ChatGPT: Meta in turn draws a generative AI to counter Microsoft and Google

Before the metaverse, artificial intelligence was already at the heart of the development of Meta. And after having squandered more than 20 billion dollars in two years to build a digital understudy of the physical world, the American group is putting AI back at the center of the game. Not only does this area attract more investors than the metaverse, but above all, the Menlo Park firm must react as Microsoft invests heavily in OpenAI, the start-up behind ChatGPT, and Google unveiled its conversational robot Bard in early February.

To show that it is as invested as its two American rivals in this booming market, Meta has announced the launch of its own generative AI. Called LLaMA (Large Language Model Meta AI), it is based on 45 billion parameters, according to Mark Zuckerberg’s group. Meta has also launched versions of this AI based on a reduced number of parameters to make it less resource-intensive. The different versions of LLaMA “showed great promise for generating text, having conversations, summarizing written documents, and more complicated tasks like solving mathematical theorems or predicting protein structures”assured the boss of Meta on Facebook.

For now, this generative AI is only available to researchers. The Californian firm prefers to play the card of caution, while ChatGPT was put in the hands of the general public as soon as it was launched on November 30, 2022. Recently integrated into the Bing search engine, the OpenAI tool has since been restricted by Microsoft to limit its slippages, insofar as it sometimes adopts strange behavior, which makes users uncomfortable. ChatGPT is “a fascinating tool [qui n’est] at this stage than an approximate parrot who sometimes somewhat awkwardly restores the astronomical sums of information he has compiled on the Internet”thus estimated a few days ago Jean-Noël Barrot, Minister Delegate for Digital.

If it was essential for Meta to show its muscles on the AI ​​market, the American giant prefers to wait before demonstrating the potential of LLaMA to the general public. Meta probably wants to avoid the mishap of Google, which had paid dearly for a blunder of its conversational robot Bard in an advertisement presenting the tool. Once the error was discovered, Google had lost $100 billion on Wall Street in one day. However, after having lost 600 billion dollars on the stock market in 2022, Meta cannot afford such a poor performance, especially since 2023 has been decreed by the firm “year of efficiency”.