Meta’s Pocket-Sized AI

The world of AI is changing fast, and today we're looking at three big stories that could change how we use AI in the future.

First up, imagine your phone becoming super smart. Meta (the company that owns Facebook) has created something called Llama 3.2. It's like a mini-brain for your phone that can do amazing things. It can understand pictures, talk in different languages, and even help manage your daily tasks. The cool part? It does all this without sending your personal info to the cloud.

Next, we've got some drama at OpenAI, the company that made ChatGPT. Some of their top people, including a smart leader named Mira Murati, have decided to leave. It's like when star players leave a sports team – it's a big deal and might change how the company works in the future.

Lastly, Intel (you might know them for making computer chips) has created a new super-fast AI computer called Gaudi 3. It's like a race car for AI, and it's even faster than the previous champion made by another company called NVIDIA. This new computer could help make AI work faster in big data centers.

These changes are exciting because they're not just about new gadgets. They're about making AI smarter, faster, and more personal for all of us.

Meta Unleashes Llama 3.2

Meta has unveiled Llama 3.2, an upgraded suite of models designed specifically for edge AI and mobile devices. This release includes small and medium-sized vision models (11B and 90B) alongside lightweight text-only models (1B and 3B), making advanced AI capabilities accessible to developers with limited computing resources. The models can operate on select mobile devices, enabling the creation of personalized applications that prioritize user privacy by processing data locally. They are equipped to handle complex image reasoning tasks, such as interpreting documents with charts or maps, and generating captions based on visual inputs. In addition to enhancing image processing, Llama 3.2 supports multilingual text generation and tool calling, allowing developers to build efficient applications that can summarize messages or manage calendars directly on devices. Meta has collaborated with over 25 companies, including tech giants like Google Cloud and Microsoft Azure, to facilitate the launch of Llama 3.2, which also introduces the Llama Stack API for easier model customization. With significant advancements in performance, Llama 3.2's vision models are competitive with leading foundation models in various tasks, reflecting Meta's commitment to responsible innovation and community-driven development in edge AI applications.

OpenAI Faces Leadership Exodus: Mira Murati and Top Executives Depar

OpenAI is experiencing a leadership shake-up as Chief Technology Officer Mira Murati, along with Research Chief Bob McGrew and Vice President of Research Barret Zoph, announced their departures. Murati, who played a crucial role in developing technologies like ChatGPT and DALL-E, expressed a desire to explore new opportunities. In response, CEO Sam Altman has appointed Mark Chen as Senior Vice President of Research and Jakub Pachocki as Chief Scientist, among other new roles. These changes come as OpenAI seeks to raise significant funding and transition from a nonprofit to a for-profit benefit corporation, aiming to maintain its competitive edge in the rapidly evolving AI landscape.

Intel Launches Gaudi 3: A Game-Changer for AI Acceleration!

Intel has introduced the Gaudi 3 AI Accelerator, designed to enhance enterprise AI workloads with high performance and efficiency. Built on the Intel Gaudi platform, these accelerators are optimized for demanding training and inference tasks, promising to deliver significant improvements over competitors. Notably, the Gaudi 3 is reported to be 1.5 times faster than NVIDIA's H100 for both training and inference, while also achieving 1.4 times higher power efficiency. This makes it a compelling option for data centers and cloud environments, capable of scaling from single units to large clusters. The Gaudi 3 accelerators support a wide range of AI applications, including generative AI, language translation, and visual question answering. They are designed for easy adoption, whether starting from scratch or transitioning from GPU-based systems. Intel emphasizes that these accelerators will integrate seamlessly with existing Ethernet infrastructures, making them a practical choice for businesses looking to enhance their AI capabilities. With a focus on rapid return on investment and future-proofing AI deployments, Intel aims to position the Gaudi 3 as a key player in the evolving landscape of enterprise AI solutions.

Hand Picked Video

In this video, we'll guide you through the process of installing Meta's powerful language model, Llama 3.1, locally on your computer.

Top AI Products from this week 

• Highlighter - Augie Studio’s Highlighter makes pulling social video highlights fast and easy. Upload your video, set a prompt, and let Augie automatically spotlight the best moments.

• Sensay Replicas - Sensay.io empowers businesses and individuals by preserving knowledge and automating tasks through AI replicas. Enhance productivity, streamline workflows, and ensure continuity effortlessly, with 24/7 support and engagement via text, voice, or video.

• Wispr Flow - Wispr Flow is a Mac dictation app that lets you speak naturally and writes in your style across every application, 3x faster than typing.

• interview.co - Imagine a world where interviews run themselves. interview.co is the all-in-one solution for recruiters who demand efficiency and top talent. interview.co—where interviews are smarter, and, dare we say, enjoyable? If interviews could get a glow-up, this is it!

• Temperstack - Temperstahttps://www.videosdk.live/character-sdk?ck enhances existing monitoring tools with alert health, automation, incident management, AI-runbooks across Infra/Apps.

• Video SDK 3.0 - CharacterSDK allows developers to create multimodal AI characters, capable of real-time interactions and contextual understanding.

This week in AI

• Meta Connect 2024 Highlights - Meta Connect 2024 showcased Quest 3, LLaMA 3, and new AI wearables, emphasizing advancements in mixed reality and AI technology for enhanced user experiences.

• Quick AI Image Analysis - The tutorial teaches how to analyze images using AI in seconds, providing step-by-step guidance and practical applications for efficient image processing.

• AlphaChip Overview - AlphaChip is an open-source framework using distributed deep reinforcement learning for efficient chip floorplan generation, revolutionizing chip design across industries.

• OpenAI's Voice Mode Update - OpenAI launched an advanced voice mode featuring more voice options and a refreshed interface, enhancing user interaction and experience with its AI models.