Stability AI uses latent diffusion models to allow users to create art in Stable Diffusion
Stability AI, an artificial intelligence (AI) startup based in London, raised $101 million in funds for its company. The seed round was led by Coatue, Lightspeed Venture Partners, and O'Shaughnessy Ventures LLC.
Parent company to Stable Diffusion
The company also runs Stable Diffusion, an open-source text-to-image generator that launched in August, and it works in a similar way to other AI systems such as DALL-E 2 by Open AI and Designer by Microsoft due to the fact that it has the ability to generate a high-quality image from a text description. This AI tool uses machine learning to generate images from text. According to the company website, the platform Stable Diffusion has been downloaded by more than 200,000 users from around the world.
Technology behind Stable Diffusion
The machine learning used behind Stable Diffusion is unique because it generates images using latent diffusion models (LDMs). The LDM uses diffusion methods, but instead of only compressing an image, it reconstructs the image. Although Stable Diffusion is an image generator, like the previously mentioned systems, it does not use an auto-regressive technique used by tools such as DALL-E. Auto-regressive techniques use probability distribution to create the output of images. In other words, it processes images randomly based on previous values entered. The past values that were previously entered into the software will affect the present and future values.
However, in the case of LDM, images are generated by denoising data of neural networks called autoencoders, in a latent representation space, which is information needed to represent specific data that’s embedded close together. Then, the representation is decoded to form the full image. This is how the pictures are generated.
The funding will be used to advance AI models within the digital media firm
Stability AI will use the funding to further advance the creation of AI art within its company. It will be used to “accelerate the development of open AI models for image, language, audio, video, 3D, and more, for consumer and enterprise use cases globally,” the company stated in a press release.
The world’s first community driven open-source AI
The digital media firm says that it is the “world’s first community-driven, open-source artificial intelligence company.” It believes in allowing everyone to have access to the technology that lets users create digital AI art. The company not only wants to help creators but is also looking ahead to make innovative applications that can be used in the future.
Potential for AI
“AI promises to solve some of humanity's biggest challenges. But we will only realize this potential if the technology is open and accessible to all,” said Emad Mostaque, founder and CEO of Stability AI.
“Stability AI puts the power back into the hands of developer communities and opens the door for ground-breaking new applications. An independent entity in this space supporting these communities can create real value and change,” he continued.
DreamStudio and Stability AI’s unique value
The company also operates DreamStudio, an AI system that allows users to create realistic art from a description. So far, the system has over a million registered users and has generated over 170 million pictures.
Stability AI is unique because of its open-source technological model that allows creators to produce different applications related to various art forms, including advertising, design and augmented reality. The company also collaborates with other AI developer communities, making it a group participation amongst everyone involved. The opening site for Stability AI says, “AI by the people, for the people,” using augmented technology and collective intelligence to design and create images.
Accessibility for everyone
The investment firms funding Stability AI mentioned the value in making artificial intelligence accessible to everyone. Gaurav Gupta, a partner at Lightspeed Venture Partners, one of the investment firms that contributed funding, stated, “we believe that an independent company like Stability AI is best positioned to democratize generative AI.”
Sri Viswanath, general partner at Coatue, another investor in Stability AI, stated that “open-source AI technologies have the power to unlock human creativity and achieve a broader good.”
AI images seem to be the digital art of today, and of the future. The technological advancements prove that combining AI and machine learning with imaginative users’ words can really produce something amazing and innovative, but also familiar in terms of imagery for the creators.