AI Inside We Got Slopped in 2025
11 snips
Dec 31, 2025 Explore the intriguing world of AI as Nvidia licenses Groq's inference chips, signaling a tech consolidation wave. Dive into Meta's $2 billion acquisition of Manus, and the potential implications for AI strategy. Witness the rise of anti-AI sentiment in U.S. politics, with Bernie Sanders leading critiques of the technology's impact. Unpack the challenges and promises of AI-generated content, alongside reflections on 2025's developments and predictions for 2026's evolving AI landscape.
AI Snips
Chapters
Transcript
Episode notes
Inference Is The New Battleground
- NVIDIA licensed Groq to target ultra-fast, low-latency inference rather than training chips.
- This bet treats models as commodified and focuses on making model use cheaper and faster for apps.
Small SRAM, Big Performance Edge
- Groq's on-chip memory yields predictable, low-latency inference that suits chatbots and real-time uses.
- Combining NVIDIA's ecosystem with Groq tech could create platform dominance across training and inference.
Require A Verification Loop
- Use a chain-of-verification prompt to improve LLM reliability by forcing internal checks.
- Ask the model to list potential gaps, cite evidence, then revise its answer accordingly.