OpenAI’s Latest Upgrade to DALL-E Pushes Boundaries of Generative AI
Generative AI software like OpenAI’s ChatGPT isn’t just for quickly answering your questions. Some of these programs can also create images based on our prompts, and we have seen amazing results from software like Midjourney. However, it looks like OpenAI is already working on an upgrade to its DALL-E image generator language model, with capabilities that exceed what is currently possible with the last version of Midjourney.
What is OpenAI’s DALL-E 3?
Unfortunately, DALL-E 3 is still in the works, and it’s only available to around 400 users, according to one YouTuber. When OpenAI does finally launch DALL-E 3, it might lose some of its current capabilities that OpenAI is using for training purposes.
YouTuber MattVidPro AI got in touch with a person currently testing this secret version of DALL-E, sharing a few examples of what it can do.
The Amazing Capabilities of DALL-E
The clip at the end of the post shows amazing progress between DALL-E test versions from May and July, with the latter samples including AI-generated images that have absolutely no censorship rules.
As you’ll see in the clip, you can use this secret DALL-E tool to create images as accurately as photos. Or pictures that resemble paintings, down to the brush strokes. This secret DALL-E version can also use gore, blood, and copyrighted content.
The YouTuber called the upgrade “extremely exciting. This blows anything we’ve seen before out of the water. It’s insane.”
Earlier in the clip, Matt teased that “Midjourney cannot compete at this level. I don’t even think Midjourney version six would be able to compete at this level.” That was before showing the audience samples from the new DALL-E model.
The results are mesmerizing, proving that we’re only in the early days of AI-generated imagery. The photos are incredible, with the DALL-E model understanding prompts perfectly. It’s also able to generate text correctly most of the time, and the secret generative AI software makes impressive progress with the generation of hands.
The video also includes a comparison to Midjourney’s latest model, which shows it’s nowhere near as sophisticated as the secret version of DALL-E. OpenAI’s software can create images that stick to the rules in the prompt much better than Midjourney can. That’s a key differentiator, with image quality being another.
One such impressive image is the one above. We have a painting of a zombie Jesus Christ eating a sandwich from Subway. First of all, I would obviously watch such a TV show on Netflix, as long as humans created it without the help of generative AI. I’m assuming this zombie Jesus Christ would also be a fan of Subway. Or that he’d be really hungry for real food before going for pieces of flesh from humans.
Future Expectations and Competition
The point here is that OpenAI’s DALL-E might deliver a massive upgrade over what’s currently possible with generative AI software for images.
However, it’s unclear when and if this DALL-E 3 upgrade will drop. And when it does, expect it to feature several restrictions. It might not feature as much blood and gore, if any. And it might not let you create images of copyrighted content. Or reproduce works of art. I’m just speculating here, but it’s almost certain that OpenAI will implement various safeguards in the future DALL-E release.
Also, this exciting demo just put Midjourney, Microsoft, Google, and Adobe on notice. They’ll certainly have to come up with their own upgrades for their DALL-E rivals. According to the same YouTuber, the generator of the images in the clip said he or she has “zero interest in using Midjourney after using this.”
You should check out MattVidPro AI video below in full, as it contains plenty of images generated with OpenAI’s secret DALL-E upgrade and comparisons to Midjourney.
Editor Notes
OpenAI’s latest upgrade to its DALL-E image generator language model is truly groundbreaking. The capabilities demonstrated by this secret version of DALL-E are a giant leap forward in generative AI software. It’s incredible to witness the progress being made in AI-generated imagery, and the potential for creating highly realistic and accurate images is both exciting and promising.
As OpenAI continues to develop DALL-E 3, it’s important to recognize the potential limitations and restrictions that may be implemented to ensure responsible use of this technology. Copyrighted content, violence, and explicit imagery are likely areas where safeguards will be put in place.
This demo also highlights the need for competitors like Midjourney, Microsoft, Google, and Adobe to step up their game and develop their own upgrades to compete with DALL-E. It’s an exciting time for generative AI, and we can’t wait to see what innovations the future holds.
For more news and updates on AI technology, visit GPT News Room.
