As promised back in September, OpenAI has now rolled out access to its newest image generating AI model, DALL-E 3, to users of its ChatGPT Plus subscription service (starting at $20 monthly) and ChatGPT for Enterprise (variable pricing).
In a blog post announcing the news, OpenAI writes, “compared to its predecessor, DALL-E 3 generates images that are not only more visually striking but also crisper in detail. DALL·E 3 can reliably render intricate details, including text, hands, and faces. Additionally, it is particularly good in responding to extensive, detailed prompts, and it can support both landscape and portrait aspect ratios.”
In addition, as previously reported by VentureBeat, DALL-E 3 also offers the ability for users to generate text and typography baked into images, which is especially helpful for marketing, branding, and other business-related visual content such as promotional imagery or sales materials. In that way, it offers capabilities beyond some of the image generating AI competition including Adobe Firefly 2 and Midjourney.
OpenAI provided several examples of what people can use DALL-E 3 in ChatGPT Plus to do, including generate art for school projects and corporate logos.
In VentureBeat’s own tests of the DALL-E 3 ChatGPT Plus integration, which we’ve had for several days, one valuable feature it offers other over image generation services is the ability to have a conversation with the AI, asking it to alter images and move elements around or change them, without generating a whole different image or having the human user have to go in and manually edit sections. See the screenshot below for an example of how we used this feature to create this article art:
But that’s not all, OpenAI also today released a research paper on how it developed DALL-E 3 and said it was working on an image classifier that can reliably tell within 95-99% accuracy if an image was generated by DALL-E 3, a valuable tool for fighting AI-produced disinformation and propaganda, which has been on the rise in the last several days in the midst of the Irsaeli-Hamas conflict.
As OpenAI writes in their blog post:
“We’re researching and evaluating an initial version of a provenance classifier—a new internal tool that can help us identify whether or not an image was generated by DALL·E 3. In early internal evaluations, it is over 99% accurate at identifying whether an image was generated by DALL·E when the image has not been modified. It remains over 95% accurate when the image has been subject to common types of modifications, such as cropping, resizing, JPEG compression, or when text or cutouts from real images are superimposed onto small portions of the generated image. Despite these strong results on internal testing, the classifier can only tell us that an image was likely generated by DALL·E, and does not yet enable us to make definitive conclusions. This provenance classifier may become part of a range of techniques to help people understand if audio or visual content is AI-generated.”
The classifier is an especially interesting move, clearly an attempt by OpenAI to show its sense of responsibility for the products it creates and some of their more negative or harmful effects on society, but it follows OpenAI’s release and withdrawal of a classifier for AI-generated text, which OpenAI (and many researchers and users) concluded was ultimately not accurate enough — and inaccurately labeled human-generated text as AI generated, especially for those English-as-as-second-language writers.
There’s no word yet from the company on when the new AI image classifier will be released.
VentureBeat presents: AI Unleashed – An exclusive executive event for enterprise data leaders. Network and learn with industry peers. Learn More
As promised back in September, OpenAI has now rolled out access to its newest image generating AI model, DALL-E 3, to users of its ChatGPT Plus subscription service (starting at $20 monthly) and ChatGPT for Enterprise (variable pricing).
In a blog post announcing the news, OpenAI writes, “compared to its predecessor, DALL-E 3 generates images that are not only more visually striking but also crisper in detail. DALL·E 3 can reliably render intricate details, including text, hands, and faces. Additionally, it is particularly good in responding to extensive, detailed prompts, and it can support both landscape and portrait aspect ratios.”
In addition, as previously reported by VentureBeat, DALL-E 3 also offers the ability for users to generate text and typography baked into images, which is especially helpful for marketing, branding, and other business-related visual content such as promotional imagery or sales materials. In that way, it offers capabilities beyond some of the image generating AI competition including Adobe Firefly 2 and Midjourney.
OpenAI provided several examples of what people can use DALL-E 3 in ChatGPT Plus to do, including generate art for school projects and corporate logos.
Event
AI Unleashed
An exclusive invite-only evening of insights and networking, designed for senior enterprise executives overseeing data stacks and strategies.
In VentureBeat’s own tests of the DALL-E 3 ChatGPT Plus integration, which we’ve had for several days, one valuable feature it offers other over image generation services is the ability to have a conversation with the AI, asking it to alter images and move elements around or change them, without generating a whole different image or having the human user have to go in and manually edit sections. See the screenshot below for an example of how we used this feature to create this article art:
AI image classifier for fighting disinformation and propaganda
But that’s not all, OpenAI also today released a research paper on how it developed DALL-E 3 and said it was working on an image classifier that can reliably tell within 95-99% accuracy if an image was generated by DALL-E 3, a valuable tool for fighting AI-produced disinformation and propaganda, which has been on the rise in the last several days in the midst of the Irsaeli-Hamas conflict.
As OpenAI writes in their blog post:
“We’re researching and evaluating an initial version of a provenance classifier—a new internal tool that can help us identify whether or not an image was generated by DALL·E 3. In early internal evaluations, it is over 99% accurate at identifying whether an image was generated by DALL·E when the image has not been modified. It remains over 95% accurate when the image has been subject to common types of modifications, such as cropping, resizing, JPEG compression, or when text or cutouts from real images are superimposed onto small portions of the generated image. Despite these strong results on internal testing, the classifier can only tell us that an image was likely generated by DALL·E, and does not yet enable us to make definitive conclusions. This provenance classifier may become part of a range of techniques to help people understand if audio or visual content is AI-generated.”
The classifier is an especially interesting move, clearly an attempt by OpenAI to show its sense of responsibility for the products it creates and some of their more negative or harmful effects on society, but it follows OpenAI’s release and withdrawal of a classifier for AI-generated text, which OpenAI (and many researchers and users) concluded was ultimately not accurate enough — and inaccurately labeled human-generated text as AI generated, especially for those English-as-as-second-language writers.
There’s no word yet from the company on when the new AI image classifier will be released.
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.
Author: Carl Franzen
Source: Venturebeat
Reviewed By: Editorial Team
