Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Google to pause Gemini image generation of people after issues

theverge.com

Post reply on HN
Print view View original

261–270 of 1001 posts

Re: Google to pause Gemini image generation of people after issues

#261

Post by cassac »

wokwokwok wrote: post #183
hajile wrote: post #156

Earlier quoted context omitted.

Not all publicity is good. How many people will never again trust Google's AI because they know Google is eager to bias the results? Competitors are already pointing out that their models don't make those mistakes, so you should use them instead. Then there's the news about the original Gemini demo being faked too. This seems more likely to kill the product than help it.

> How many people will never again trust Google's AI because they know Google is eager to bias the results? Seems like hyperbole. Probably literally no one is offended to the point that they will never trust google again by this. People seem determined to believe that google will fail and want google to fail; and they may; but this won’t cause it. It’ll just be a wave in the ocean. People have short memories. In 6 mo…

The real surprise is that anyone trusted google about anything in the first place.

Re: Google to pause Gemini image generation of people after issues

#262

Post by chasd00 »

perihelions wrote: post #39

There's an amusing irony here: real diversity would entail many competing ML companies from non-Western countries —each of which would bring their own cultural norms, alien and uncomfortable to Westerners. There's no cultural diversity in Silicon Valley being a global hegemon: exporting a narrow sliver of the world's viewpoints to the whole planet, imposing them with the paternalism drawn from our own sense of superi…

> imposing them with the paternalism drawn from our own sense of superiority.

The pandemic really drove this point home for me. Even here on HN groupthink violations were delt with swiftly and harshly. SV reminds me of the old Metallica song Eye of the Beholder.

Doesn't matter what you see Or intuit what you read You can do it your own way If it's done just how I say

Re: Google to pause Gemini image generation of people after issues

#263

Post by xetplan »

_heimdall wrote: post #224

We humans haven't even figured out how to discuss race, sex, or gender without it devolving into a tribal political fight. We shouldn't be surprised that algorithms we create and train on our own content will similarly be confused. Its the exact same reason we won't solve the alignment problem and have basically given up on it. We can't align humans with ourselves, we'll absolutely never define some magic ruleset tha…

We figured this out a long time ago. People are just bored and addicted to drama.

Re: Google to pause Gemini image generation of people after issues

#264

Post by rohtashotas »

Jensson wrote: post #139

> Of course the politically sensitive people are waging war over it. Just like politically sensitive people waged war over Google identifying an obscured person as a Gorilla. Its just a silly mistake, how could anyone get upset over that?

It's not a silly mistake. It was rlhf'd to do this intentionally.

When the results are more extremist than the unfiltered model, it's no longer a 'small mistake'

Re: Google to pause Gemini image generation of people after issues

#265

Post by oceanplexian »

anonzzzies wrote: post #188
frozenlettuce wrote: post #112

How would a product like that be monetized one day? This week openai released the Sora video, alongside the prompts that generated them (the aí follows the description closely). In the same week, Google releases something that looks like last year's MidJourney and it doesn't follow your prompt, making you discard 3 out of 4 results, if not all. If that was billed, no one would use it. My only guess is that they are t…

> How would a product like that be monetized one day? For video (Sora 2030 or so) and music I can see the 'one day'. Not really so much with the protected/neutered models but: - sell/rent to studios to generate new shows fast on demand (if using existing actors, auto royalties) - add to netflix for extra $$$ to continue a (cancelled) show 'forever' (if using existing actors, auto royalties) - 'generate one song like…

I think the technology curve will bend upward much faster than that, as humans we’re really bad at perceiving exponential change over time. By next year this will be used to generate at least parts of films and TV shows.

By the 2030’s this technology will be on-device, real time, and anyone will be able use it. You won’t need to buy movies when you can generate them, probably causing a collapse of the entertainment industry. AR/VR will use this technology shortly after, resembling something like the Holodeck from Star Trek where you simply prompt it and it creates a customized simulation.

Re: Google to pause Gemini image generation of people after issues

#266

Post by sega_sai »

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 globe or US ?

In my opinion, some corrections are worthwhile. In this case they clearly overdone it or it was a broken implementation. For sure there will be always people who are not satisfied. But I also think that the AI services should be more open about exact guidelines they impose, so we can debate those.

Re: Google to pause Gemini image generation of people after issues

#268

Post by Semaphor »

indy wrote: post #233
Semaphor wrote: post #104

Earlier quoted context omitted.

Not a great link for an international audience. Here in Germany, the top image is a white person: https://i.imgur.com/wqfdJ95.png

That's an Asian person

Interesting, even on a second look I'm not able to tell that.

Re: Google to pause Gemini image generation of people after issues

#270

Post by tourmalinetaco »

fenomas wrote: post #193
lwansbrough wrote: post #3

Issue appears to be that the uncensored model too closely reflects reality with all its troubling details such as history.

Surely it's more likely that Google is just appending random keywords to incoming prompts, the same way DALLE used to do (or still does)?

It wouldn’t shock me either way, Google loves to both neuter products into uselessness and fuck with user inputs to skew results for what they deem is best for them.
Post reply on HN