🍟Meta Releases New AI Chips

Hey TEDchats Fam!

Happy Friday! Better late than never? Meta is living up to this phrase as it releases its first custom AI chip, entering the field much later than its competitors. As Meta positions itself at the starting line, the question arises: will they surge ahead, or just play catch-up?

👀 SNEAK PEEK

🚨FEATURE STORY

The Race for Generative AI Leadership: Meta's Infrastructure Investment

In a world enthralled by the rapid advancements of generative AI, Meta throws its hat into the ring with a groundbreaking reveal that could alter the course of AI development. But are they leading or playing catch-up?

Billions on the Line

The social media giant, Meta, is all about pushing the envelope when it comes to new technologies, owning popular platforms like Instagram, Facebook, Threads, and WhatsApp. The company has ambitious plans to develop state-of-the-art AI infrastructure and is prepared to invest a significant portion of its anticipated $30 to $37 billion budget in 2024 to bring these plans to life. Now, Meta has revealed that a large portion of this spending will go toward proprietary hardware to support its generative AI goals.

Meta is now diving headfirst into this billion-dollar venture, even though it is behind major players like Google, Amazon, Microsoft, and Intel. Intel, by the way, has recently gained significant attention for its impressive AI hardware, Gaudi 3. Despite its position in the market, Meta is pushing forward with its own custom AI hardware.

Introducing Meta's AI Hardware…

The debut of the next-gen Meta Training and Inference Accelerator (MTIA) is a major milestone. MTIA is a long-term project aimed at offering the most efficient architecture for Meta's specific workloads, particularly deep learning recommendation models that enhance a variety of experiences across their products.

Meta's next-gen MTIA chip is its solution to the industry's call for more efficient, powerful AI hardware for advanced AI applications. The chip boasts a 5 nm process, more processing cores, and triple the performance of its predecessor. This custom hardware, designed specifically for Meta's use, allows the company to manage the entire stack, resulting in greater efficiency compared to commercially available graphics processing units (GPUs).

A Leader or A Follower in the Tech Marathon?

The next-gen MTIA is now live in 16 data center regions, less than nine months after the first unit was manufactured. This marks not only a technical milestone but also a strategic step in the ongoing AI arms race. Will Meta's investment in custom AI chips give it an edge over competitors, or is it a case of arriving late to the game? (see Meta’s announcement).

🧑‍💻 DAILY PROMPT

🍕 QUICK BITES

Google AI Photo Editing for Everyone

Starting May 15, Google is transforming photo editing by making its advanced AI tools in Google Photos available to all Android and iOS users. Features like Magic Editor, which were previously exclusive to Pixel devices and Google One subscribers, will now be accessible to everyone. While Pixel owners and paid subscribers will enjoy unlimited edits, other users will be able to enhance their photos with up to 10 saved edits per month (see Google’s announcement).

Texas Will Use AI for Grading School Assessments Alongside Humans

Texas is boldly stepping into the future with its latest move to implement an AI-powered system for grading state-mandated exams, according to The Texas Tribune. Named the “automated scoring engine”, this initiative aims to efficiently tackle the growing number of open-ended questions in exams, potentially saving the Texas Education Agency up to $20 million each year. While the combination of cutting-edge technology and traditional oversight promises a significant change in educational assessments, the plan has raised concerns among educators and parents who remain cautious about the influence of machine learning, automation, and AI (see the article).

Autoparts Manufacturer Trials Humanoid Robots

Sanctuary AI's humanoid work robot, Phoenix™, is stepping into the automotive world through a partnership with Magna, a key player in European car manufacturing. Magna believes that by incorporating general-purpose AI robots into manufacturing facilities for specific tasks, their ability to deliver high-quality products to customers will improve. This collaboration also aims to evaluate the potential of humanoid robots in optimizing production processes, marking a significant step towards the future of automation in the automotive industry (more details here).

🧰 AI TOOLBOX 

🤯 MIND SPARK

ChatGPT’s Newest Version Just Got Released: OpenAI has released their new GPT-4 Turbo for paid ChatGPT users, offering enhanced capabilities in writing, math, logical reasoning, and coding. This update provides responses that are "more direct, less verbose, and use more conversational language" (see OpenAI’s announcement here and here).

Vision AI models fall short of human performance, but there's a new way to test them: Meta has just launched OpenEQA, the Open-Vocabulary Embodied Question Answering Benchmark, to shake up the AI scene and motivate research. This tool challenges AI agents with open-ended questions like 'Where did I leave my badge?' to measure how well they understand and navigate physical environments — in other words, to test how well they 'see' (see Meta’s announcement).

Get Local Weather Updates with AI! The Weather Company has introduced ReelSphere, a new AI tool designed to automate the creation of hyperlocal weather videos, complete with custom graphics, captions, and AI-generated voiceovers akin to local meteorologists (see more).

🎨 AI ART