We think that we all knew that Meta will not stay away from the AI trend. Mark Zuckerberg's company Meta announced that it has trained and is now making available to researchers a new large language model that is a key component of artificial intelligence systems that have become hugely popular in recent months. However, this model will be smaller than its competitors, and the company sees an advantage in this.
Large Language Model Meta AI, or LLaMA, is a competitor to ChatGPT AI
Meta is joining the battle for dominance between AI companies in the same way that Google did some time ago. LLaMA will soon be available under a non-commercial license to researchers and those associated with government, civic affairs, or academia. Like other similar technologies, it will be able to process huge amounts of text to summarize data and create content.
Meta announced that its model requires "significantly less" computing power than other offerings and trains simultaneously in 20 languages using the Latin and Cyrillic alphabets. Making such a tool available to researchers is a step toward broader testing of a model that Meta will likely want to use in future products based on content-generating artificial intelligence.
Another chatbot, this time from Zuckerberg?
Meta has little experience when it comes to generative AI like ChatGPT or Barda, but artificial intelligence is a good reason for the company to invest in the tech industry. The recent economic downturn in the industry has led to massive layoffs at many companies, including Meta. For Zuckerberg's company, this is also partly the consequence of focusing on Metaverse, which so far brings more losses than profits.
The company claims that LLaMA can outperform competitors in terms of the number of parameters and variables taken into account by the algorithm. OpenAI's ChatGPT has a whopping 175 billion parameters, but Meta says the same can be achieved with fewer. LLaMA will be available in several options, from just 7 billion parameters to 65 billion.
Recommended:
