So check it out, OpenAI is dropping some big news today. They’re expanding access to their new text-to-image generator, the DALL-E 3 model. This bad boy is gonna be available for ChatGPT Plus and Enterprise customers, straight from the ChatGPT app. OpenAI claims they’ve put together a safety mitigation stack to make sure this release is ready for prime time.
Now, if you’re not up to speed, DALL-E 3 was announced last month. And let me tell ya, it’s an improvement over the previous DALL-E 2. This thing allows users to get all creative with their prompts in ChatGPT, so they can write longer and more visually descriptive cues for generating images. And get this, Microsoft’s platform actually beat ChatGPT to the punch, introducing DALL-E 3 to the wider public through Bing Chat and Bing Image Generator. They’ve got that public access game locked down, man!
But let’s not forget, there have been some bumps in the road when it comes to these text-to-image generators. Users have generated some questionable stuff, my friends. We’re talking images of the World Trade Center with SpongeBob SquarePants flying planes towards it. Yeah, that’s not what we wanna see. And even after Microsoft tried to block certain prompts, people found workarounds and kept producing similar results. So, it’s been a bit controversial, to say the least.
We’ve seen it before with other generators like Midjourney and Stable Diffusion, and even previous versions of DALL-E. Copyright infringement, nonconsensual nudes, altering ethnicities of subjects, and creating photo-realistic misrepresentations of public figures. Yeah, these things have caused quite a stir.
But here’s the deal, OpenAI is assuring us that they’ve really stepped up their game this time. They’ve put in some serious effort, my friends. They even have a website showing off all the research and hard work that went into DALL-E 3. They’re promising to limit the model’s generation of content in the style of living artists and images of public figures. And hey, they’re also aiming to improve the diversity of the generated images. Gotta respect that, right?
Oh, and they’ve got this fancy tool called the “provenance classifier” that can supposedly detect with 99 percent accuracy if an image was created by DALL-E 3. Now that’s some advanced tech right there.
So, here we are, OpenAI making big strides in text-to-image generation. They’re taking the necessary precautions and setting higher standards to avoid any sketchy or controversial stuff. Let’s see how this new DALL-E 3 model performs, my friends.