Meta Platforms released Llama 3 on Thursday, the latest update to its publicly available language model, which further intensifies AI competition in Silicon Valley. The new AI functionality is being used to improve the Meta AI tool, which is now more prominently integrated into its apps and is also being incorporated into the search functions of its main platforms like Facebook, Instagram, WhatsApp, and Messenger. The updated tool can also summarize links and will be accessible through the website meta.ai.
Meta Launches with Two Versions of the Model, Including One with 8 Billion Parameters and Another with 70 Billion Parameters. Parameters are the variables in an AI system used to measure the size and capabilities of a model. According to the company, the smallest version of Llama 3 is as powerful as the largest version of its predecessor. Meta will also use its Llama 3 model to enhance its Meta AI Assistant Chatbot.
"Based on the benchmarks we see today, this will be the smartest and most powerful model of its kind available for free," said Chief Product Officer Chris Cox in an interview.
Meta Released Llama 2 Publicly in July and Introduced the First Version of Its Meta AI Chatbot within Instagram, WhatsApp, and Messenger in September. Unlike Other Companies in This Field, Meta Has Released Its Model as "Open Source" Software, Meaning That Others Can Use and Modify It for Free.
Meta Announces Plans to Release More Llama 3 Models with Additional Capabilities in the Coming Months, Including the Ability to Communicate in Multiple Languages. The Company Is Working on an Even Larger Model with 400 Billion Parameters, Which Will Be Released After the Company Completes Its Training.
With the launch of Llama 3, Meta AI will be more capable and will roll out to 13 additional English-speaking markets, including Australia and Canada. It will not be available in the European Union at the start.
The image generation feature of Meta AI, Emu, is also being improved to deliver faster results that are of higher quality and show users preview results as they type their request.
Meta focuses on democratizing these AI advancements for free for its users, said Cox. "What we as a company have tried to do, is to make the tools usable for everyone everywhere," added Cox. He mentioned that the company would consider a monthly subscription model, similar to OpenAI's ChatGPT Plus and Google One AI Premium, when larger versions of Llama become available.
It is noteworthy that Meta has used synthetic data, i.e., texts generated by other AI models, to train Llama 3. As technology companies reach their limits regarding the available data for training their AI models, they must find solutions for how they can obtain or create more data.