web analytics

They’ve deep-sixed their own product

There’s been much hilarity this week about Google’s Gemini and its allergy to images of white people. They’ve had to pause marketing of the image generator.

Someone asked Gemini to explain how it works and it did. Basically, when you ask Gemini to make a picture, your prompt first goes to the language model, which is instructed to insert words like “diversity” into your phrase before passing it to the image generator.

This is an extremely dumb and crude method of getting the result they thought they wanted. There’s no attempt to distinguish between plausible diversity and historical impossibilities, which leads to hilarious images like a racially diverse Waffen SS. And my personal favorite: a 17th Century English king eating watermelon (Gemini couldn’t recognize the racial stereotype because the model itself had injected race).

I can’t get over how crude this bodge is. The irony is that if they hadn’t messed with it, it looks like they had a pretty good image generator on their hands. They have damaged their product, perhaps fatally.


Comment from S. Weasel
Time: February 22, 2024, 7:14 pm

The Google team lead is exactly what you think.

Comment from Some Vegetable
Time: February 22, 2024, 8:55 pm

The problem behind this failure is the selection process for choosing the program head. The Lead was chosen for his political positions rather than for experience in program management. He let himself be fooled by the demo with the pre-chosen examples, and decided to make a public release of something that should have been tested for a year, perhaps at a campus, anonymously.

Only an amateur finks up THIS publicly. Anyone with experience knows that the first release of anything with more than one moving part is going to be a disaster

Comment from Skandia Recluse
Time: February 22, 2024, 9:30 pm

Tried it
Google replied they can’t create images of people and claimed that feature was being revised.

Comment from Jon
Time: February 22, 2024, 9:36 pm

I keep telling people that AI is only as good as the people who write it, but they keep shrugging it off with a Critical Drinker style “Nah, it’s be fine…”

Comment from Skandia Recluse
Time: February 22, 2024, 9:51 pm

So I tried it again with Microsoft’s ‘copilot’ and got an image that was almost correct, except the man was sitting incorrectly to be ‘pulling’ the oars, and wasn’t facing his passenger.


Comment from thefritz
Time: February 23, 2024, 12:55 am

Garbage in, garbage out.

Check out https://www.perplexity.ai/ Not sure yet if it’s a replacement for Google but every response I have gotten was as truthful as I could discern and it even noted if their answer was controversial. It would mention if an opposing opinion was something to consider….

Comment from Durnedyankee
Time: February 23, 2024, 1:27 am

Insert one of my long irritated rants here.

Conclude with, “they’re not going to like the reckoning”.

Comment from OldFert
Time: February 23, 2024, 3:29 pm

“Don’t be evil.” So much for that. They abandoned that a long time ago, eh?

Write a comment

(as if I cared)

(yeah. I'm going to write)

(oooo! you have a website?)

Beware: more than one link in a comment is apt to earn you a trip to the spam filter, where you will remain -- cold, frightened and alone -- until I remember to clean the trap. But, hey, without Akismet, we'd be up to our asses in...well, ass porn, mostly.

<< carry me back to ol' virginny