Social media giant Meta is a pioneer in artificial intelligence (AI) and uses it to power its products and services, such as News Feed, Facebook Ads, Messenger, and VR. The company announced new hardware and software projects to support the next generation of AI applications. The event featured speakers from Meta who shared their insights and experiences on building and deploying AI systems at a large scale.

Meta announced that it had built its own AI data center optimized for both AI training and inference, using Meta’s own silicon, the Meta training and inference accelerator, and the Research Supercluster (RSC) to help accelerate AI workloads across various domains.

AI infrastructure is a priority for companies like Meta and Baidu. The goal is to enable even more advances in AI.

IBM’s new AI supercomputer, Vela, uses x86 silicon alongside Nvidia GPUs and ethernet-based networking. IBM’s goal is to build out new foundation models that can help serve enterprise AI needs. Google announced its upcoming AI supercomputer, Google System, on May 10, using Nvidia GPUs and custom-designed infrastructure processing units.

Meta is jumping into the custom silicon space with its MTIA chip. Custom-built AI inference chips are becoming more common. Meta Research scientist Amin Firoozshahian commented that traditional CPUs are not suited to handle inference demands from the applications that Meta runs. “MTIA is a chip that is optimized for the workloads we care about and tailored specifically for those needs.”

Meta is a large-scale machine-learning company using PyTorch. They want to have highly optimized silicon for running PyTorch workloads. The MTIA silicon is a 7nm (nanometer) process design and can provide up to 102.4 TOPS of throughput. They are planning to optimize AI operations, including networking, data center optimization, and power utilization. The Data Center of the Future is going to be built for AI.

Meta has built its data center for over a decade to meet the needs of its billions of users. As AI use grows, the company needs more computing capacity. To address this, it is rethinking cooling and providing liquid cooling for chips.

  • Meta has launched the AI Sandbox, a testing playground for advertisers to experiment with generative AIpowered ad tools. The sandbox could revolutionize building ads and improve campaign results.

  • IBM announced WatsonX, a new AI development platform for enterprises. With watsonx, IBM is offering an AI development studio with a suite of foundation models that enterprises can use to build, train, tune, and deploy traditional machine learning and new generative AI capabilities. The IBM watsonx.ai studio is expected to launch in July 2023.

  • Canva, the popular visual communication platform, has unveiled a new suite of AI-powered design tools at the Canva Create event. The suite includes eight new features, one of which is Magic Design, which allows users to upload an image and select a style, after which the tool will curate a personalized selection of templates. Another new addition is Magic Edit, which allows users to add or replace anything in an image simply by describing it to the tool.

Read more related articles:

  • Adobe Unveils New Generative AI-powered Services in Adobe Experience Manager

  • Meta creates new AI-powered platform that could aid medical research, upgrade avatars

  • Google AI Search Update: New Chatbot ‘Bard’ Unveiled to Challenge ChatGPT-powered Bing

The post Metas AI-first Vision Takes Shape with Unveiling of New Data Centers and Supercomputer appeared first on Metaverse Post.