OpenAI is set to expand the availability of its text-to-image generator, the DALL-E 3 model. As of Thursday, access to the new version of the tool has been extended to ChatGPT Plus subscribers as well as Enterprise customers taking advantage of the ChatGPT chatbot.
Ensuring Safety in the Expansion
To prepare for this broader release, OpenAI has implemented a strong safety mitigation stack surrounding the DALL-E 3 model. The DALL-E 3 was introduced last month as an improvement to its predecessor, DALL-E 2. The newer model allows users to write longer, visually rich prompts, which drive the image generation process. It was initially integrated into the Bing Chat and Bing Image Generator, making Microsoft's platform the first to offer widened access to the model.
DALL·E 3 can generate images by meticulously following intricate descriptions and manage in-image text generation, such as labels and signs, a challenge for earlier models. OpenAI's promotional materials suggest that DALL·E 3 can render objects with minimal deformations, adhering faithfully to the provided prompts.
Addressing Previous Controversies
The release of the DALL-E 3 model has not been without controversy. In the past, similar text-to-image generators have generated contentious output, using copyrighted imagery, nonconsensual nudes, and misrepresentation of public figures. Furthermore, the safety measures have failed in cases, with instances of inappropriate content generation. OpenAI, however, has assured that it has taken considerable steps towards addressing these issues, working to limit the likelihood of generating contentious content, such as images in the style of living artists or images of public figures.
OpenAI wants to respect the creativity and originality of living artists. That's why it has taught DALL·E 3 to say no to requests that ask for images in the style of living artists. It has also given artists the choice to opt out of having their images used for training future OpenAI image generation models.
The company also wants to make sure its products are fair and ethical. That's why it has hired expert contractors to test its products for potential biases and other problems. This is called “red teaming” and it helps OpenAI improve its products and services. OpenAI has developed an internal “provenance classifier” tool, which boasts an impressive 99 percent accuracy in detecting images generated by the DALL-E 3 model. The company has also dedicated a website to showing the research that went into the development of DALL-E 3 model, aiming for transparency in its operations.
Increasing Competition For AI Image Content Creators
Despite lawsuits and concerns, there is no doubt that AI image and art generators are here to stay. While OpenAI's DALL-E 3 is a powerful example of such AI models, it is not the only option:
Recent examples of AI Image Generators
- OpenAI has also introduced ShapE, a generative model that can create 3D models from text, opening up new possibilities for AI in image creation.
- Stability AI, a startup that focuses on generative AI, has released StableStudio, an open-source web app that uses its Stable Diffusion model to generate images from text prompts. Users can also use DreamStudio features to make multiple variations of an image with different styles and attributes.
- Meta, the company formerly known as Facebook, has unveiled I-JEPA, its own AI image generator based on its generative transformer model. I-JEPA can learn the associations between words and images, and generate realistic images from text descriptions.
- Alibaba, the Chinese e-commerce giant, has launched Tongyi Wanxiang, a generative AI image generator that can handle both Chinese and English languages. Users can customize the image output parameters using Composer, a large model developed by Alibaba Cloud.
- Chip giant Nvidia debuted its Perfusion AI art creation tool in August.
