Cloud computing
Artificial intelligence models trained on millions of images can occasionally reproduce near-perfect copies, researchers have found, which could be significant for ongoing copyright infringement lawsuits
Technology
6 February 2023
By Alex Wilkins
Left: an image from Stable Diffusion’s training data with the caption “Ann Graham Lotz”. Right: the image produced by Stable Diffusion when prompted with “Ann Graham Lotz”
Cornell University/Extracting Training Data from Diffusion Models
Do image-generating artificial intelligence (AI) models infringe artists’ copyright? With legal action already under way against some of the companies behind these AIs, researchers have now found that the generators can occasionally reproduce some of the specific images used to train them – a discovery that could play into the upcoming lawsuits.
Popular text-to-image generators like Stability AI’s Stable Diffusion and OpenAI’s DALL-E 2 use …
No commitment, cancel anytime*
Offer ends 14th March 2023. *Cancel anytime within 14 days of payment to receive a refund on unserved issues.
Existing subscribers, please log in with your email address to link your account access.
Inclusive of applicable taxes (VAT)
Cloud computing MORE FROM NEW SCIENTIST
Cloud computing Filling the internet with AI-created images will harm future AIs
The performance of image-generating AIs drops when trained using images previously created by other AIs, which may create problems for them given the proliferation of these images
Cloud computing How will AIs that generate videos from text transform media online?
Artificial intelligences that turn text into video are making rapid progress, but are we ready for a world in which the authenticity of every video is in question?
Cloud computing Why do AIs keep creating nightmarish images of strange characters?
Image-generating AIs seem to produce mythical characters that look consistently recognisable – but their appearance has a rational explanation
Cloud computing Mistakenly calling AIs ‘sentient’ is more dangerous than we think
A Google engineer recently claimed an AI was alive and that it had hired a lawyer. If judges were to accept these claims, it could lead to AIs being frozen in their biased states, writes Annalee Newitz
