More disruption at OpenAI - NYTimes just now I think the paywall is only for artiles older than two weeks but not sure since I've had a subscription for decades:
https://www.nytimes.com/2023/11/20/business/openai-staff-exodus-turmoil.html

Supposedly it was today but AI ACT agreement in Europe is not reached yet.
This week France, Italy and Germany came with the idea of Self-regulation. In some cases self regulation can be interpreted as no regulation at all.

Oh yes, German politicans just love "self-regulation". 

Chinese court declares that AI-generated image has copyright
https://www.technollama.co.uk/chinese-court-declares-that-ai-generated-image-has-copyright

This isn't exactly news, but I thought it was still worth sharing, because this is just so bizzare:

I just stumbled upon this movies webpage and they use AI generated images instead of photos of actors.  motionpicturemagazine.com



Some of the actors are hardly recognizable.

And I also found a website for Asian cooking recipes that also uses AI generated images.
asian-foodie.de



And some of the results are, well, let's say questionable.
Okonomiyaki. The real thing on the left and the AI version on the right. I'd be scared if my Okonomiyaki looked like this.

from uncle pete

https://www.msn.com/en-us/money/companies/ny-times-sues-openai-microsoft-for-infringing-copyrighted-works/ar-AA1m75sX?ocid=00000000&pc=U528&cvid=36cdc2530ee347549955a4670eb08328&ei=17

NEW YORK (Reuters) - The New York Times sued OpenAI and Microsoft on Wednesday, accusing them of using millions of the newspaper's articles without permission to help train chatbots to provide information to readers.

The newspaper's complaint, filed in Manhattan federal court, accused OpenAI and Microsoft of trying to "free-ride on The Times's massive investment in its journalism" by using it to provide alternative means to deliver information to readers.

"There is nothing 'transformative' about using The Times's content without payment to create products that substitute for The Times and steal audiences away from it," the Times said.

The case is New York Times Co v Microsoft Corp et al, U.S. District Court, Southern District of New York, No. 23-11195.

It is a good article.

There are a lot of legal things to clear up with how to use ai responsibly.

I certainly hope midjourney, stable, open ai all license our content properly from the agencies.

Summary of Nvidia ai tools. They are working with. everyone in the industry, Adobe, getty, shutterstock

https://www.nvidia.com/en-us/gpu-cloud/picasso/


NVIDIA Picasso
A foundry for building and deploying generative AI for visual design.
Get Notified
What Is NVIDIA Picasso?
NVIDIA Picasso is a foundry for custom generative AI for visual content, providing a state-of-the-art model architecture to build, customize, and deploy foundation models with ease. Enterprise developers, software creators, and service providers can choose to train, fine-tune, optimize, and infer foundation models for image, video, 3D, 360 HDRi, and PBR materials to meet their visual design needs. Picasso streamlines foundation model training, optimization, and inference on NVIDIA DGX Cloud.
Check out related products.

BioNeMo�
NVIDIA NeMo�
 
Announcements
CES Keynote
Getty Images Releases Generative AI By iStock Powered by NVIDIA Picasso
Watch Announcement
Image courtesy of i-stock
NVIDIA Picasso Workflows
For visual foundation models.
Build a Foundation Model

Train a state-of-the-art Edify model with your enterprise data for a proprietary foundation model for image, video, 3D, PBR materials and 360 HDRi and run inference through APIs.
Customize a Pretrained Foundation Model

Fine-tune a pretrained NVIDIA Edify model on your custom data to meet your unique needs and run inference through APIs.
Accelerate Any Foundation Model

Bring your own foundation model for visual design, ‌optimized by NVIDIA AI experts to run at fast inference speeds on DGX Cloud.
The Picasso Service model and the workflow that it provides to enterprise AI applications.
NVIDIA Edify State-of-the-Art Generative AI Foundation Models
Organizations and developers can train NVIDIA�s Edify model architecture on their proprietary data or get started with models pretrained with our early adopters.
Generate and Modify Images
Expert denoising network to generate photorealistic 4K images.

Text to Image
Generate images with text prompts.

Inpainting
Modify images with text prompts.

Outpainting
Expand images to various aspect ratios with text prompts.

Video
Temporal layers and novel video denoiser generates high-fidelity videos with temporal consistency.

3D
Novel optimization framework for generating 3D objects and meshes with high-quality geometry.
A generated image of a 360 environment
360 HDRi
 State-of-the-art architecture to generate photorealistic environment maps and lighting for 3D scenes.

Materials
Generate tileable materials and textures with text prompts�like brick or mosaic�for virtual scenes.
Supercharge Your Applications With Generative AI
NVIDIA Picasso opens new worlds of possibilities. Differentiate your application with custom generative AI models. Offer in-house teams and customers best-in-class generative AI tools to start their creative journey. Deliver interactive experiences while saving cloud inference costs using powerful inference optimizations in NVIDIA DGX Cloud. Access the best of NVIDIA AI innovation through cloud APIs.
Amazingly Accurate
Model Selection Flexibility
Use state-of-the-art pretrained Edify models from NVIDIA or bring your choice of model for building your generative AI.
Deploy Picasso Anywhere
Performance-Tuned
Achieve best performance on training and inference by using NVIDIA AI on NVIDIA DGX Cloud.
Managed APIs
Data Authenticity
Use models that are pretrained on licensed, enterprise-grade data.
Connect Workflows With OpenUSD
Develop Generative AI-Powered 3D Workflows
Develop custom 3D pipelines and workflows connected to NVIDIA Picasso-based tools with the NVIDIA Omniverse platform.

Early Adopters
AI-generated landscape using the Edify model trained on the Getty Images dataset.
Getty Images
Generative AI by Getty Images is a commercially safe and legally indemnified service that allows creators to use text prompts to generate realistic images. It was built using NVIDIA Picasso and trained on Getty Images� licensed creative library. Enterprise can access the model through API calls.
Bria
Bria employs a tailored, holistic visual generative AI solution for businesses, trained on fully licensed data to ensure accuracy and compliance. With generative AI pretrained models, APIs, web integrations, and more, Bria provides a comprehensive suite of tools to enhance operations and drive innovation.
Bria logo
Adobe will implement Picasso throughout the Adobe Cloud for creators to leverage.
Adobe
Adobe and NVIDIA will codevelop generative AI models with a focus on responsible content attribution and provenance to accelerate workflows of the world�s leading creators and marketers. These models will be jointly developed and brought to market through Adobe Cloud flagship products like Photoshop, Premiere Pro, and After Effects, as well as through Picasso.
Masterpiece-X
An early adopter of generative AI, Masterpiece is using it to create a text-to-3D playground for rapid exploration and prototyping. By using text prompts, assets like mesh, textures, and animations can be generated. These 3D models are compatible with popular apps and game engines such as Blender, Unity, and Unreal Engine.
Masterpiece X logo

Shutterstock
Shutterstock is developing models to generate 3D and 360HDRi assets trained on fully licensed content from Shutterstock, using NVIDIA Picasso. These models can be used to generate high-fidelity 3D assets from simple text prompts, which can in turn be used in game development, animation, and other 3D workflows.
Success Stories

Courtesy of Runway
Runway
Runway uses AI to generate video in any style. The AI model is able to imitate specific styles prompted through given images or through a text prompt. The user can now create new video content using existing video.
This flexibility enables creators to explore and design video in a whole new approach to filmmaking and content creation.
Screen used by Seyhan Lee to build virtual productions.
Courtesy of Seyhan Lee I Cuebric
Seyhan Lee | Cuebric
Cuebric uses generative AI to build and edit virtual productions. This includes building out environments using generative AI models and then distilling the final asset into a volume that engages with the individual in real time.
Their cost-effective method enables filmmakers, production studios, and artists to partner with CGI specialists much earlier in the post-production process.

Courtesy of WOMBO
WOMBO
WOMBO Dream brings generative AI to the masses through their mobile app. The user uses a text prompt to generate a desired image and selects a style prompt, and their image is generated within seconds. The app also allows users to share their generated art with members of the WOMBO community.
More Resources
Watch the Picasso Announcement

Did you miss NVIDIA CEO Jensen Huang�s presentation on generative AI? From LLMs to digital avatars, catch up on everything NVIDIA is doing in generative AI.
Watch Keynote 
GANs to Diffusion�the Path to Generative AI

Join Arash Vahdat, senior NVIDIA Researcher,  to learn about the path from GANs to diffusion models. In this talk, he�ll share his thoughts on how these fundamental technologies drove various applications such as text-to-2D images, video, and 3D content generation.
Watch on NVIDIA On Demand
Read Our Generative AI Ebook

test

test

Why is the thread not showing up on the starter page?

Interesting article co-written by a journalist and a law professor about where the fair use argument has succeeded - and failed - in prior cases. The key issue, IMO, is that things that might be fair use by an individual for personal use, or researchers for academic use are not so if done by a for-profit company for commercial use.

https://arstechnica.com/tech-policy/2024/02/why-the-new-york-times-might-win-its-copyright-lawsuit-against-openai/

After a discussion of why Google won its case about building a search engine for books, including that it was built to be unable to produce no more than snippets, the article says:

"Ultimately, the fate of these companies may depend on whether judges feel that the companies have made a good-faith effort to color inside the lines. If generative models never regurgitated copyrighted material, then defendants would have a compelling argument that it is transformative. The fact that the models occasionally produce near-perfect copies of other people�s creative work makes the case more complicated and could lead judges to view these companies more skeptically."

Google is now also playing in the AI league.
With the Gemini model Google is attacking at the forefront. Gemini is a multimodal model that can process text, images, video, music and other documents such as PDFs as inputs and generate corresponding outputs.

Google offers a cloud-based API platform for using and developing your own AI models, which can be created from the smaller Gema models.
Even some third-party open source models can be used.

In my opinion, Google created a strong overall package of AI models, API and cloud environment.
Despite the current problems I would assume that Google has a big chance to take the first place from Open AI in the long term.

https://blog.google/technology/developers/gemma-open-models/
https://deepmind.google/technologies/gemini/#introduction
https://blog.google/products/ads-commerce/gemini-models-are-coming-to-performance-max/
https://blog.google/technology/ai/google-labs-imagefx-textfx-generative-ai/

Free testing of Google's AI Image Generator: https://aitestkitchen.withgoogle.com/tools/image-fx

Quote from: Andrej.S. on February 28, 2024, 16:36

Google is now also playing in the AI league.
With the Gemini model Google is attacking at the forefront. Gemini is a multimodal model that can process text, images, video, music and other documents such as PDFs as inputs and generate corresponding outputs.

Google offers a cloud-based API platform for using and developing your own AI models, which can be created from the smaller Gema models.
Even some third-party open source models can be used.

In my opinion, Google created a strong overall package of AI models, API and cloud environment.
Despite the current problems I would assume that Google has a big chance to take the first place from Open AI in the long term.

https://blog.google/technology/developers/gemma-open-models/
https://deepmind.google/technologies/gemini/#introduction
https://blog.google/products/ads-commerce/gemini-models-are-coming-to-performance-max/
https://blog.google/technology/ai/google-labs-imagefx-textfx-generative-ai/

Free testing of Google's AI Image Generator: https://aitestkitchen.withgoogle.com/tools/image-fx

Being Woke and diverse has created the opposite of what an honest AI would do.

"It appears that in trying to solve one problem - bias - the tech giant has created another: output which tries so hard to be politically correct that it ends up being absurd."

"Gemini also generated German soldiers from World War Two, incorrectly featuring a black man and Asian woman."

https://www.bbc.com/news/technology-68412620

Reverse Bias is still bias. While trying to compensate for what the AI got from general images, they have created a mockery.

"When the AI was asked to show a picture of a White person, Gemini said it could not fulfill the request because it "reinforces harmful stereotypes and generalizations about people based on their race." 

Yes, I read that, just like Gemini couldn't answer the question of whether Elon Musk's meme tweets or Adolf Hitler did more harm to the society.
https://gemini.google.com/share/231396168632?hl=en

Google is currently lagging far behind Open AI, so they had to rush and tried to play it safe on all potential ethical issues, because it takes much time to adjust the algorithm for appropriate answers. I mean political correctness is such a complex thing that even many humans fail. In addition, we currently have political "overcorrectness" and minorities are disproportionately "favored", so you just can loose nowdays if you start from the scratch and rush.
But as long as the advantages for professional applications like chemistry, medicine, automotive, etc. outweigh, nobody will talk about such cases in the future.
 
For the image generator it would be easy just to block all such prompts by a giving error of an inappropriate request just like for example Microsoft's Bing image creator does.

////Edit:
Found this demo of Gemini's model capabilities.
https://www.youtube.com/watch?v=UIZAiXYceBI

If this isn't staged at all, it is really impressive.
Reminds me of the TARS robot from the Interstellar movie.

////Edit2:
What about the AI google ads capabilities?
https://youtu.be/kyTu3mgGfUA?t=78

Do you really believe people will still search for photos on agencies, when they will use all-in-one AI suites in 3 to 4 years?
People won't care about less quality that much. They will use it because they can save time and earn more money by just being more productive.