Zuck doesn't want to get left behind.

Second Wind

As OpenAI and Google sprint ahead in the ongoing Silicon Valley AI race, Meta — yes, the company that renamed itself just three years ago to go all in on the prior tech craze of the metaverse — appears more determined than ever to catch up.

According to new reporting from The Wall Street Journal, Meta is secretly developing a powerful new AI model designed to compete with OpenAI's GPT-4, the most advanced iteration of the rival AI firm's large language model (LLM) to date.

Very little is known about the model, and as the WSJ reports, the details of the AI — which, as of now, is reportedly intended to "help other companies to build services that produce sophisticated text, analysis and other output" — may still be subject to change. It's also worth noting that this isn't Meta's first foray into LLM territory; LLAMA-2, the company's much-hyped language model, was released just a few months ago.

But according to WSJ's sources, Meta is hoping that its new AI will be "several times more powerful" than LLAMA-2 — a sign, perhaps, that LLAMA-2 isn't entirely cutting it as a GPT-4 alternative, and that Meta knows it has to seriously up its ante if it's going to keep up with its competitors.

Dust in the Wind

Per the report, the model is being built by a Meta team — formed earlier this year by the company's founder and longtime overlord Mark Zuckerberg — that was initially tasked with expediting the company's effort to build a generative AI system able to "produce human-like expressions." Don't hold your breath; the team reportedly plans to start training the new AI by early 2024, and in the meantime, in an effort to expand its AI-supporting infrastructure, Meta has apparently been busy acquiring those coveted Nvidia H100 computing chips.

The report also noted that Zuckerberg has been pushing to make the firm's new AI open-source — a decision that's reportedly troubled Meta's lawyers, who are said to have raised some valid concerns about threats such as potential AI-abetted copyright infringement and the use of AI to generate disinformation.

Importantly, OpenAI isn't Meta's only major competitor. Google is also a leader in the gold rush-hungry pack, while smaller firms like the heavily Google-funded Anthropic have also made a big splash. Apple, meanwhile, is reportedly spending "millions of dollars a day" to train its AI.

And it's worth mentioning that Meta isn't the only company with a new model on the horizon. Google's forthcoming, Gemini AI has drawn wide anticipation, and may well wind up kicking GPT-4's ass. And if Meta doesn't get anything new and impressive into public hands before then? It all remains to be seen, of course, but Zuck could find himself left even deeper in the dust than he anticipated.

More on Meta's AI efforts: Meta Launching AI-Powered Friends to Talk to on Facebook


Share This Article