Meta, formerly known as Facebook, is leaving no stone unturned in its quest to develop a chatbot that can rival OpenAI's GPT-4. The company's ambitious plan is to kickstart the training of this new large language model early in 2024.
CEO Mark Zuckerberg is leading the charge, advocating for this AI marvel to once again be freely available for companies to create their own AI tools.
Meta has been on a shopping spree to power its AI aspirations, acquiring Nvidia H100 AI-training chips. The Verge reports the company is bolstering its infrastructure to reduce its reliance on Microsoft's Azure cloud platform for training the chatbot.
Rumors have been swirling about generative AI features that Meta has been quietly working on. In June, leaks hinted at the existence of an Instagram chatbot with a whopping 30 personalities under testing. This development closely aligns with the unannounced AI "personas" that Meta is rumored to unveil this month.
The company has grappled with a high turnover of AI researchers due to resource allocation among multiple large language model (LLM) projects. This internal challenge could hinder the speedy development of its chatbot. Moreover, Meta isn't alone in the race for generative AI dominance.
OpenAI, the creator of GPT-4, announced in April that it was planning to wait to develop a GPT-5. However, other tech giants are throwing their hats into the generative AI ring.
Apple, for instance, is reportedly investing millions of dollars daily into its "Ajax" AI model, which it believes could outperform even GPT-4. Google and Microsoft are integrating AI into their productivity tools. Google plans to use generative AI for Google Assistant. Amazon is also working on generative AI for Alexa. So, its going to take a lot of work for Meta to be the best at AI games.