Episode Description

AI usually lives in the cloud, but this episode is about where you can actually feel it. Mia opens with two vending cabinets that look similar but feel like different eras — one waits on a cashier, the other grabs and bills in under a second. David explains that gap as three generations of recognition tech: manual RFID tags, gravity sensors, and today's AI-vision combo of cameras plus weight checks. The "smart brain" inside is a neural-network module that does on-device inference, running smart scales just as smoothly as it runs cabinets. The episode closes by looking ahead to where the same kind of AI is heading next.

Chapters & Timestamps

0:00 — Hook: the “smart” vending cabinet that took three minutes in Singapore

1:00 — Cold open: Shanghai’s 0.3-second “grab-and-go” — two cabinets that feel a generation apart

2:10 — The roadmap: three generations of recognition tech behind smart retail

4:45 — Gen 1, the RFID era: manual stickers, €0.05 tags, an 8% failure rate and Paris’s “Sold Out” cabinet

6:20 — Gen 2, the gravity-sensing era: Southeast Asia & India — and Bangkok’s cashew-vs-almond mix-up

7:20 — Gen 3, the AI-vision era: cameras + gravity, 99.9% accuracy, crowds & chip flavors

9:20 — Under the hood: NPU modules — SIM9650L’s 12 TOPS and the 0.5%-of-revenue operating cost

10:40 — Design DNA: Japan’s confirm-every-step vs. China’s seamless — and accessibility ahead

12:40 — Beyond cabinets: AI self-service scales — 15-second checkout, a 0.5 s mango ID, Munich’s Pick&Go

14:40 — From bakeries to buffets: discounts priced right & a 15→3-minute checkout

16:35 — Price & market: $150–$1,000 units — an $8B→$15B market by 2030

17:20 — Wrap-up: AI as helper, not replacement — four arenas ahead (intelligent computing, autonomous driving, robotics, edge AI)