Meta’s AI, SAM capable of recognizing objects in images

Meta has launched a new artificial intelligence (AI) model to recognize individual objects within an image. The AI, called Segment Anything Model (SAM), is claimed to have the world’s largest dataset of image annotations. It can detect objects in images and videos, even if it hasn’t encountered them during training.

Meta’s research division announced in a blog post that the new AI, Segment Anything Model (SAM), could detect objects in images and videos. The company added that the tool could function even without encountering those items during its training. This is in tandem with the company’s declaration of 2023 as a “year of efficiency” where it will channel its efforts into generative AI.

How does SAM work?

SAM allows users to click on objects or write text prompts to select them. In a demonstration, the tool drew boxes around cars in a photo after a user wrote the word “cat.” Meta’s CEO Mark Zuckerberg has announced that the company plans to incorporate “creative aids” powered by generative AI into its apps this year.

Drawing from the explosive popularity of Open AI’s ChatGPT, the biggest tech companies have injected billions of dollars into generative AI. 

Although Meta has not yet developed a product, it has teased several features that use the type of generative AI pioneered by ChatGPT. CEO Zuckerberg recently announced that the company would add AI tools to Whatsapp, Instagram, and other platforms. Examples include a technology that makes surrealist movies from text prompts and another that creates children’s book illustrations from prose. 

Zuckerberg said the company wants to incorporate generative AI “creative aids” into its apps this year. The company’s research division focuses on generative AI as part of the company’s declaration of 2023 as a “year of efficiency.” Meta is already using SAM-like technologies internally for tasks such as tagging photos and choosing which posts to promote to Facebook and Instagram users.

Meta hopes SAM’s release will increase access to that kind of technology. The SAM model and dataset will be made accessible for non-commercial use. Users who contribute their photographs to an accompanying prototype must also agree to use it for research purposes solely.