Meta's Behemoth AI Delay: Impact and Opportunity with Gene Munster

Behemoth AI Model Delay

• 📉 Reports indicate Meta is delaying the rollout of its flagship AI model, Behemoth, from June to the fall or later.
• 🧠 Behemoth is described as a 400 billion parameter model, roughly half the size of the latest rumored GPT models from OpenAI.
• ⚠️ This delay is considered a "black eye" for Meta, highlighting the difficulty in advancing these large AI models.

Meta's AI Progress and Opportunity

• 📈 Despite the delay, Meta is seeing benefits of AI at scale, evidenced by a 6% growth in daily active users in the March quarter.
• 🎯 AI improvements contribute to better content and targeting, enhancing user engagement.
• 🚀 While the delay is a headwind, it doesn't change Meta's overall AI opportunity.

Investor Expectations and AI Development

• 📊 Investors have expanded multiples based on the idea of AI as a massive growth engine, but patience is needed for AI technologies to mature.
• 💡 The advancement of AI models is proving harder than anticipated, with less of an exponential step-up in model improvements than some expected.
• 💰 Companies are likely to increase spending on hardware to compensate for slower model improvements.

The Challenge of AI Development

• 🧩 The primary issue is that the performance of models like Behemoth may not match public statements about their capabilities, indicating they are not as advanced as optimistically projected.
• 🌍 Developing AI at scale is extremely difficult, suggesting only a handful of companies will power future intelligence.
• 🔒 This difficulty is good for large companies like Meta, as it creates a barrier to entry for competitors.

Open Source and Llama

• 🌐 Meta's Llama model is open-source, which has historically scored lower than other top models.
• ⚠️ Operating in an open-source environment can present limitations that affect the speed of model improvement.