This model, I-JEPA, uses background knowledge about the world to fill in gaps in images, instead of just looking at nearby pixels like other generative AI models.
Metaplatforms announced Tuesday that it will give researchers access to components of its new “human-like” artificial intelligence model. The model can analyze and complete unfinished images more accurately than existing models.
According to the company, the model, I-JEPA, uses background knowledge about the world to fill in gaps in images, rather than just looking at nearby pixels like other generative AI models.
The approach incorporates human-like reasoning, as advocated by Meta’s top AI scientist Yann Lucan, to help the technology avoid common errors in AI-generated images, such as extra fingers on the hand. It is said to be useful.
you’re exhausted
Limited number of free calls per month.
To read more,
Simply register or sign in
subscribe and read more
Please select a plan
all access
access to premium story
digital only
access to premium story
This premium article is free for now.
Subscribe to continue reading this story.
This content is for subscribers only.
Subscribe for unlimited access to premium articles exclusive to The Indian Express.
This content is for subscribers only.
Subscribe now for unlimited access to premium articles exclusively from The Indian Express.
Meta, which owns Facebook and Instagram, publishes a ton of open-source AI research through its in-house labs. CEO Mark Zuckerberg shares models developed by Meta researchers to help the company drive innovation, identify safety gaps and reduce costs said it can.
“For us, it would be much better if the industry standardized on the basic tools we use and could benefit from improvements others made,” he told investors in April.
Company executives ignored warnings from industry insiders about the technology’s potential dangers, and last month saw top executives from OpenAI, DeepMind, Microsoft and Google equate the risks to their companies with pandemics and war. refused to sign a statement endorsed by
Considered one of the “Godfathers of AI,” Lekunn lashed out at “AI fatalism” and argued that he was in favor of building safety checks into AI systems.
Meta is also beginning to incorporate generative AI capabilities based on text prompts into its consumer products, such as an advertising tool that can create image backgrounds and an Instagram product that can change a user’s photo.
Date first published: Jun 14, 2023 09:29 IST
