Meta plans AI strategy around three key pillars

Mark Zuckerberg has unveiled Meta’s core AI strategy during the Q4 2023 earnings call, shedding light on three key pillars: formidable compute infrastructure, a steadfast commitment to open source, and access to an extensive pool of training data from Facebook and Instagram.

Zuckerberg highlighted the pivotal role of compute power, stating, “Our long-term playbook is to become leaders in building the most advanced AI products and services, as well as building the full general intelligence.” The company is set to possess around 600k H100 equivalents of compute by yearend, a result of lessons learned from Instagram Reels development.

Addressing Meta’s open source approach, despite recent criticism, Zuckerberg emphasised its strategic benefits. “Open source software is typically safer and more secure, as well as more compute efficient to operate,” he remarked, underlining its role in fostering industry standards and attracting top talent.

The vast training data at Meta’s disposal was a focal point, with Zuckerberg revealing “hundreds of billions of publicly shared images and tens of billions of public videos” on Facebook and Instagram. This data, he asserted, surpasses the Common Crawl dataset, complemented by “large numbers of public text posts in comments across our services.”

Zuckerberg concluded with Meta’s proactive stance in the AI race, stating, “We are playing to win.” The company’s commitment is evident through aggressive investments in AI infrastructure, hinting at significant capital expenditure in 2024.