Synthetic Media & Generative AI News and Discussions

[Tadasuke]

>
Google claims its text-to-image AI delivers 'unprecedented photorealism'

https://www.engadget.com/google-imagen- ... 05123.html

Google has shown off an artificial intelligence system that can create images based on text input. The idea is that users can enter any descriptive text and the AI will turn that into an image. The company says the Imagen diffusion model, created by the Brain Team at Google Research, offers "an unprecedented degree of photorealism and a deep level of language understanding."

This isn't the first time we've seen AI models like this. OpenAI's DALL-E (and its successor) generated headlines as well as images because of how adeptly it can turn text into visuals. Google's version, however, tries to create more realistic images.
To assess Imagen against other text-to-image models (including DALL-E 2, VQ-GAN+CLIP and Latent Diffusion Models), the researchers created a benchmark called DrawBench. That's a list of 200 text prompts that were entered into each model. Human raters were asked to assess each image. They "prefer Imagen over other models in side-by-side comparisons, both in terms of sample quality and image-text alignment," Google said.

The examples are curated and it's not publicly available. Still needs improvement in social biases and limitations of large language models.

[Lorem Ipsum]

>
Google claims its text-to-image AI delivers 'unprecedented photorealism'

https://www.engadget.com/google-imagen- ... 05123.html

Not that impressive at the moment other than its accurate text generation.

[Yuli Ban]

[Yuli Ban]

[Yuli Ban]

[Yuli Ban]

My prediction is that by 2025 there will exist a protoform of the Metaverse as was defined before Zuckerberg— a multimodal internet realm unique to each person, of AI generated text, images, audio, and short videos that one can lose themselves within.

[Lorem Ipsum]

My prediction is that by 2025 there will exist a protoform of the Metaverse as was defined before Zuckerberg— a multimodal internet realm unique to each person, of AI generated text, images, audio, and short videos that one can lose themselves within.

Like a software suite?

[caltrek]

Give this AI a few words of description and it produces a stunning image – but is it art?
by Aaron Hertzmann
June 10, 2022

Introduction:

(The Conversation) A picture may be worth a thousand words, but thanks to an artificial intelligence program called DALL-E 2, you can have a professional-looking image with far fewer.

DALL-E 2 is a new neural network algorithm that creates a picture from a short phrase or sentence that you provide. The program, which was announced by the artificial intelligence research laboratory OpenAI in April 2022, hasn’t been released to the public. But a small and growing number of people – myself included – have been given access to experiment with it.

As a researcher studying the nexus of technology and art, I was keen to see how well the program worked. After hours of experimentation, it’s clear that DALL-E – while not without shortcomings – is leaps and bounds ahead of existing image generation technology. It raises immediate questions about how these technologies will change how art is made and consumed. It also raises questions about what it means to be creative when DALL-E 2 seems to automate so much of the creative process itself.

A staggering range of style and subjects

OpenAI researchers built DALL-E 2 from an enormous collection of images with captions. They gathered some of the images online and licensed others.

Read more here: https://theconversation.com/give-this- ... rt-184363

[waitingforthe2030s]

My prediction is that by 2025 there will exist a protoform of the Metaverse as was defined before Zuckerberg— a multimodal internet realm unique to each person, of AI generated text, images, audio, and short videos that one can lose themselves within.

By 2025? Far too early.

[Yuli Ban]

[Tadasuke]

Things are slowly getting closer to what I envisioned as a "personal paradise for everyone" by the year 2100, when an affordable personal computer should be a trillion times faster than a PS5 and a quadrillion times faster than a PS2.

Today the problem is that you need to browse the web to find what you like and that browsing can take time and also be disturbing or disgusting (at least for me). By the year 2040, AI will be capable to either quickly find what you like on the web or to create what you like (to an extent). Whatever image you imagine will be possible to be created nearly instantly in the year 2040. Whatever fantasies you may have will be made possible in image form by AIs. Web stories will be always illustrated and memes will be much more original. Also dubbing will be done by AIs by 2040. Of course things are becoming possible even now, but in a form that needs a lot of improvement to be perfected. We will probably see whole comics done using AI programs. By 2040 some people will even try animations done entirely by AIs.

I think it will be a better future overall. Of course people will also be creating disturbing, disgusting things with AIs, but the pros will top the cons. Just like the WWW today is better than no worldwide web at all.

[Yuli Ban]

[Yuli Ban]

[Yuli Ban]

I am not involved in the visual arts industry so I wouldn't know, but I am aware of some panic that has been simmering for a few months.

I distinctly remember a fairly vicious pushback I had back in 2020 when I argued that such a thing like DALL-E 2 was imminent, with the comment accusing me of not understanding art and overexaggerating AI progress. Here we are 2 years later. It's not going to take another 2 years for disruption to eviscerate multiple industries.

It feels good to argue that we have 5 to 10 years left because that gives us humans just enough time to make bank, make a career, and high-tail it out of here.

Poppycock! DALL-E 3 will be a 2023 invention (2024 at latest) and it'll mop the floor with humans so magnificently that if it was publicly released the same day, you could almost immediately shut down whole colleges of art, modeling, and graphic design and not look back.


Should stress that I DO have some experience in the literary arts industry, and even when I was more active, I was using artificial intelligence to help me out. That was 2019-2020. Not just because I was already obsessed with synthetic media, I could read the writing on the wall (no pun intended) and tell that there was at max 5 to 7 years left before AI completely dominated that field. Only because of the nature of long-form coherent storytelling do I stick with that prediction— an AI that can generate a whole coherent novel with a beginning, middle, and end that all ties into itself without going on bizarro tangents can also pass the Turing Test reliably.

Creating ultra-coherent art and photographs won't take much more than what we have now. Just a bit more scaling.

[Yuli Ban]

Blast from the past: synthetic media from 2015



Think about just how far we've come in just seven years!

[Yuli Ban]

More news from 2015

Crazy to think this used to be the state of the art, the absolute best we could do in terms of image synthesis. It looks so quaint, so primitive by modern standards.


And don't you worry, one day we'll be saying the same thing about this:

[Yuli Ban]

I feel people didn't react to this enough

Dr. Wood says “This is simply the most impressive AI result I have personally seen in my career. Long range coherence is a challenge even for modern language models with massive parameter counts. Will, Saeid, Vaden, and Christian have taken a huge step forward by being able to stably generate coherent, photo-realistic 1hour+ long videos; 70x’s longer than their longest training video, and more than 2000x’s longer than the maximum of 20 frames they ever look at at once during training. There is something very special in the training procedure they have developed and the architecture they employ. Never have we been closer to being able to formulate AI agents that plan visually in domains with life-like complexity.”

It did all that, over an hour of novel video... off only 78 million parameters. Literally smaller than GPT-1.

[Yuli Ban]

A video call between Berlin Mayor Franziska Giffey and a man pretending to be Kyiv Mayor Vitali Klitschko broke off as suspicion grew he was an imposter, the Berlin Senate Chancellery said on Friday.

"There were no reference points that the video conference was not being conducted with a real person. Apparently, it is a deepfake. Police have been called in to investigate," the chancellory wrote on Twitter.

[Yuli Ban]

[Yuli Ban]