On Wednesday, Facebook's parent corporation Meta unveiled an artificial intelligence (AI) model. Both a collection of picture annotations and specific image objects can be chosen by it. The largest model of its sort has ever been constructed. In a blog post, the company's research division claimed that its Say Anything model, or SAM, can recognize items in images and videos for which it was not specifically taught.
How does SAM works?
You can choose an object by clicking on it or by using SAM to enter text. The AI model in one example drew boxes around each of the cats in the image when the user typed the word cat. Big IT businesses are bragging about their AI triumphs now that OpenAI is having issues with ChtaGPT, and the competition for investment and supremacy in the AI market has already started.
There have been hints of numerous ChatGPT-like capabilities in Meta. Unlike other AIs, the AI model used by Meta produces entirely new content rather than merely identifying or categorizing data. It hasn't yet released any products, though. A tool that generates pictures for stories in a children's book and another that creates a very lifelike video from text input are two examples.
Meta talks about SAM
The SAM model and dataset are available for download from the company under a non-commercial licence. Users of the accompanying prototype who add their own photographs certify that they will only use the tool for research. "In the future, SAM could be utilised to help power applications in many domains that require to locate an object in any given image," wrote Meta in a blog post. SAM can be integrated into bigger AI systems for a more comprehensive multimodal understanding of the world, such as comprehending both the visual and linguistic content of a webpage, according to the AI community and others. SAM could give users in the AR/VR space the ability to choose an object based on their gaze and then "raise" it in 3D.
