Meta Platforms Launches Llama 3 and Enhances Meta AI Assistant

In a bid to compete with leading generative AI technologies, Meta Platforms has unveiled its latest large language model, Llama 3, along with a real-time image generator. These releases aim to bolster Meta’s position in the AI market against competitors like OpenAI.
Llama 3 model performanceThe Llama 3 model, integrated into Meta’s virtual assistant, Meta AI, is one of the most advanced among free-to-use AI assistants. Meta showcased its performance in various domains, including reasoning, coding, and creative writing, comparing it favorably against offerings from rivals such as Google and Mistral AI.

Meta plans to elevate the Meta AI assistant’s visibility within its ecosystem, including Facebook, Instagram, WhatsApp, and Messenger apps, as well as through a standalone website. This move positions Meta to directly challenge Microsoft-backed OpenAI’s popular ChatGPT.

The updated Meta AI assistant promises to assist users in tasks ranging from creating vacation packing lists to engaging in trivia games or receiving homework help. Meta’s efforts to push generative AI products to its billions of users have entailed substantial investments in computing infrastructure and the consolidation of research and product teams.

Llama 3 boasts new coding capabilities and has been trained on both images and text data. Although it currently outputs only text, future versions are expected to generate both text and images, enhancing its utility in various applications.

Chris Cox, Meta’s Chief Product Officer, emphasized the goal of simplifying users’ lives by assisting them in various tasks, from interacting with businesses to planning trips. Moreover, the inclusion of images in Llama 3’s training is expected to enhance its performance in identifying objects, a feature that will integrate with Meta’s smart glasses product.

Meta also announced collaborations with Google and Intel to enhance its AI assistant’s capabilities and performance. Additionally, Meta’s expansion of its AI assistant to markets outside the U.S. reflects its ambition to reach a global audience.

Despite the impressive performance of Llama 3, challenges remain, as developers have raised concerns about previous models’ understanding of context. Meta aims to address these issues by leveraging high-quality data to train Llama 3.

Intel, Qualcomm, and IBM have announced collaborations with Meta to optimize and deploy Llama 3 models across various devices and platforms, signaling broader industry support for Meta’s AI initiatives.

INTEL

Intel has validated its AI product portfolio for the first Llama 3 8B and 70B models across Intel Gaudi accelerators, Intel Xeon processors, Intel Core Ultra processors and Intel Arc graphics.

“Meta Llama 3 represents the next big iteration in large language models for AI. As a major supplier of AI hardware and software, Intel is proud to work with Meta to take advantage of models such as Llama 3 that will enable the ecosystem to develop products for cutting-edge AI applications,” Wei Li, Intel vice president and general manager of AI Software Engineering, said.

QUALCOMM

Qualcomm Technologies and Meta announced a collaboration to optimize the execution of Meta Llama 3 large language models (LLMs) directly on smartphones, PCs, VR/AR headsets, vehicles, and more. The ability to run Llama 3 on devices offers significant advantages, including superior responsiveness, enhanced privacy, improved reliability, and more personalized experiences for users.

IBM

IBM announced the availability of Meta Llama 3 on its watsonx AI and data platform. This expands IBM’s watsonx.ai model library to help enterprises innovate with its in-house Granite series of models, as well as those from leading model providers like Meta.

Baburajan Kizhakedath

Related News

Latest News

Latest News