Meta is acquiring AI training chips and expanding its data centres to develop a more powerful chatbot that it believes will be as advanced as OpenAI’s GPT-4. With CEO Mark Zuckerberg apparently pushing for it to once again be free for businesses to construct AI products with, the company reportedly aims to start training the new huge language model early in 2024.
According to The Journal, Meta has been increasing its infrastructure and purchasing additional Nvidia H100 AI-training processors so that it won’t have to use Microsoft’s Azure cloud platform to train the new chatbot this time. According to reports, the corporation put together a team to create the model earlier this year in an effort to hasten the development of AI capabilities that can mimic facial expressions.
That objective seems to be a logical expansion of the generative AI characteristics that Meta has allegedly already been developing. The unknown AI “personas” the business is rumoured to be unveiling this month sound a lot like the Instagram chatbot with 30 personalities that was allegedly under testing in June.
According to reports, Meta has had to cope with a lot of AI researcher turnover because computational resources have been divided among several LLM projects this year. OpenAI declared in April that it would not train a GPT-5 and “won’t for some time,” but Apple is rumoured to have been pouring millions of dollars every day into its own “Ajax” AI model, which it reportedly believes is superior to even a GPT-4. Google aims to employ generative AI in Google Assistant, and Microsoft and Google have also been increasing the usage of AI in their productivity tools. Additionally, Amazon is working on generative AI projects that could lead to an Alexa chatbot.