Artificial intelligence
Facebook Group publishes new AI model
Meta’s Llama technology (Large Language Model Meta AI) is available as open source software. photo
© Tony Avelar/AP/dpa
Llama-3 is intended to take apps and services to a new level. Faster image generation, animations, clever assistance.
Llama-3 will initially only be available in English, Meta manager Nick Clegg told the German Press Agency when the program was announced. However, more than five percent of the data used to train Llama-3 was in other languages.
Among other things, Meta brings his AI assistant into the networked glasses developed together with Ray Ban, which have a camera, microphone and loudspeaker. For example, while skiing, you can ask the assistant when and how Cleopatra died or what the weather is like in Berlin, said Clegg. The software will also be able to generate images from text specifications more quickly and also display them as animations.
Meta is currently not thinking about business models for AI software, emphasized Clegg, who is the group’s head of policy. The Facebook Group wants to first develop technology that people find useful or interesting – “and then we’ll figure out later how you can make money with it.”
Unlike, for example, the ChatGPT developer OpenAI, Meta makes its Llama technology (Large Language Model Meta AI) available as open source software where the source code is publicly viewable. There was a growing view that open source models were more secure because many people could test them and “you don’t have to rely on a company to iron out the vulnerabilities in your software.” At the same time, there is currently a lack of a uniform basis for assessing the risks of artificial intelligence, criticized Clegg.