Zuckerberg took Instagram today to explain that Meta would incorporate the new Meta AI assistant, powered by Llama 3, into products including Whatsapp, Instagram, Facebook and Messenger.
Meta said in its blog post announcing Llama 3 that it had focused heavily on improving the training data used to develop the model. It received seven times more data than its predecessor, Llama 2, the company said. Some artificial intelligence experts pointed out that the figures published by Meta also showed that the creation of Llama 3 required enormous amounts of energy to power the necessary servers.
The growing capabilities of open source AI models have prompted some experts to worry that they could facilitate the development of cyber, chemical or biological weapons, or even become hostile toward humans. goal has released tools which he says can help ensure that Llama does not emit potentially harmful expressions.
Others in the AI field say that Goal Flame models are not as open as they could be. The company’s open source license on the models imposes some restrictions on what researchers and developers can build.
“It’s great to see more and more models openly releasing their weights,” Luca Soldaini, an applied research scientist at the Allen Institute for AI, a nonprofit lab, said in a statement after the launch of Llama 3. “But the open community you need access to all other parts of the AI process: your data, training, logs, code, and evaluations. This is what will ultimately accelerate our collective understanding of these models.”
Stella Bidermanan AI researcher involved with EleutherAI, a nonprofit open source AI project, says that Meta’s license for Llama 2 limited the experiments AI researchers can run with it, adding that Llama 3’s license seems even more restrictive. “Meta frees up weights, but is notoriously restrictive about what you can do with them,” Biderman says.
One part of the model license says that companies with “more than 700 million monthly active users” must apply for a special license from Meta, a clause apparently designed to prevent the project from helping the company’s closest rivals.
Still, it seems likely that Llama 3 will spark a new explosion of AI experimentation. Clément Delange, CEO of HugsFace, a repository of open AI models, including Llama 3, says that developers created more than 30,000 variants of Llama 2. “I’m sure we’ll also see a flood of new models based on Llama 3,” it says. “Impressive community movement in Meta.”