Yesterday, Facebook-owned Meta unveiled the largest collection of image annotations ever, along with an artificial intelligence model that can recognize certain objects inside an image. The business’s research division said in a blog post that its Segment Anything Model, or SAM, could identify objects in images and videos even when it hadn’t seen them during training. Users of SAM may choose things by clicking on them or by typing instructions.
SEE ALSO: Tim Cook Likely To Visit India For Apple’s First Retail Launch In Mumbai In April
In one case, writing the word “cat” led the tool to create boxes around each of the several cats in the image. Big tech companies have been boasting about their artificial intelligence advancements since Microsoft-backed OpenAI’s ChatGPT chatbot became popular in the fall, igniting a wave of investments and a race to rule the space.
Although it hasn’t yet released a product, Meta has teased a few features that make use of the ChatGPT-famous generative AI, which creates whole new content rather than merely recognizing or categorizing input as traditional AI does. Examples include software that transforms language into visuals for children’s books and another that converts text into surrealist films.
SEE ALSO: Microsoft Employee Fired After Over Two Decades Of Service; Describes It As An Emotional Rollercoaster Ride
The addition of these generative AI “creative tools” to Meta’s products is a high priority for this year, according to Chief Executive Mark Zuckerberg. Internally, Meta does utilize SAM-like technology to categorize images, scrub content for offensiveness, and choose which articles to highlight for Facebook and Instagram users. According to the firm, SAM will make that technology more widely available. The SAM model and dataset must be obtained under a non-commercial license. Users must also provide permission for the related prototype to use any photos they send only for research purposes.