Meta Platforms is reportedly working on a new and more powerful large language model (LLM) that will exert competition in the generative AI industry.
According to reports, the new model is expected to be more powerful than ChatGPT-4 from OpenAI.
All mighty and powerful
The Mark Zuckerberg-owned Meta anticipates the new model to be ready next year after it starts training it early 2024, the Wall Street Journal reported citing source close to the developments.
It follows therefore that the anticipated new model will be “several times more powerful” than its current commercial version known as Llama 2. This model was launched in July this year and is distributed by Azure services to rival ChatGPT and Google’s Bard.
According to the report, although its details could change, the planned new model will help other firms develop services that produce “sophisticated text, analysis and other output.”
Meta is reportedly planning to train the new model using its own infrastructure, although its previous model Llama 2 was made available using Microsoft’s Azure cloud computing platform.
The company is also reportedly in the process of getting H100 AI training chips, the most advanced chips produced by Nvidia. Meta is also building data centers to train the model, according to the reports.
Also read: Meta Touts Metaverse Reality as Apple Prepares Vision Pro VR Headset
Staying in the race
OpenAI spurred the race of generative AI worldwide with the launch of its ChatGPT in November last year. Even in China, where ChatGPT is restricted, the chatbot went viral with the market accessing it via VPNs.
To meet the demand for the model, tech firms in that country developed their own ChatGPT-alternative, for example Baidu’s chatbot known as Ernie Bot.
Businesses are not stopping there, but flocking the cut-throat generative AI market for “newer capabilities and refining business processes.”
Recently, Apple was reportedly working on its own AI model, which engineers dubbed Apple GPT to rival ChatGPT and Google’s Bard.
Now, for the Zuckerberg-owned Meta, this is an attempt to reassert its market position as a force to reckon with in the AI industry, despite falling behind peers in the sector like OpenAI, Microsoft and Google.
Earlier, it was also reported the company has built teams of AI researchers to work on AI offerings including a chatbot with 30 different personalities on its other platform – Instagram. The Chatbot could be released to the public this year according to HT.
Risky approach
Reports also suggest that Zuckerberg is pushing for the new model to be open-sourced and available to other companies for free, enabling them to build their own AI powered products. This will be similar to its other AI products and services.
However, the Wall Street Journal suggests this could be a risky approach with potential downsides, such as spread of misinformation coupled with copyright infringements.
Meanwhile, Zuckerberg is expected to join other tech gurus like OpenAI CEO Sam Altman, Google CEO Sundar Pichai and Tesla CEO Elon Musk at the newlyn formed AI forum hosted by US Senate Majority Leader Chuck Schumer.
and then