Ch 8 · The Chicane: This AI Beat the Algorithm in Practice, Not in the Exam

Last chapter, Sennacher got round two hairpins by never going fast. Today it has no choice: a straight of 144 metres and, at the end, a left-right-left chicane on a track just 6 metres wide, too tight to take flat out. On its cold run the chapter seven brain brakes, but not enough, and hits a marker post at the entry. Its first finish comes on attempt 54, at a crawl: 21.2 s. Then the Instructor brings a drill we hadn't seen: half of its tries start from a standstill anywhere between the 60 and 135 m marks, so it reaches the chicane at a different speed every time. Little by little it dares, and its best practice run, 15.59 s, beats the classic algorithm (15.65 s). But the official exam is a single run: 16.04 s, 0.39 s behind, because it starts braking at the 107 m mark and the classic waits until 117 m. ◀ Previous: Chapter 7 · Double Hairpin:    • Ch 7 · Double Hairpin: This AI Forgot How ...   ▶ Next: Chapter 9 · Chicane School:    • Видео   ▶ The whole series in order, from chapter 1:    • Sennacher · Season 1: School (Chapters 1-25)   Sennacher · Chapter 8: The Chicane Sennacher is an AI driver learning to race from scratch: from a cadet kart to Formula 1 lap records, over 300 chapters. One brain, never reset. Every result is real, failures included. A new chapter every 3 days. Subscribe:    / @sennacherai   Contents 0:00 Through the chicane at a crawl 0:19 A chicane six metres wide 0:41 The two yardsticks 0:54 Cold run: into a marker post 1:08 New here? The brain and how it learns 1:53 12 worlds, one chicane 2:07 First finish 2:32 Can it do it again? 2:45 Was it luck? 2:57 The Instructor is back 3:14 The drill: start anywhere 3:32 Braking for the chicane 3:52 Racing its own records 4:25 A clean habit 4:37 The official exam 5:29 The Instructor's verdict 6:19 Next: Chicane School 6:51 My funny highlights Real numbers from this chapter • Official exam: 16.04 s (classic algorithm: 15.65 s; physics limit: 14.96 s); its biggest slide: 3° • The chicane: left, right, left, 8 m radius, after a 144 m straight, on a track 6 m wide • First valid finish: attempt 54, after 12 minutes of driving (21.2 s) • The Instructor's new drill: about half the attempts start from a standstill anywhere between the 60 and 135 m marks • Best practice run: 15.59 s, already faster than the classic algorithm • Where the classic wins: in the exam it starts braking for the chicane at the 107 m mark; the classic algorithm at 117 m, the physics limit at 129 m • A clean habit: 93 of its last 100 attempts from the start line finish without sliding more than 25° • 6,790 attempts (3,188 in the Instructor's drills), 3,234 valid finishes from the start line, 226 crashes, 152 spins, 1,320 km • Brain: the same neural network since chapter 1 (630,000 adjustable numbers), never reset, trained on 12 karts at once Credits Voices: Kokoro-82M synthetic voices (narrator: af_heart; the Instructor: am_michael). AI voices, not real people. Music and sound: procedural, generated by code. Simulation: SennacherAI physics 1.0, built to scale. Animated, not real racing footage. Kart: Tony Kant Mini 60 (60cc Mini cadet kart). Track: School Track 08 · The Chicane. Subtitles: English, हिन्दी, Español, العربية, Français, Bahasa Indonesia, Português, Deutsch, 日本語, Tiếng Việt. #AI #Sennacher #ReinforcementLearning #MachineLearning #Karting