There is still a thick cloud of mystery aroundMeta’s artificial intelligencewhich is secretly working on its own alternative to ChatGPT on OpenAI e Google Bard.
According to the latest rumors, Mark Zuckerberg’s company would have acquired new chips to work on a more powerful, precise and accurate AI model even than GPT4the latest AI engine that powers ChatGPT (and Bing Chat).
Will Meta’s AI outperform GPT-4?
The news comes from the Wall Street Journal (via The Verge), according to which Meta purchased a large quantity of chips NVIDIA H100. These will make up the hardware part for AI training, which should begin by the beginning of 2024.
The chip game, again according to the Journal, will go to enhance Meta’s internal AI infrastructurewhich will thus not have to resort to the cloud platform Azure (owned by Microsoft, financier of OpenAi).
The result will be (or at least it is planned) a more realistic, precise and humanoid language model than ChatGPT with GPT4. A model that therefore aims to overcome the limits expressed by the OpenAi chatbot, first of all the lack of accuracy of certain information.
But Meta isn’t the only big tech vying to overtake OpenAi. There is also an appeal Applewhich is investing significant resources (in the order of millions of dollars per day) for what is called Project Ajax. More muted instead Amazonwhich may soon roll out its own language model in the Alexa voice assistant.
Are we expecting a technological revolution in 2024? Probably yes.