Meta Platforms on Thursday shared new particulars on initiatives it was pursuing to make its knowledge facilities higher suited to supporting synthetic intelligence work, together with a customized chip “household” that it stated it was creating in-house.
The Fb and Instagram proprietor stated in a collection of weblog posts that it designed a first-generation chip in 2020 as a part of the Meta Coaching and Inference Accelerator (MTIA) program, which was geared toward enhancing effectivity for the suggestions fashions it makes use of to serve advertisements and different content material in information feeds.
Reuters beforehand reported that the corporate was not planning to deploy its first in-house AI chip extensively and was already engaged on a successor. The weblog posts portrayed the primary MTIA chip as a studying alternative.
“From this preliminary program, we’ve got discovered invaluable classes that we’re incorporating into our roadmap,” it wrote.
The primary MTIA chip was targeted completely on an AI course of referred to as inference, wherein algorithms educated on enormous quantities of information make judgments about whether or not to point out, say, a dance video or a cat meme as the subsequent put up in a person’s feed, the posts stated.
A Meta spokesperson declined to touch upon deployment timelines or elaborate on the corporate’s plans to develop chips that would prepare the fashions as properly.
Meta has been engaged in a large undertaking to improve its AI infrastructure this previous 12 months, after executives realized it lacked the {hardware} and software program wanted to help demand from product groups constructing AI-powered options.
As a part of that, the corporate scrapped plans for a large-scale rollout of an in-house inference chip and began work on a extra bold chip able to performing each coaching and inference, in line with the Reuters reporting.
Meta acknowledged in its weblog posts that its first MTIA chip stumbled with high-complexity AI fashions, though it stated the chip dealt with low- and medium-complexity fashions extra effectively than competitor chips.
The MTIA chip additionally used solely 25 watts of energy — a fraction of what market-leading chips from suppliers akin to Nvidia eat — and used an open-source chip structure referred to as RISC-V, Meta stated.
Along with detailing its chip work, Meta supplied an replace on plans to revamp its knowledge facilities round extra fashionable AI-oriented networking and cooling programs, saying it could break floor on its first such facility this 12 months.
The brand new design can be 31 p.c cheaper and could possibly be constructed twice as rapidly as the corporate’s present knowledge facilities, an worker stated in a video explaining the adjustments.
© Thomson Reuters 2023
