The Wall Street Journal (WSJ) reported on the 18th that Meta, the parent company of Facebook and Instagram, is developing artificial intelligence (AI) models for image and video generation. It is seen as a challenge to Google's Nano Banana, which has been praised as "no need for Photoshop," and OpenAI's Sora. With Meta joining Google, OpenAI, and xAI of Elon Musk, the CEO of Tesla, the race for leadership in the AI image generation market is heating up.
Alexander Wang, Meta's chief AI officer (CAIO), said during an internal Q&A session that the company is developing an AI image and video generation model with the code name Mango, WSJ reported, citing sources familiar with the matter. Wang added that in addition to Mango, Meta is also preparing an AI model based on a large language model (LLM) with the code name Avocado. The models mentioned are set to be unveiled in the first half of next year.
Wang named "improving coding ability" as one of Avocado's main goals. Unlike Llama, Meta's open model that released information such as weights until recently, these models are expected to be launched as closed models.
He also said the company is conducting research on so-called world models that understand the physical world. A world model is a next-generation AI model in which AI directly observes the real world and predicts and reasons according to the laws of physics.
This summer, as Meta reorganized its AI division, it recruited Wang, the founder of Scale AI, to lead the newly established superintelligence lab, Meta Superintelligence Labs.
Wang, a Scale AI founder who became Meta's CAIO in his twenties, was brought in as Mark Zuckerberg, Meta's CEO, set "superintelligence" that surpasses humans as the company's AI goal. To strengthen Meta's AI capabilities, CEO Zuckerberg personally led large-scale hiring and recruited 50 AI experts, including more than 20 OpenAI researchers.
Big tech's AI competition is expanding from text to images and video. Google recently launched its image generation and editing tool Nano Banana Pro, and OpenAI hit back by releasing GPT-Image 1.5. OpenAI also unveiled Sora, which specializes in video generation.