ChatGPT has long won the race and got the title for the best text-generating AI based on the prompts. Though it excelled in text generation, the AI chatbot lacked image or art-creating ability. However, over time, the San Francisco startup found ways to connect its chatbot with other resources and tools available there. And recently, they finally came up with a way to generate images using ChatGPT.
Within a year of its release, ChatGPT has evoked a race among the Silicon Valley giants. They all started to compete in terms of AI and its advancements. Before this image-generation news about ChatGPT, Google released a new version of Brad, its chatbot. As we know, Brad connects itself with several other services from Google like Outlook, Docs, Gmail, and YouTube. Apart from Google, two other image generators also updated themselves recently, Stable Diffusion and Midjourney. However, let us get back to ChatGPT and know what’s new with it.
How did they make it happen?
Open AI recently released the DALL-E 3. It is the latest version of their image-generating AI that is more powerful and precise than its predecessors. Initially, ChatGPT and DALL-E were two different tools of Open AI.
Now, they just added the ChatGPT to the DALL-E 3 to create a hub for generative AI. The Open AI released it to a few testers to test the result of their incorporation of ChatGPT with an image generator, DALL-E. The results were shockingly detailed images created based on the prompts written in detail. The image generation followed the prompt closely and didn’t miss to include any of the mentioned requirements by the testers.
However, the audience has to wait for at least a month for Open AI to release DALL-E 3. At that time, DALL-E 3 will be accessed via ChatGPT with an additional charge of $20 monthly.
Though, Open AI previously had options to connect ChatGPT with other resources like Wikipedia, Open Table, and Expedia, this time it was different. This was their first attempt to incorporate an image generator with ChatGPT. Now, Open AI aims to make ChatGPT eligible to create not only texts and images, but also, videos, software, sounds, and several other digital media.
What do the experts say about ChatGPT generating images?
As mentioned above, the results of DALL-E 3 incorporation with ChatGPT provided us with great detailed images. And they were made by the AI following the multipara instruction line by line. However, it is bound to make minute mistakes just like any other AI system. This was a review by one of the Open AI researchers, Gabriel Goh, who tested the system as well.
According to another Open AI researcher, Sandhini Aggarwal, the generated images were more stylized than realistic. As a person who looks over the safety and policy, she aired her concerns for DALL-E results. She also mentioned that they need to tinker it more to get more realistic images and of course, the results will also depend highly on the place of its use.
Looking at this leap by Open AI to develop a powerful chatbot that can generate both text and related images on its own, are we moving towards something surreal? What are your thoughts?