Meta currently relies on AI to advance its development after its Metaverse goals failed. On Wednesday, Meta launched the Segment Anything Model. Segment Anything helps people find things in images with a few clicks. Segment Anything, which is still in demo phase, can take a snapshot and identify each pixel in the image so that one or more things may be isolated from the rest.
In unveiling the new model, Meta noted, “Segmentation—identifying which image pixels belong to an object—is a core task in computer vision and is used in a broad array of applications, from analyzing scientific imagery to editing photos. Meta stated technical professionals need AI training infrastructure and significant amounts of precisely annotated in-domain data to create reliable segmentation models for specific jobs.
“We collect an unprecedented dataset to achieve greater generalization than previous approaches.” Meta research scientist Ross Girshick emailed Decrypt. “Crucially, in this dataset, we did not restrict the types of objects we annotated. “Thanks to the scale of the data and its generality, our resulting model shows impressive capabilities to handle types of images that were not seen during training, like ego-centric images, microscopy, or underwater photos,” Girshick said.
Generative artificial intelligence develops text, graphics, and other media based on prompts. ChatGPT and Midjourney are examples of this technology. Meta claims the Segment Anything AI system was trained on 11 million photos. As Girshick mentioned, Meta is making Segment Anything available to researchers under a liberal open license, Apache 2.0, via the Segment Anything Github.
Lyle Solomon, Oak View Law Group’s principal attorney, told Decrypt that privacy regulations require data collecting to be open and with consent. AI face recognition without consent creates privacy legal concerns. Companies should also avoid sharing face data with third parties without authorization and follow privacy rules.” Girshick claims Segment Anything is under research and not for manufacturing. Potential AI usage raise privacy issues.
Meta launched a generative A.I. product division in February, abandoning its metaverse objectives. After eliminating Instagram NFT, the firm lay off nearly 10,000 staff. After OpenAI’s ChatGPT launched, global leaders were weary of artificial intelligence’s rapid advancement and began investigating its effects on user privacy and safety. Italy banned the famous chatbot.
“Many users do not understand how this process works or what the consequences can be long term if their face is used to train a machine learning model without their consent,” Kristen Ruby, president of social media and A.I. consultancy Ruby Media Group, told Decrypt.
“The biggest challenge many companies have is obtaining large-scale training data, and there is no better source than what people provide on social media networks,” she added. Ruby proposes investigating if a firm has a machine learning clause that tells consumers how their data is used and if they may opt out of future training models. She observes that many firms now have an opt-in default option but may switch to opt-out.
Girshick stated, “We have employed various privacy-preserving techniques, such as blurring faces and other personally identifying information (e.g. license plates). “Users can report offensive content by emailing [email protected] with the image id, and we will remove it from the dataset.”
