

NCI: AI Chip Wars: LPUs, TPUs & GPUs w/ Jonathan Ross, Founder Groq
13 snips Jan 17, 2024
Founder of Groq, Jonathan Ross, discusses the Genesis of Grok, AI chip technology, business model, CPU vs. GPU vs. TPU vs. LPU, AI's future impact, AI in games, limitations of human knowledge, winners and losers in AI, Google's role, AI's impact on the semiconductor industry, open source importance for AI safety.
AI Snips
Chapters
Transcript
Episode notes
Groq's Speed Demo
- Jonathan Ross demonstrated Groq's AI speed on a Llama 2 70B model.
- It answered a stock query instantly, much faster than OpenAI's GPT.
Groq's Business Model
- Groq offers an API and sells systems/chips, focusing on latency-sensitive applications.
- They primarily target enterprise customers needing faster responses.
CPU, GPU, TPU, and LPU Comparison
- CPUs prioritize ease of programming, while GPUs excel at parallel processing for graphics.
- LPUs combine parallel computation with sequential processing, ideal for language and other sequential data.