Updates: Meta - Underappreciated GenAI Potential (Pt.1)

Summary

• In Part 1 we review the battle for GenAI supremacy and how Meta's Llama has emerged as a foundation model leader, despite the odds.
• We discuss the challenges Mistral, the previous open-source LLM leader, is facing when competing against Meta and closed source LLM leaders.
• In Part 2 we delve into an extensive discussion regarding the possibilities of Meta monetizing its open-source Llama models.
• We shall also be publishing a Notes report on Meta, discussing how they might control the GenAI value chain.

Since the beginning of the GenAI hype in late 2022, we've seen multiple rounds of GenAI battles and changes in leadership (less so for GPT-4 until recently) including:

• OpenAI & MSFT leading the field first, with GPT and Phi, respectively.
• GOOGL's DeepMind trying very hard to fight back with PaLM and then Gemini & Gemma (Open Source variant).
• META initially entered the GenAI space as an underdog with its open-source LLM, Llama, along with several new models for voice, image, and video. However, they quickly emerged as a leading contender in the field.
• Anthropic quickly reached parity with OpenAI and now they are partially ahead with the latest Claude model.
• Mistral gained some initial momentum as a startup with their open-source models, but then they chose to close source their larger models due to investor pressure. However, they have since partially reverted to being open again, following Llama 3's performance.
• MosaicML initially found success with its open-source MPT model, but after being acquired by Databricks, its open-source models became outdated as updates were paused, while META and Mistral continued advancing with next-gen releases.
• From late 2023 through to 2024, X.ai rapidly emerged as a tier 1 player with super ambitious talent, a DC buildout, and the release of Grok.
• SNOW entered the GenAI market later but demonstrated strong leadership in RAG and enterprise applications with its open-source model, Arctic
• Other smaller players like AI21 Labs, Cohere, Reka AI, DeepSeek (subsidiary of a Chinese quant trading firm), Qwen (BABA), and others.

The competition in GenAI has intensified to the point where OpenAI strategically brought forward the launch of its GPT-4o to the 13th May 2024, just a day before Google I/O 2024, to dominate headlines. Google I/O 2024, an annual developer conference, revealed advancements in their Gemini AI models and introduced the Search Generative Experience (SGE), which enhances search with AI capabilities. Meanwhile, META keeps pace with continuous updates to Llama, making it increasingly central to its strategic and financial objectives. This series will delve into Llama's development, competition, and future trajectory.