OpenAI is an AI research and deployment company that creates multiple AI models that serve all kinds of purposes. To give an example, an algorithm that writes an entire essay with only one or two sentences as input (Beta OpenAI, 2022).
Their latest finding was DALLE 2 which is available since April 2022. This new AI system can create unique and realistic images from a description in natural language using 100 percent AI. All the created images never existed before. If the input is for example “An award-winning professional photo of a smiling marmot skiing in the Alps in winter” DALLE 2 returns the following photos below (OpenAI, 2022).
The CLIP language model, which is used by DALL.E2, enables connections between images and text descriptions. DALL.E2 begins to provide a rudimentary intermediate solution that, according to CLIP, includes the key visual elements of the text description. The intermediate solution is then enhanced by DALL.E2 using a diffusion model until the image completely complies with the CLIP description. Diffusion models are “image enhancers” that learn to recreate the original image by first adding random pixels to images during training (OpenAI, 2022).
DALL.2 surpasses systems that produce deep fake images. The system appears to comprehend the links between items. For the first time ever, this enables the meaningful combination of various concepts (such as marmot, skiing, Alps) in an image. The system can also be used to edit existing images (Arnold V, 2022).
To get the most out of the system a very clean and detailed explanation is required. The system can be used for many purposes. For example, to visualize an idea such as the architecture of a building or a movie set. But it can also create designs like wall art or a cover of an album. The photos are unique and of high-quality (Arnold V, 2022). Some are even more beautiful than any professional photographer can shoot. Besides that, it can adjust existing photos. Because everything is possible in the photos abuse of the system can have negative consequences. Therefore, rules are set like no content including nudity, faces, politics, or violence (Arnold V, 2022).
Personally, I think this new system is interesting and smart. It has a lot of potentials to create value for multiple purposes. However, I am curious about the long-term impact it will have. Is the future of professional photographers questioned? are artists becoming redundant?
References
Arnold V. (2022). DALL-E 2: The new text to image generator by OpenAI. Neuroflash. https://neuroflash.com/blog/dalle-2-open-ai/
Beta OpenAI (2022). Essay outline. Beta OpenAI. https://beta.openai.com/examples/default-essay-outline
OpenAI (2022). DALL·E 2 is a new AI system that can create realistic images and art from a description in natural language. OpenAI
.
Hi Liselot, thank you for your post! I find DALL-E 2 to be a very interesting new technology, as I believe it will truly disrupt the creative industry. However, I don’t think it will make designers, photographers and artists redundant anytime soon. Like you mentioned, DALL-E 2 requires human input, and I believe that this implies creative AI will be a complementary tool; they won’t replace humans (just) yet! What do you think? Would you agree?