OpenAI announced today that the research laboratory is removing the waitlist for its DALL-E beta, allowing anyone to sign up — citing improved safety systems and lessons learned from real-world use. In addition, it is testing a DALL-E API with several customers and says it plans to soon offer it more broadly so that developers and customers can use it to build apps on DALL-E.
The announcement comes in the context of a growing number of startup players offering more accessible text-to-image AI generators, including Midjourney, which was released in open beta in mid-July, as well as Stability AI’s open-sourced Stable Diffusion, which was released in August (its new CIO, Daniel Jeffries, wrote a new blog post today saying that the company is “putting the open back in AI”).
According to a newly released blog post, Open AI said “responsibly scaling a system as powerful and complex as DALL·E – while learning about all the creative ways it can be used and misused – has required an iterative deployment approach.”
OpenAI noted that there are currently more than 1.5 million users creating over 2 million images a day with DALL-E, with about 100,000 users sharing their creations and feedback in its Discord community.
MetaBeat will bring together thought leaders to give guidance on how metaverse technology will transform the way all industries communicate and do business on October 4 in San Francisco, CA.
New DALL-E features and more robust filters
Artists, in particular, have provided important input and feedback since DALL-E 2’s paper was announced in April, the blog post said, leading to features such as Outpainting, added in August, which lets users continue an image beyond its original borders and create bigger images of any size.
According to OpenAI, in recent months it has made its filters more robust in order to “reject attempts to generate sexual, violent and other content that violates our content policy, and building new detection and response techniques to stop misuse.”
When DALL-E 2 was released, it elicited rhapsodic responses for its use of advanced deep-learning techniques to generate and edit photorealistic images simply by comprehending text instructions. There was suddenly a sharing tsunami of images of avocado-shaped teapots and chairs, as well as loud concerns about biased datasets that reinforced stereotypes about women and people of color.
But OpenAI maintains that “lessons learned from deployment and improvements to our safety systems make wider availability possible.”
Thorny DALL-E copyright questions remain
However, thorny questions about DALL-E image ownership remain: In July, when OpenAI announced it would expand beta access to DALL-E 2, it also offered those users full usage rights to commercialize the images they create with DALL-E, including the right to reprint, sell, and merchandise.
According to OpenAI’s spokesperson, user feedback found that full usage rights are what creators want. OpenAI, however, retains ownership of the original image “primarily so that we can better enforce our content policy.”
But creative workers, as well as legal experts, find the issues around ownership and copyright to be unclear.
Bradford Newman, who leads the machine learning and AI practice of global law firm Baker McKenzie, in its Palo Alto office, said the answer to the question “Who owns DALL-E images?” is far from clear. And, he emphasized, legal fallout is inevitable.
“If DALL-E is adopted in the way I think [Open AI] envisions it, there’s going to be a lot of revenue generated by the use of the tool,” he told VentureBeat in August. “And when you have a lot of players in the market and issues at stake, you have a high chance of litigation.”
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.