Why is the release of image generation AI ``Stable Diffusion'' version 1.5 delayed?

In recent years, image generation AI that can output a desired image simply by inputting text has been attracting a great deal of attention. Among the many image generation AIs, ``Stable Diffusion'' developed by AI startup Stability AI has been open sourced in August 2022, and has a very active community, such as the development of multiple GUI applications . Born. Daniel Jeffries, Chief Information Officer of Stability AI, explains why the release of version 1.5 of Stable Diffusion is delayed.

Why the Future of Open Source AI is So Much Bigger Than Stable Diffusion 1.5 and Why It Matters to You
https://danieljeffries.substack.com/p/why-the-future-of-open-source-ai

The version of Stable Diffusion released at the time of writing the article is 1.4, and the release of version 1.5 with enhanced functions is awaited. However, since the information on version 1.5 has not been released even after several months have passed since the release of version 1.4, there are voices of suspicion from the community that `` Stable Diffusion will no longer be open source ''. increase. Regarding the rumor that Stable Diffusion will be unpublished, Mr. Jeffries said, ``It is not true.We are working on open source as a central element of Stable Diffusion,'' and Stable Diffusion will continue to open Emphasis on staying the source.

According to Jeffries, Stable Diffusion is ' brain scans for medical research, ' ' interior design ,' ' Hollywood-style film effects, ' ' seamless textures for games, ' 'a new kind of fast animation, ' and ' plugins for Photoshop and Figma. ' It is said that it is used in various fields such as. Mr. Jeffries pointed out that the scope of utilization of third-party image generation AI developed with closed source is limited, and said, ``Software (developed with open source) like Linux has value to society as a whole. History has proven that it is the best way to provide technology and a source of innovation.” I am appealing.

As mentioned above, Stable Diffusion is used in a variety of fields, but at the same time, it is pointed out that there are many works that contain explicit sexual expressions and rights issues of the data set from which Stable Diffusion was learned. It has been. According to Jeffries, Stability AI has organized an open source committee and is working to establish policies on sexual expression and guidelines on model data. Mr. Jeffries said, ``If you don't respond to what the community has pointed out, the open source AI project itself may disappear,'' and the release of version 1.5 is to establish guidelines to deal with various problems related to image generation AI. I explained that I was late.

Criticism that ``companies are escaping legal liability by leaving the creation of datasets for AI learning to universities and non-profit organizations''-GIGAZINE

Stability AI not only established an open source committee and is proceeding with the establishment of guidelines, but also opened a bounty program of $ 200,000 (about 30 million yen) to deal with deep fakes. Mr. Jeffries said, ``Please cooperate to make AI truly open, not just a name,'' calling for cooperation from the community.