Slashdot Mirror

← Back to Stories (view on slashdot.org)

Facebook To Design Its Own Processors For Hardware Devices, AI Software, and Servers (bloomberg.com)

Posted by BeauHD on from the do-it-yourself dept.

Facebook is the latest technology company to design its own semiconductors, reports Bloomberg. "The social media company is seeking to hire a manager to build an 'end-to-end SoC/ASIC, firmware and driver development organization,' according to a job listing on its corporate website, indicating the effort is still in its early stages." From the report: Facebook could use such chips to power hardware devices, artificial intelligence software and servers in its data centers. Next month, the company will launch the Oculus Go, a $200 standalone virtual-reality headset that runs on a Qualcomm processor. Facebook is also working on a slew of smart speakers. Future generations of those devices could be improved by custom chipsets. By using its own processors, the company would have finer control over product development and would be able to better tune its software and hardware together. The postings didn't make it clear what kind of use Facebook wants to put the chips to other than the broad umbrella of artificial intelligence. A job listing references "expertise to build custom solutions targeted at multiple verticals including AI/ML," indicating that the chip work could focus on a processor for artificial intelligence tasks. Facebook AI researcher Yann LeCun tweeted about some of the job postings on Wednesday, asking for candidates interested in designing chips for AI.

2 of 56 comments (clear)

  1. Baked-in spyware by Anonymous Coward · · Score: 5, Insightful

    What Zuckerberg really wants to do is bake spyware right into the CPU so it can collect data on you in a totally unimpeachable, nigh-unto undetectable way.
    DO NOT WANT.

  2. Not for the public ; a response to TPU by Guybrush_T · · Score: 3, Insightful

    This seems like an answer to Google's TPU. Nothing like a general purpose CPU they'd want to sell to anyone, more like a dedicated piece of hardware to accelerate ultra common deep learning workloads (like, image recognition).

    Just like Google, Facebook has to process immense volumes of images. GPUs are much more efficient at doing that than CPUs, but so there is still a bit of room for improvement when doing very specific tasks.