What to expect from Artificial Intelligence in 2023? Art Generator not seen well by global artists

The future of artificial intelligence for 2023 is a real mystery. This is because the US regulation on the subject intends to control the worst that exists in this segment. Additionally, incredible and transformative new forms of AI are expected to arrive during this year, but these are all still industry hypotheses.

In 2022, with the success of the Lens, an AI-powered selfie app developed by Primas Labs, many more AI-powered apps are expected to follow. The market even believes that they are capable of being tricked into creating NSFW images to sexualize and alter the appearance of women.

In this regard, senior policy researcher at the Mozilla Foundation, Maximilian Gahntz, points out that the integration of these applications with generative artificial intelligence can amplify the effects of these systems, whether that is a good thing or a bad thing.

As an example, we have Stable Diffusion, which was fed with millions of images from the internet until it was able to associate certain concepts and words with certain images. This is because current text generation models are easily tricked into producing deceptive or fraudulent content.

VEHICLE ACCESSORIES - FREE MARKET

Thus, a member of the open research group Knives and Paintbrushes, Mike Cook, claims that generalized artificial intelligence will be a problematic force of change in 2023. On the subject, he pointed out that “It is not enough to motivate a community of experts [to create new technology] – for technology to become a part of our lives in the long term, it must either make someone a lot of money or have a significant impact on everyday life. of the general public”

Artists Lead the Effort to Deactivate Datasets

At the end of 2022, the DevianArt released an artificial intelligence art generator, developed in Stable Diffusion, and tweaked with art from its own community. For former members of DeviantArt, the art generator was not looked upon favorably, who criticized the platform's lack of transparency in using uploaded art to train the system.

In this sense, the Breeders of these generators say they have taken steps to limit the amount of harmful content their systems produce. However, looking from generations on social media, it is exposed that there are still obstacles to be fought.

On the subject, one of the heads of Stability AI, responsible for the generator, stated that “The datasets require active curation to resolve these issues and must be subjected to significant scrutiny, including from communities that tend to stay on the short end of the spectrum. of the baton”. For him, there are controversies regarding the control that exists in social media. 

Additionally, after pressure from the public, Stability Al signaled that it allowed artists to opt out of the dataset used to reignite the next-generation Stable Diffusion model. Thus, through the HaveIBeenTrained.com website, rights holders will be able to request cancellations before training begins in a few weeks.

Finally, it is worth mentioning that OpenAL, also responsible for the artificial intelligence of DevianArt, does not allow this exclusion mechanism, preferring to partner with organizations such as Shutterstock to license parts of its image galleries. However, given the issues it has with the Stability AL, it is likely that it will quickly offer users the opt-out option.