[perceptronas]

Their product embodied their values. It turned out that their values are quite radical when exposed to general public. In my opinion, unless there are people and cultural changes - its quite hard to imagine their long term success in this space

[foota]

I don't think it's radical that when prompted with something like "Generate photos of doctors", that it's reasonable to return a set of images that shows diversity (e.g., instead of being a bunch of white men), even if that isn't representative of a "population sample".

I guess though there were unintended consequence where I imagine they're prompting the model with something along the lines of "and remember to be diverse!", and there are obviously some cases where this isn't a good idea. In particular, when the prompt itself is for something that is explicitly racial or where the result is "charged".

E.g., if someone asks for photos of white people, the AI shouldn't generate photos of people that aren't white (and fine, it might return a disclaimer that it only generate white people because you asked it to).

More nuanced though are situations like asking it about historically evil people (e.g., Nazis, as was one of the examples I've seen) but also more benign things like British monarchs or something. I think trying to figure out what kind of results to "inject" diversity into isn't easy though, since it feels like there are many edge cases.

[mike_hearn]

> I don't think it's radical that when prompted with something like "Generate photos of doctors", that it's reasonable to return a set of images that shows diversity

Historically Google had a very simple solution to globally differing expectations about query results: IP or account geolocation. Query personalization by geography is one of the biggest quality wins in web search. Generalizing, an AI built with the same values and ethos as classical Google web search would respond to "Generate a photo of doctors" differently depending on where in the world you asked it from.

That solution also fixes many other cases that aren't third rails, like "Show me a good nearby restaurant serving local food" which you can't solve by attempting to hallucinate a non-existent restaurant that serves a menu of every conceivable dish weighted by population size.

It's unclear why this solution wouldn't resolve all their stated concerns, so we might infer that their actual goals differ from their stated goals. For example, influencing the people who use their services.

[changingwords]

That doesn't work well in America, maybe works well in less diverse places like Europe/Asia.

If you're in NYC/SF and you search for "generate photos of doctors", you expect to see people of all colors represented. Yet the training data for a lot of this is based off white-centric Anglo-centric media.

"Good restaurant near me"? There's literally a dozen amazing cuisines around.

All this said, I'm actually not a fan of this forced 'diversity' in results. Just show me the data and hope that we'll have more diverse data sources.

[dkn775]

Another issue I see based on your comment is that segmenting based on locale (diverse mix in SF, white majority in Kansas) is that it can just take what knowledge and norms exist now and harden them.

[mike_hearn]

There's no reason the results have to match the training data. Obviously, the current outputs it generates don't.

[Ekaros]

And I think it is radical. If the consumer wants certain race, gender, disability and so on they can do it via prompt.

If they want diversity do it by corpus. Get images from Africa and Asia and so on... Feed that to model to get there...

[bcye]

> If the consumer wants certain race, gender, disability and so on they can do it via prompt

No in that case you'll just have a default. Believing that the default should be diverse/random is imo not a radical view.

[alexey-salmin]

If it's "random" as in "proportional to the actual distribution" then it's perfectly reasonable.

But random as in "proportional to an imaginary world that some people want to present as reality" is questionable.

[Ekaros]

And here the "proportional to the actual distribution" means distribution in training data. If that is not diverse enough, they can very well spend couple billions and go get more from areas that increase that diversity, like mentioned Africa and Asia and maybe South America...

[bcye]

I think it's probably (accidentally?) proportional to actual distribution (worldwide not in the west)

[crooked-v]

> do it by corpus

That would be a default. Defaulting to current demographics fitting whatever context is requested (e.g. "a set of US doctors" matching US population demographics) would be an entirely reasonable default, but it would still be a default.

[jart]

I don't think Google sets the agenda. They get their orders like the rest of us, and then probably handed it over to an army of low-paid offshore contractors to implement. How else could you explain Gemma believing Abe Lincoln was black? If Googlers had done the mind-killing work of RLHF training themselves, things would have turned out differently. They were however nice enough to give us a version of the model that doesn't have RLHF training and it's illuminating to compare how it responds to certain queries, because oftentimes the sentences will be the same and only have specific key words or phrases within the sentence changed.

[Taurenking]

Wishful thinking. The CEO of Gemini expressed openly racist views on Twitter. Those views are reflected in the tool built.

> I don't think Google sets the agenda

Soo, you're saying they did not internally test this at all, and delegated that to whoever sets their agenda?

[bitcharmer]

Can you share their tweets?

[hnburnsy]

Just sharing, dont shoot...

https://twitter.com/eyeslasho/status/1760650986425618509

[dekhn]

buthe's not the CEO of gemini. He was a director which is a fairly vague term at Google (Peter Norvig says he's Director of Research at Google, but he was a director of research, not the director). Also he (Krawcyzk) locked his twitter and his linkedin. I can imagine Sundar is also looking for ways to get this guy out of the decision path for Gemini.

[jart]

Don't believe Fox News when they say this guy is a Google exec. It looks like he's just a product manager. I wouldn't hold being spicy on Twitter four years ago against any man, but someone torpedoed their Gemini product and a product manager actually would have the authority to dictate these kinds of decisions. For example, the way the the language model has been programmed to rewrite the user's question before handing it over to the vision model. That's something that's most likely attributable to him. I mean it's one thing to change the training data to remove all information about personal appearance, a curious thing for that algorithm to fail to include Internet rumors about Abe Lincoln being black in that category, an annoying thing for it to refuse to generate content it considers offensive, and another thing entirely to program an information system to meddle with user queries on top of that since that shows an unprecedented level of disrespect.

[visarga]

Their values, but there is also common sense. If they start pushing out unbelievable things like a black queen of England ... they are hurting their own goals. Diversification should stop at the boundary of believability.

[astrange]

A company isn't "their", it's made of people, and not all of those people see all of their products before they're launched.