Bright Simons

1. I had my first driverless car experience last evening. 2. I was like the cat that got the cream. 3. No, I had no anxiety during the 20-minute ride. The highly organised traffic system and transit infrastructure made everything seemed like a breeze. Even before I got in, I had sized up the situation and was sure I was on firm ground. 4. A couple of things brought home to me some constraints on how fast AI will dominate our lifestyles in this decade. 5. Whilst locking the door, part of the seatbelt got jammed. There was no way to pull over and fix this without ending the trip completely. Lighting couldn't be adjusted by the rider either. 6. Someone called to remind me about the seatbelt. I explained the situation. They remotely unlocked the door for me to fix things. But they couldn't remotely adjust lighting either. 7. It became clear what was happening. Someone has to keep remote watch throughout the ride. There isn't yet contextually-intelligent AI in the car or remotely that can handle things. That probably means you still have one employee per car like a normal cab service. 8. Much of this is risk-management. I once wrote ponderously about how automation creates novel and rarely anticipated risks that require new and costly investments to surmount. See here: 9. In manufacturing, this "automation premium" is easily and quickly addressed by boosting volumes very early in the transition. Thus, we barely notice. 10. In complex services, like commercial self-driving cars, boosting volumes rapidly during the transition is not possible. Making it harder to drive down unit costs. 11. Further, much of what is needed to address risk and productivity-slack are not in the AI layer itself. They are in the "contextual" and "ancillary" layers. You have to make the underlying non-AI infrastructure more intelligent to allow the AI to integrate into society. That is a slog and costly as hell. 12. All the same, like I said, I was like the cat that got the cream.😉