#191 - Sora leak, Pixtral Large, OpenAI email archives
Summary
This episode covers the leaked OpenAI Sora video generator, discussing its quality and the controversy surrounding artist relations. The hosts critique Google Gemini’s performance compared to competitors and highlight Mistral’s new LeChat interface features. They also explore emerging trends in AI agents, voice interfaces via Eleven Labs, and Perplexity’s new shopping integration, debating the future of user experience and business models.
Topics discussed
Intro, sponsors, and show overview
Host introductions and listener reviews
Critique of Google Gemini's performance
Image generation: Ideogram vs Midjourney
OpenAI Sora video model leak and analysis
Llama 3.1 release and open source trends
Microsoft Copilot agents and AI terminology
ChatGPT personas and agent identity
ElevenLabs voice cloning and memory features
Perplexity AI shopping integration
Suno V3.5 music generation improvements
Flux image model and Runway Gen-3 Alpha
Runway image-to-video consistency tools
OpenAI leadership drama and Elon Musk
Amazon Alexa integration with Anthropic Claude
Physical AI robots and labor automation
DeepSeek R1 and academic research models
AI benchmarking and quantization research
Cerebras wafer-scale AI chips
San Francisco politics and tech industry
AI art perception and episode conclusion
Listen ad-free on Castria