The Future of AI with Perplexity CEO Aravind Srinivas

Shifting AI Development Paradigms

• 💡 AI progress is moving from pre-training to post-training, focusing on reasoning and specific tasks.
• 🚀 This shift involves training AIs to be useful assistants for complex workflows and research, a direction all model labs are pivoting towards.
• 🧠 DeepSeek is noted for its open-source reasoning model and system optimizations, accelerating progress and influencing American labs.

Perplexity's Strategic Foundation

• ✅ Perplexity made a conscious choice to leverage existing models and bet on the rise of open-source efficiency to build its user base and mind share.
• 📊 Data for post-training comes from user queries, feedback, human evaluators, and synthetic data generated by larger LLMs.
• 🎯 The company competes with incumbents like Google by exploiting their high infrastructural costs, amplified reputation risk, and ad-revenue-dependent business model.

AI's Societal and Economic Implications

• ⚠️ AI presents both utopian potential (democratized knowledge) and dystopian risks (job displacement, economic inequality).
• 🧑‍💻 Even programmers are being significantly impacted, highlighting the need for individuals to adapt and focus on creative problem-solving and