Meta Platforms, Inc., led by CEO Mark Zuckerberg, has recently announced the launch of its latest AI model, Llama 3.3 70B, which reportedly surpasses competitors such as Alphabet Inc.'s Google, OpenAI, and Amazon.com, Inc. The announcement was made on Friday and signifies a strategic move by Meta to solidify its status within the AI sector.
The new model delivers similar performance to Meta's larger Llama model, Llama 3.1 405B, but at a significantly reduced cost. Ahmad Al-Dahle, Meta's Vice President of Generative AI, indicated that Llama 3.3 utilises advanced post-training techniques to enhance its core performance efficiently. In a post on X (formerly Twitter), Al-Dahle shared a comparative chart demonstrating the model's superiority over Google’s Gemini 1.5 Pro, OpenAI’s GPT-4o, and Amazon’s Nova Pro across various benchmarks, including the Multiple-Choice Questions in the Massive Multitask Language Understanding (MMLU) test. The model is available for download on platforms such as Hugging Face and the official Llama website.
In addition to the new model, Zuckerberg made mention of Meta's AI progress on Threads, highlighting that the company’s AI products now boast nearly 600 million monthly active users. The company’s ambitions in the AI space are further underscored by its plans to construct a $10 billion AI data centre in Louisiana to support future developments of Llama models.
The announcement aligns with earlier reports suggesting Meta's intention to create an AI-powered search engine to reduce its reliance on dominant tech players like Alphabet and Microsoft. Additionally, Meta is exploring the integration of nuclear energy to support its AI projects, seeking proposals for nuclear power ventures to generate 1-4 gigawatts of capacity in the United States by the early 2030s. This strategic energy initiative reflects broader trends in sustainable technology adoption within the tech industry.
Meta's third-quarter earnings report has revealed a revenue of $40.59 billion, which comfortably surpassed analyst expectations. Following the model launch, Meta's stock experienced a notable increase of 2.44%, closing at $623.77, although it saw a slight dip of 0.08% in after-hours trading. Over the course of 2023, Meta shares have surged by an impressive 80.13%, significantly outperforming the Nasdaq 100 index, which has risen by 30.7%.
Alongside these developments, Meta has rolled out an internal AI tool known as Metamate, aimed at enhancing productivity within the company. Built upon the Llama model, Metamate assists with coding tasks, research, and drafting communications, though it is exclusively an internal tool and isn't intended for external release. Prashant Ratanchandani, Meta's VP of Engineering overseeing Metamate's development, stated that their goal is to create “the world’s best enterprise assistant.” This aligns with Meta's broader efforts to monetise its AI capabilities and establish a stronger foothold in the enterprise market.
The market for AI agents, which are designed to handle complex tasks and reasoning, is projected by MarketsandMarkets to expand from $5.1 billion to $47 billion by 2030. Meta's recruitment of Clara Shih, former CEO of AI at Salesforce, to lead a new "Business AI" group exemplifies its commitment to making cutting-edge AI accessible to various businesses.
Despite these initiatives, analysts have remarked that Meta's Metamate is still in the nascent stages when compared to similar products offered by tech giants Microsoft and Google. Unlike the more consumer-driven AI chatbots, Metamate currently functions primarily through text and lacks the autonomous features that allow it to manage scheduling or summarise meetings. Nevertheless, the tool is being actively used by Meta employees for various internal tasks, although Ratanchandani noted that improvements to enhance its functionality are under consideration.
The ongoing developments in Meta's AI landscape underscore a significant trend within the technology sector, as firms race to harness the capabilities of AI to drive future growth and innovation. However, widespread adoption of such technologies is projected by analysts to remain limited until at least 2027, suggesting that while the potential for AI in business is significant, the path to widespread implementation remains complex.
Source: Noah Wire Services