in their faces this week, because they slipped up and showed us exactly what they’re trying to do — which is to erase white people at every possible opportunity."

https://www.dailywire.com/news/just-when-you-think-the-google-gemini-ai-story-cant-get-any-worse-it-does

https://www.dailywire.com/news/just-when-you-think-the-google-gemini-ai-story-cant-get-any-worse-it-does

Daily wire?

We aren’t teenagers here.

Be serious.

It will get worse but that is another topic.

My surprise was to see J S Bach as ‘a black man’

and Gemini’s refusal to produce an image of ‘a white family’

Not at all intelligent…

Hilarious. The BBC have been salting 'adaptations' of classic fiction with non-whites to such an extent that Jane Austen will soon be regarded as a diversity activist.

If you fake history, you hide the prejudice and remove the chance to learn from its mistakes.

the AI was programmed and trained to be able to “overcompensate” by woke liberal humans.

Oh dear, maybe they’re trying to replace you…

Show me a historical Viking leader.

Black woman shown.

it’s intentional.

Bias in bias out. I wonder if I will show up as a black transgender man.

This is just an extension of the rewriting of history, the wanting to please everyone regardless of facts and reality. Cleopatra, Hannibal, etc...in Hollywood, places like Canada in supposed historical places with government locations and reenactments of battles and forts garrisons and 17th century battles between the French and English, inserting women and ethnic soldiers in live reenactments and books to the point that when asked many youth today in Canada actually think that the British and french in the 1600 and 1700 had females soldiers, etc..in the ranks of their armies.

So with the above being acceptable and AI programs designed to "learn" from what they collect around them (AKA from the web) and unable to tell the difference from what is being "created" based on modern "inclusion" theories vs what was reality before modern times and the internet, one shouldn't be surprised if these AIs make George Washington a woman, or Henry the 8th an Asian, etc...

Before certain people say my previous post is off topic let me explain better.

The examples I gave showing what governments and Hollywood are doing to history and historical facts can distort our view of the facts and reality.

But we are human, we are not limited to images and information on the internet, we can go out, talk to actual historians, visit actual museums, libraries where much has still not been digitized and put on the web.

So these AIs are limited to what is published on the web, they have only the concept that is placed here and in computers.

So if it finds the photos if a 17th century battle between the French and English from a "reenactment" it cannot make the connection between the reenactments being updated for diversity and the fact of what the men who fought in those battles actually look like because no photos exist.

This can be the same for the AI mistaking "historical drama" made by Hollywood as a historical documentary, etc...

These AI main flaw is they are binded to the web and only what they can collect on the web.

Garbage in, garbage out. This was intentional and what happens when DEI goes haywire. And Google has been rightly ridiculed for it. The apology was nothing more than a lame face-saving measure that's only made the situation worse for Google. Maybe they'll learn, but they'll have to clean out the diversity hires first.

The AI should rely on primary texts and not modern interpretations and distortions of history or else it is just fiction.

Ok. I guess you are arguing with someone else then.

Exactly. Which was my point:

it's not even on the same plane of intelligence as the people who program it. AIs are still very simple, nothing even remotely close to the complexity of humans.

Redemption

Today 02:32 pm JST

The AI should rely on primary texts and not modern interpretations and distortions of history or else it is just fiction.

Well the problem with that is that our schools, government, even many so-called historians are actively rewriting history and ignoring the cultural, ethics and habits of the times historical events were happening!

Yes the cancel culture is real, the idea of holding people especially famous historical figures to today's standards are well documented !

With that as a background, how are AIs supposed to figure things out?

Show me a historical Viking leader.

Black woman shown.

Wildly off topic - but research shows that the Vikings were certainly not all blue-eyed and blonde haired as per the stereotype. They were very racially diverse.

So, maybe the algorithm has things half right here?!

With that as a background, how are AIs supposed to figure things out?

Yes, that is my point. In high school I had an enlightened history teacher and he forced us to review primary sources, such as letters written by George Washington, actual newspapers articles from the time and so on, and discuss history in that context coming to our own conclusions.

South Park will have a field day with this.

Fighto!

Today 03:46 pm JST

Show me a historical Viking leader.

> Black woman shown.

> Wildly off topic - but research shows that the Vikings were certainly not all blue-eyed and blonde haired as per the stereotype. They were very racially diverse.

> So, maybe the algorithm has things half right here?

Actually yes and no!

This type of comment is the exact type the AI can pickup on and get the wrong ideas!

No they were not blonde hair blue eyes only but they were Caucasian and in fact only one instance recorded of a viking leader having a dark shin Asian wife a former captive thought to be Mongolian was ever recorded and she is supposed to have switched her own children with someone else's as to not let her husband know his children were dark skin with her Asian features!

So the idea a black woman would be a viking leader has no historical fact or evidence and spreading this as a possibility then leads to the AI learning unsupported non historical information.

And hence to problem with AI, they cannot make a difference between someone wishing something be true like what you wrote and the historical facts.

AI is turning out to be junk data.

'Garbage in, garbage out', that's not applicable here and also the algorithms and the machines work exactly error-free as they are instructed. The problem is, that this 'techology' can't and won't work properly right from beginning, which is mathematically intrinsic. There are many reasons that it all has slept untouched in the researchers' desk drawers for many decades, since the late 1940's btw. So either you accept is as is, which means it doesn't work, or you put it back into the desk drawers, something I would recommend or prefer.

It can't comprehend anything. It just processes data and generates images according to its programming and the training data it's provided.

I suppose it depends on one’s definition of ‘comprehend’. Certainly transformer neural networks, such as Gemini and Sora, have come to develop ‘neurons’ (i.e. a specific node or set of nodes that capture specific information) that recognize such disparate things as sentiment, entities, and fluid dynamics without ever receiving explicit programing in such. If you want to fault TNNs for not 'comprehending' things, then I'd argue that most people do not really 'comprehend' things most of the time so much as 'recognize' them.

Generative AI market's answer to Google's long-standing search monopoly. So long as not asking opinions, results far better for most, no $clickbait and time savings. E.g. Hotels with X, Y and Z near airport ABC = Great Answer

If you get into opinion-based queries, AI interference models become AI gas-lighting/biased in order to satisfy agenda of KEY Hidden Stakeholders = Terrible Answer

Your welcome

It seems pretty clear that Big Tech hates White people.

It's not just Google. It's Microsoft too:

Headline: Microsoft Brags That It Pays White Employees Less Than Non-White Employees Despite ‘Same Job Title’

Link: https://www.dailywire.com/news/microsoft-brags-that-it-pays-white-employees-less-than-non-white-employees-despite-same-job-title?fbclid=IwAR1erCuh4f1FrBuVeYjjSWsjeV2fFLy2euORnvr7H71ZVWXStp60dHW5Vq8

Such discrimination against any other group of people would not be tolerated by society, so why is it tolerated for White people?

What Microsoft has admitted to doing it blatantly illegal, and I hope someone sues the pants off of them.

Keep in mind, Steve Balmer makes +1B$ Per Year in MSFT dividends and he's a white as you get. Bill Gates and Paul Allen, co-founders, ditto. Business agenda to appeal to nonwhites, who make up most of world.

So massive gas-lighting with salary info, as anyone good gets serious stock options.

SCOTUS done away with affirmative action in college admissions, hiring, etc. and they're busy gutting DEI.

Good times for class action lawyers in US!!!

TokyoOldMan

Feb. 25 10:21 pm JST

@Quo Primum - Originally Indian workers were imported to take over American jobs, because they were cheap, now that they have gotten into management, this pay inversion is sweet revenge for them.

Revenge on who?

MS? No they are still getting their work done!

On the white guy that lost his job because imported workers were will to take it at a lower salary?

Not sure who you think is getting revenged on!

Let's not forget most of those cheaper foreign workers like here in Japan are brought in by their own people who low bid these companies then pay their own people (workers) a pitance, don't provide health insurance, vacation time, etc...

The end game of those that believe themselves to be the victims of systemic racism, reverse-systemic racism!

That'll work a treat.

As a white male from the poor side of Europe I can testify I haven't benefited from any kind of privilege in my life. Americans should solve their problems internally, not project them over the rest of the world. It has became so strongly anti-white males biased it's not even funny any longer. We're worth nothing when compared with otherwise similar blacks and will lose evertime in a selection (job interview for example).