Aurora, the new AI image generator from Elon Musk’s xAI was quickly taken down over the weekend after being made live. It is planned to be baked into Grok, X’s (formerly Twitter) generative AI chatbot, and has already drawn attention for its capabilities.
X hasn’t officially commented on the removal of Aurora, with speculation being that it was made live too early. Elon Musk simply said it was a “beta” and “will improve very fast”.
However, it might be drawing more attention thanks to limited guardrails around the software. As with Grok’s previous Flux AI creator, it appears Aurora is quite capable of generating copyrighted materials – as well as celebrities.
Muy impresionante Aurora, la función hiperrealista de Grok para generar imágenes por IA con rostros conocidos pic.twitter.com/cdq9eYAVuq
— Ricardo Sametband (@rsametband) December 10, 2024
Multiple images were posted by users who managed to get in before Aurora was pulled. Some depict Elon Musk in the army or as Thor. Two users found that it’s exceptionally good at recreating Taylor Swift, Jeff Bezos, and other tech luminaries.
This shouldn’t come as a surprise, as these massive AI language models are trained on huge quantities of data. In the last few months, companies like Nvidia were found to be harvesting YouTube and Netflix for sources. OpenAI, creator of ChatGPT, has said it’s “impossible” to train these models without touching on copyrighted material.
However, X pushed the limits with Grok, as the platform became flooded with copyrighted images on Flux AI’s initial release.
Musk flexes X’s AI capabilities with AuroraxAI's Grok 2 Stuns with Ultra-Realistic AI Images
xAI just made waves by releasing Grok 2, paired with Aurora their groundbreaking AI image model (though quickly pulled offline).
The level of photorealism achieved is absolutely mind-blowing