NeuroAI NEUROAINEUROAI.SITE
ESC

Wuhan handed its streets to driverless taxis — and a quiet commute war began

Baidu's Apollo Go (萝卜快跑) now runs fully driverless rides across Wuhan, serving more than half the city and forcing a real conversation about jobs and trust.

2026-10-18 · 1005 words · NeuroAI
Wuhan handed its streets to driverless taxis — and a quiet commute war began

A man in Wuhan watched his neighbour tap a phone and summon a car with no driver inside. He drives for a living, and the empty vehicle that pulled up felt less like a gadget than a competitor moving into his lane.

That small scene is now the daily default across large parts of central China's biggest transport hub. Baidu's robotaxi service, Apollo Go (萝卜快跑, literally "Radish Run"), has turned Wuhan into the clearest real-world test of what happens when driverless cars stop being a demo and start being a habit.

The car that crossed the river

Wuhan is a city split by the Yangtze, and crossing it used to mean a bridge crawl. On 27 February 2024, an Apollo Go vehicle completed the first fully driverless crossing of the Yangtze in the city — a 9-kilometre trip that took about 27 minutes, running over the Yangsigang and Baishazhou bridges.

According to the Wuhan municipal government, that ride was a milestone rather than a stunt. It was the first time an autonomous vehicle had been cleared to use an elevated bridge in the city, and it stretched the service from the Hankou and Hanyang districts south into Hongshan and the Jiangxia area, reaching as far as the Optics Valley (光谷) zone.

The same official update reported that by early 2024 the service had already logged more than 680,000 paid autonomous ride orders, with a single longest trip exceeding 95 kilometres. Wuhan had also opened its Tianhe Airport autonomous shuttle on 25 August 2023 — the first airport robotaxi connection of its kind in China.

How big the network already is

The scale is what makes Wuhan different from a closed test track. Reuters, reporting from the city in mid-2024, found that Apollo Go runs 24/7 and reaches about 7.7 million residents — more than half of Wuhan's population. The vehicles operate without a safety driver behind the wheel, though a remote service assistant can be contacted if something goes wrong.

Baidu has been explicit about where this is going. In a May 2024 statement cited by Reuters, the company said it planned to deploy 1,000 robotaxis in Wuhan by the end of 2024, and it had earlier forecast operating in 100 cities by 2030.

Wuhan's own numbers back up the footprint: the city says its open autonomous-vehicle test roads now total more than 3,378 kilometres of one-way mileage, covering roughly 3,000 square kilometres and touching over 7.7 million people.

Why Wuhan, why now

Wuhan is not an accident. It was among the first Chinese cities to publish commercial pilot rules for fully driverless operations, and it has leaned into what Beijing calls "new productive forces" (新质生产力) — a national push to lead in strategic technologies from EVs to autonomous driving.

The national backdrop is broad. Reuters reported that at least 19 Chinese cities were running robotaxi or robobus trials, and that seven had approved driverless testing without a human monitor on board, across players including Apollo Go, Pony.ai, WeRide, AutoX and SAIC Motor. China has moved faster than the United States in green-lighting road trials, betting that local competition will produce a home-grown lead.

The jobs question nobody avoids

Here is the conflict that will not go away. Ride-hailing and taxi drivers are among the first workers anywhere to face displacement from AI on the road. Reuters spoke with Wuhan drivers who said the robotaxis had taken a large share of short-distance trips and were already eating into incomes.

Baidu's line is careful: in an August 2024 statement it called the transition "gradual and well-regulated" and said its fleet "currently complements, rather than replaces, existing transport options," while also creating monitoring, testing and data-analysis roles. Economists quoted by Reuters noted the technology is hitting "pockets of the economy" rather than all of it — but warned that a nationwide rollout could push the government to intervene on unemployment.

That tension is the actual story. The technology works well enough to ride; the social contract around it is still being written.

What "driverless" really means here

The cars are built for Level 4 autonomy — they handle driving and environmental monitoring on their own within defined conditions. Baidu describes its current fleet as the world's first large-scale self-driving model supporting L4 at volume.

A fair caveat: reporting has noted that safety operators can still step in through 5G cloud-driving cabins when a vehicle meets something it cannot resolve. So "driverless" on the road does not mean "no human anywhere." It means the human moved from the passenger seat to a screen.

What readers can do now

  1. Try it as a traveller, not a spectator. If you visit Wuhan, Beijing, Shanghai, Shenzhen or Chongqing, check whether Apollo Go or a rival app operates there and take one short ride — the experience of a car with an empty driver's seat is the fastest way to understand the shift.
  2. Track the permits, not the hype. The real signal is which cities approve driverless (no-monitor) trials. A growing list of those cities tells you where autonomy is becoming infrastructure.
  3. Watch the labour debate. The Wuhan taxi-driver reaction is the canary. Where robotaxi expansion outruns retraining and transition policy, expect pushback.

Honest limitations

This article relies on two main sources — Reuters dispatches from Wuhan (August 2024) and the Wuhan municipal government portal (February 2024) — plus context on national trial counts from Reuters. We did not independently audit Baidu's internal ride-volume or safety records, and the "6 million orders" figure that appears in some secondary commentary was excluded because it came from a derivative outlet rather than a primary or corroborated source.

Our viewpoint is observational, not an endorsement: we describe what is deployed and the social friction around it, but we do not model accident rates, insurance liability, or long-term employment effects. Dimensions we left uncovered include the exact fare comparison with human-driven taxis, the remote-operator staffing model, and how riders with disabilities experience the service. Those need dedicated, sourced follow-up.

Related coverage

More in “AI in Action” → · Back to home · Markdown version