The Math Behind “AI Will Replace Engineers” Is Embarrassingly Wrong

29.04.2026 Автор: Jovan EEN Welcome to my channel! I’m passionate about technology and content creation, and this channel is my way of sharing knowledge, and disproving misconceptions. 00:43 –Purpose of the Video 01:42 –Fear vs Reality in Career Decisions 02:22 –What “AI Replacing Jobs” Actually Means 03:07 –How Neural Networks Work (Basic Math) 04:03 –Transformers & Attention Mechanism 05:18 –Why Transformers Are Faster Than Older Models 07:31 –Why AI Scaled So Fast 08:42 –The S-Curve of Technology Growth 10:13 –Why Exponential Growth Doesn’t Last Forever 11:26 –Where AI Currently Sits on the Curve 12:07 –OpenAI Scaling Laws Explained 13:15 –Bigger Models = Smaller Gains 14:07 –“Test-Time Compute” & Thinking Models 15:12 –Efficiency Tricks (Quantization, MoE, etc.) 17:04 –Infrastructure Cost Problem 18:12 –Why AI Growth Was a One-Time Alignment 19:08 –Demos vs Real-World Deployment 20:22 –The Core Claim Revisited 21:01 –Hardware Limits: Memory Bottleneck 23:07 –Why Models Must Use Multiple GPUs 24:08 –Training vs Inference Complexity 25:06 –KV Cache & Conversation Memory 26:20 –Scaling Problem with AI Agents 27:18 –Why Faster Chips Don’t Solve It 28:08 –Amdahl’s Law & GPU Scaling Limits 30:31 –Manufacturing Bottlenecks 32:37 –AI = Electricity → Compute → Heat 33:17 –Data Center & Power Constraints 34:08 –Energy Math: Replacing Workers 36:27 –Why This Scale Is Unrealistic 37:25 –Jobs = Bundles of Tasks 38:07 –Accountability & Risk Problem 39:17 –Hallucinations &...