Driverless Cars Are a Dead-End Technology

Why It Matters

The prohibitive data, power, and cost requirements make fully driverless cars unsustainable, forcing the industry to reconsider business models and regulatory frameworks.

Key Takeaways

• Level‑3 systems generate 34 GB of data per minute, straining resources.
• Full self‑driving would need computing power exceeding Facebook’s data centers.
• Current autonomous fleets still rely on remote operators for edge cases.
• High sensor costs (~$180k per vehicle) make profitability doubtful.
• Human‑in‑the‑loop safety failures highlight ergonomic and liability risks.

Summary

The video argues that fully autonomous vehicles are a technological dead‑end, citing the massive data, power and cost burdens they impose. It highlights Mercedes‑Benz’s Drive Pilot (Level 3) producing 34 GB of sensor data each minute and estimates that a fleet of driverless cars would consume four orders of magnitude more computing power than all of Facebook’s data centers, demanding electricity comparable to a small country.

Key data points include 4,000 GB of daily data per car, the need for ultra‑low‑latency processing, and sensor packages costing roughly $180,000 per vehicle. The discussion also notes that companies like Waymo still depend on remote operators for edge‑case handling, and that Uber’s 2018 fatal crash exposed the dangers of human‑in‑the‑loop designs where drivers lose situational awareness.

Notable examples cited are Mercedes’ Level 3 “Drive Pilot,” Waymo’s Gen‑5 “fleet response” system that calls a remote operator, and Uber’s negligent‑homicide case after a self‑driving car struck a pedestrian. The speaker emphasizes that even the most advanced prototypes require a human safety net and that the economics are untenable without a clear path to profitability.

The implications are clear: investors and regulators should temper expectations for fully driverless taxis, focus on incremental Level 2‑plus assistance, and address safety‑critical ergonomics before scaling. The high energy demand, prohibitive hardware costs, and liability risks suggest the industry may pivot toward hybrid models rather than pure autonomy.

Original Description

Watch Episode 262 of #Railnatter here: https://www.youtube.com/watch?v=1vaMlaugtBI

Support #Railnatter at https://patreon.com/garethdennis. Merch at https://merch.railnatter.uk. Join in the discussion at https://discord.railnatter.uk. You can also buy my book #HowTheRailwaysWillFixTheFuture: https://bit.ly/HowTheRailways