Veo and Imagen 3: Announcing new video and image generation models on Vertex AI | Hacker News Reader

Veo and Imagen 3: Announcing new video and image generation models on Vertex AI

cloud.google.com

30 pointssrameshc15 comments

Google has learned nothing in the last 3 years. Still trying to push people to use their clunky GCP to do AI when all other platforms are a one click signup.

I clicked on Veo link in the article, went to deep mind veo landing page, signed in, redirected to test kitchen with Google, signed in again, to see another button saying join the waiting list.

Do they ever do usability test at Google?!

The PMs scored a win, who cares if it's actually usable. They still congratulate eachother on getting some DL papers out in 2016, and the juggernaut will keep chugging along burning ad monopoly money.

They're so inwards focused, it's like mainstream politics. Noone cares that they are turning into an obsolete also-ran as long as there's internal fame to be made.

presumably they expect that the people who are going to bring in most of the money already have >$100k gcp accounts

Google has rolled out a few AI features in GCP before making them available to general consumers. It would be more accurate to compare the GCP setup process to AWS Bedrock AI model setup. Gemini is very easy to set up in GCP.

Okay, but can Imagen 3 create dozens of pictures of the same character in a variety of situations, with consistent but variable outfits?

Because PASanctuary can…

Couldn’t find anything about PASanctuary. Did I miss a joke ?

It's here: https://civitai.com/models/835578/pasanctuary-sdxl-illustrio...

Definitely has the best character consistency of any SDXL model I've ever seen, and a huge range of characters as well. So long as what you want is some brand of anime, it'll work; in combination with latent coupling and in/outpainting, I can always get precisely the pictures I want.

Having a range of characters is useful even if I want it to draw original ones, because I can go "This character here, except with brown hair and older" and it'll still consistently draw the same person.

Yes. There's an example in the linked announcement.

There’s an example of doing it with a dress. People are far more sensitive to flaws in faces, so I believe the lack of a demonstration is suggestive.

Though I’ll certainly be experimenting.

Apparently I will not be experimenting, because it's only available to 'trusted testers'. How typical.

Videogen is still not generally available. It's just a beta sign up, as it's been for what- a year?

As with most corporate gen-ai, it's going to remain in private 'safety testing' until open source (or at least some other corporation) has caught up.

From the article:

> Generative AI is leading to real business growth and transformation. Among enterprise companies with gen AI in production, 86% report an increase in revenue1, with an estimated 6% growth.

Most of that business growth is going to be in more vapid media for more effective advertising, which is why Google is investing in it. Is that really a good thing for anyone? I mean, of course it provides short-term economic gain for some but that is only because we are constantly modifying society to make it more lucrative to push consumerism as a way of life. We need less of this AI trash, not more.

To be honest I am absolutely sick of this sort of technology that wastes enormous amount of energy, makes us more dependent on using more energy (which in some places that will be fossil-fuel use), and which does not give us any real benefit, especially if we consider that we could instead work on transitioning to a lower-energy economy.

I believe long into the future, if someone looks back at us, companies like Google will be seen as the height of stupidity of humankind. If we have any intelligence left, we should dismantle this AI machine before it's too late.

Yes yes lofty ideals but will you please think of the shareholder value?!

Also, taken at face value this quote is unadulterated bullshit:

> Among enterprise companies with gen AI in production, 86% report an increase in revenue, with an estimated 6% growth.

Zero causation. The average company reporting 6% growth is par for the course in the current age of growth at all costs. There is nothing in this statement to indicate that presence of GenAI pet projects helps improve growth.

True, but AI is definitely making a measurable impact in some businesses where gen AI is already replacing some illustrators and graphic designers. I think there is already strong evidence for causation.

Reply on news.ycombinator.com