The total number of hyperparameters (sum of all the model blocks) is 16.25B, which is large but less than expected.
I assume you meant just "parameters" since "hyperparameters" has a specific alternate meaning? Sorry for the pedantry lol.
Imagen Video: high definition video generation with diffusion models
41–50 of 500 posts
Re: Imagen Video: high definition video generation with diffusion models
#42What everyone is missing is that these AI image/video generators lack _taste_. These tools just regurgitate a mishmash of images from it's training set, without any "feeling". What you're going to tell me that you can train them to have feeling? It's never going to happen.
That's purely subjective. We can definitely model AI to give a certain mood. Sentiment analysis and classification is very advanced, it just hasn't been put in these models.
If you think AI will never catch up to anything a human can do, you're simply wrong.
Re: Imagen Video: high definition video generation with diffusion models
#43This sort of AI related work seems to be accelerating at an insane speed recently.
I remember being super impressed by AI Dungeon and now in the span of a few months we have got DALLE-2 , Stable Diffussion, Imagen, that one AI powered video editor, etc.
Where do we think we will be at in 5 years??
Re: Imagen Video: high definition video generation with diffusion models
#44Probably only 6 months until we get this in stable diffusion format. Things are about to get nuts and awesome.
Isn't Imagen a diffusion model?
From the abstract: > We present Imagen Video, a text-conditional video generation system based on a cascade of video diffusion models
Re: Imagen Video: high definition video generation with diffusion models
#45> We have decided not to release the Imagen Video model or its source code
...until they're able to engineer biases into it to make the output non-representative of the internet.
Re: Imagen Video: high definition video generation with diffusion models
#46Earlier quoted context omitted.
byte alignment has always been a consideration for high performance computing. this alludes to a fascinating, yet elementary, fact about computer science to me: there’s a physical atomic constraint in every algorithm.
that's not byte alignment, though- those constraints are what can be held in GPU RAM during a training batch, which is subject to a number of limits, such as "optimal texture size is a power of 2 or the next power of 2 larger than your preferred size". Byte alignment would be more like "it's three channels of data, but we use 4 bytes (wasting 1 byte) to keep the data aligned on a platform that only allows word-level…
thanks for the insight. you obviously understand the domain better than me. let me try and catch up before I say anything more.
Re: Imagen Video: high definition video generation with diffusion models
#47And there you have it. As an aspiring filmmaker and an AI researcher, I'm going to relish the next decade or so where my talents are still relevant. We're entering the golden age of art, where the AIs are just good enough to be used as tools to create more and more creative things, but not good enough yet to fully replace the artist. I'm excited for the golden age, and uncertain about what comes after it's over, but regardless of what the future holds I'm gonna focus on making great art here and now, because that's what makes me happy!
Re: Imagen Video: high definition video generation with diffusion models
#48I’m going to post an Ask HN about what am I supposed to do when I’m “disrupted”. I work in film / video / CG where the bread and butter is short form advertising for Youtube, Instagram and TV. It’s painfully obvious that in 1 year the job might be exceedingly more difficult than it is now.
Here's the link to kranke155's submission: https://news.ycombinator.com/item?id=33099182
Re: Imagen Video: high definition video generation with diffusion models
#49Earlier quoted context omitted.
Whatever insights and expertize you've gained up until now can probably be used to gain enough of a competitive advantage in this future industry to be employed. I doubt the people that will spend their time on this professionally will be former coders etc. (I've seen the stable diffusion outputs that coders will tweet. It's a good illustration that taste is still hugely important.)
I like your optimism but OP's job is to take text instructions and turn them into video, for advertisements. If Google (who already control so much of the advertising space) can take text instructions and turn them into advertisements, what's left for OP to do here? Even if there's some additional editing required this seems like it will greatly reduce the hours an editor is needed. And it can probably iterate option…
Maybe OP's future involves being able to do their work 10x faster, while producing much higher quality results than people who have been given access to a generative AI model without first spending a decade+ learning what makes a good film clip.
The optimistic view of all of this is that these tools will give people with skill and experience a massive productivity boost, allowing them to do the best work of their careers.
There are plenty of pessimistic views too. In a few years time we'll be able to look back on this and see which viewpoints won.
Re: Imagen Video: high definition video generation with diffusion models
#50Probably only 6 months until we get this in stable diffusion format. Things are about to get nuts and awesome.
Isn't Imagen a diffusion model? From the abstract: > We present Imagen Video, a text-conditional video generation system based on a cascade of video diffusion models
"Stable Diffusion" is a particular brand from the company Stability AI that is famously open sourcing all of their models.