On Wednesday, Facebook-owned Meta released a dataset of image annotations that it claimed was the largest ever of its kind combined with an artificial intelligence model that can identify specific objects within an image.
In a blog post, the company’s research division claimed that its Segment Anything Model, or SAM, could recognise things in pictures and videos even when it hadn’t seen them during training.
SAM allows users to pick objects by clicking on them or by entering text commands. Writing the word “cat” in one demonstration caused the tool to draw boxes around each of the multiple cats in a picture.
Since Microsoft-backed OpenAI’s ChatGPT chatbot became a sensation in the autumn, sparking a surge of investments and a race to dominate the area, big tech companies have been bragging about their artificial intelligence advances.
Though it hasn’t yet made a product available, Meta has teased a number of features that use the generative AI made famous by ChatGPT. Which generates entirely new material rather than just detecting or categorising data like conventional AI.
Examples include a programme that creates surrealist videos from text prompts. And another that turns prose into pictures for children’s books.
According to Chief Executive Mark Zuckerberg, adding these generative AI “creative tools” to Meta’s products is a top focus for this year.
Internally, Meta does make use of SAM-like technologies for tasks including tagging photographs, removing objectionable information. And choosing which articles to promote to Facebook and Instagram users.
The corporation said that the introduction of SAM will increase access to that kind of technology.
A non-commercial licence will be required to obtain the SAM model and dataset. Users who upload their own photographs to an associated prototype must also consent to use them strictly for study.
Visit CxO Global FORUM or CxO News Live for all the latest updates.
