Meta Investing Heavily in AI to Catch Up with Rivals: Report

Summary:

  • Meta has been investing heavily in AI after acknowledging a “significant gap” in the area.
  • The company has been spending an additional $4 billion a quarter on AI hardware and development capabilities.
  • Efforts ramped up significantly after OpenAI launched ChatGPT in November 2022.

Meta has spent much of the past year making major investments in artificial intelligence (AI) hardware and development, according to a report from Reuters.

Company CEO Mark Zuckerberg evidently held a five-hour-long meeting with top Meta executives last year to discuss the company’s AI capabilities. In an internal memo dated September 20, Meta acknowledged that it had a “significant gap” in its AI development capacity and was falling behind rivals.

“We have a significant gap in our tooling, workflows and processes when it comes to developing for AI. We need to invest heavily here,” the company’s new head of infrastructure, Santosh Janardhan, wrote in the memo.

Janardhan added that Meta would have to “Supporting AI work would require Meta to “fundamentally shift our physical infrastructure design, our software systems, and our approach to providing a stable platform” in order to catch up.

In the months since, Meta has been working to overhaul its AI infrastructure and development capabilities.

The endeavour has increased Meta’s capital expenditures by about $4 billion per quarter at a time when the company has been feeling some financial heat, laying off a not insignificant portion of its workforce while also cutting projects and expansion plans. Last week, a report indicated that Meta once again plans to cut thousands of jobs.

As part of its ambitions to upgrade and fortify its AI capabilities, Meta has ordered billions of dollars worth of GPUs from Nvidia, started retooling its data centers for the incoming GPUs, and reorganized its AI units, naming two new heads of engineering, including Janardhan.

Meta was reportedly also developing its own AI chip for inference, an AI process where trained models make judgments and generate responses to prompts. However, plans to roll out the in-house chip at scale were axed last year.

Now, Meta is developing a new and more ambitious chip for AI workloads that could handle both training AI models and performing inference. The company hopes to finish work on this new in-house chip around 2025, according to two sources.

Meta’s AI capacity, even after the company started expanding it, was originally focused on ads, feeds, and Reels. The social media giant wasn’t particularly focused on generative AI that creates human-like written and visual content in response to prompts.

However, that changed following the November 30 launch of OpenAI’s wildly popular AI chatbot, ChatGPT, which sparked an AI arms race amongst tech companies. Meta ramped up generative AI development in response.

That said, Meta currently lags behind rivals like Google and Microsoft, which have already put generative AI products on the market. What’s more, Meta’s struggle to catch up is taxing its already lacking resources, which are also being used to fuel Zuckerberg’s ambitions for the metaverse.

Meta hopes to release its first generative AI product sometime this year, according to recent comments from Chief Technology Officer Andrew Bosworth.

P.S. - Like our news? Support the site with a coffee/beer. Or shop with our Amazon link. We use affiliate links when possible--thank you for supporting independent media.