In The Loop Episode 1 | DeepSeek’s AI Breakthrough: Hype or Game-Changer? A No-Nonsense Breakdown
Table of contents
What is China’s DeepSeek Ai?
What did DeepSeek do different
How AI models are created and ran
What’s so great about DeepSeek?
Improvement #1: 8-bit floating-point numbers
Improvement #2: Compress key-value memory by 93%
Improvement #3: Predict multiple words at once
Improvement #4: Mixture of Experts (MoE)
Debate: Did DeepSeek use distillation?
How distillation works
What DeepSeek actually did…
What this means for AI
Why this was misunderstood
Why DeepSeek is a big deal
Reason #1: It's come at a time of big geopolitical tension
Reason #2: A new scaling law - AI Is moving from training to inference
Reason 3: Reduced costs through open-source and commoditized models
Reason #4: Big Techin frenzy: Is it time to go bigger and more expensive?
Market reaction, winners, and losers—My take
Why did DeepSeek affect NVIDIA?