AI Image Generation Pushes the Boundaries of Innovation and Ethics
Procreate vows not to use GenAI in its product offerings. xAI releases Grok-2 with minimal ethical safeguards. Read about these and other AI updates in this roundup.
Artificial intelligence (AI) is increasingly becoming an integral part of individuals’ and organizations’ daily lives. The technology has also made inroads into image and art generation, with several design and art generation platforms actively incorporating it. This has led to a massive hue and cry in the artist community, with some companies even facing lawsuits.
Amidst this ongoing tussle between artists and AI image-generation companies, here is a quick roundup of a few platforms’ latest releases and updates.
1. Procreate Vows Not To Use GenAI in Its Products
While most organizations seem to be incorporating AI into their tools and products, Procreate, the popular iPad design app, has vowed not to use generative AI (GenAI) in its products. On its website, the company said that despite machine learning (ML) being a compelling tech with a lot of merit, the current path GenAI is on is wrong for the platform. The company also said it would not chase a technology that threatens human creativity, though it may risk being left behind.
In a video posted on X, James Cuda, CEO of Procreate, said, “I don’t like what’s happening in the industry, and I don’t like what it’s doing to artists. We’re not going to be introducing any generative AI into our products. Our products are always designed and developed with the idea that a human will be creating something.”
We’re never going there. Creativity is made, not generated.
You can read more at https://t.co/9Fgh460KVu ✨ #procreate #noaiart pic.x.com/AnLVPgWzl3— Procreate (@Procreate) August 18, 2024
The company’s stance has gained widespread praise from digital artists, many of whom are unhappy with how digital art and illustration apps have embraced the technology. A few design apps and companies that have actively incorporated AI in their product offerings include Adobe, Jasper AI, and Midjourney. Adobe specifically came under fire after its updated terms of service, which seemed to imply that it would train AI models on users’ content. However, the company later clarified that it wasn’t true.
Given that Procreate’s announcement has gained widespread praise, it will be interesting to see if other platforms follow suit.
2. Midjourney Unveils a New Unified Image Editor on the Web
Midjourney has unveiled an updated version of its website containing a new editor interface that unifies its existing features, such as inpainting, outpainting/canvas extension, and more, into a single view. Further, the web editor has a new virtual brush-like tool for inpainting. The tool replaces the previous square selector and lasso tools, allowing for more precision when using AI to edit parts of an earlier generated image.
The web editor is now available to all users who created at least ten images on the platform.
Earlier, the various features could be accessed on the web nested under more disparate menus. According to a Discord message by Midjourney CEO David Holz, the goal of the new editor is to make editing AI generations easier and more seamless.
In addition to the editor and the tool, the company has released another feature to improve communication between its Discord and web communities. Messages sent in a few Web Rooms will now be mirrored in corresponding Discord channels and vice versa.
Besides these, Midjourney is opening its website to more users, even with limited free trials for people who haven’t already signed up. According to Holz, new users can generate about 25 free images. The initiative allows both new and existing users to explore the platform’s capabilities without an immediate financial commitment. It is also expected to attract diverse users, from casual creators to serious artists, improving the platform’s user base and engagement.
3. Flux Appears as the New Kid on the Image Generation Block
A new image generator named FLUX.1 has appeared on the scene and seems to be rapidly gaining kudos for its high-quality results and being an open-source model. The image generator was built by Black Forest Labs, whose founders worked earlier at Stability AI.
FLUX.1 has three versions, in increasing order of size: Schnell, Dev, and Pro. The Pro version is designed for commercial use, while the Schnell and Dev versions are optimized for various levels of speed and performance.
The primary difference between FLUX and Midjourney or other competitors is that Schnell and Dev can run on reasonably good hardware, making it accessible to a broader range of users. Further, people may not have to rely on cloud storage or internet access to run FLUX. The application also claims a few distinct advantages over Midjourney, especially when generating human figures.
If the device’s performance is good enough, people interested in using the image generator can download and run FLUX.1 locally on their computer. Enterprises can also integrate it relatively easily, as it is open source and compatible with multi-modal platforms.
See more: Artists Win Initial Phase of Copyright Infringement Lawsuit Against GenAI Companies
4. Grok-2 Allows X Users to Post Uncensored AI Images
Elon Musk’s AI company, xAI, has released two new language models, Grok-2 and Grok-2 mini. The models are connected to the FLUX image-synthesis model mentioned above and have minimal ethical boundaries. Grok also doesn’t seem to add watermarks to its outputs. This means X users can now use Grok-2 to create largely uncensored, potentially controversial, and even harmful AI images and share them on Musk’s social media platform.
The move comes against the backdrop of Musk batting for “freedom of speech” on his platform. Hence, it is highly likely that these capabilities will remain for some time.
Beyond generating controversial images, Grok-2 and Grok-2 mini are expected to be significantly advanced in their capabilities. Grok-2 supposedly outperformed a few leading competitors in recent benchmarks. According to xAI, both models showed improvements over Grok-1.5 in areas like general knowledge, graduate-level science knowledge, and math problem-solving. The company also claims high-quality results in document-based question answering and visual math reasoning. Further, the models boast high coding, chat, and image generation capabilities.
X Premium and Premium+ subscribers can access the models through an updated app interface. The company plans to release the models through an enterprise API later this month.
5. Google Releases a Free-To-Use AI Image Generator
Google has announced the release of a new AI image generator, Imagen 3. The generator, developed by Google DeepMind, is an improved version of Imagen 2. According to a company report, the image generator beats DALL-E 3, Stable Diffusion 3 Large, Stable Diffusion XL 1.0, and Midjourney v6 in performance in both human and automatic evaluations.
Imagen 3 works similarly to other AI image generators: Users type a prompt, and a few seconds later, pictures begin to appear. Google also offers an option to edit the images using the inpainting technique, which allows users to select a part of the image and type in the change they need. Further, unlike Grok-2 or Grok-2 mini, Google has placed restrictions on Imagen 3.
The AI image generator is available for free in the US. Users can access it via ImageFX and VertexAI. The tool will soon be available throughout Google AI features in Gemini and Workspace on mobile and the web.
These developments show the tough competition among AI image-generation companies and platforms to improve their offerings and attract diverse users. Simultaneously, they show how some companies are pushing the boundaries of ethics and innovation. As companies continue to explore the frontiers of this technology and find ways to harness AI’s potential, the development of models like Grok-2 accentuates the need to implement safeguards to protect against its misuse.
MORE ON ARTIFICIAL INTELLIGENCE (AI)
OpenAI Updates: Condé Nast Partnership and GPT-4o Fine-Tuning Initiative
MIT Unveils Comprehensive Database of Artificial Intelligence Risks
The State of California Partners With NVIDIA To Train Students, Workers, and Educators on AI
