When automated cars become common enough, I assume someone will develop a means of communication between vehicles, even from different makers. This will allow them to navigate road conflicts. In the case where human drivers are still common, maybe these can be usable by humans also. For example, a robotic hand on the dash board that can wave, beckon or give the finger or other readily apparent gestures to other vehicles .
Andrew, thanks for a thought-provoking piece. You nail the reality that mixing cautious software with human improvisation will create a frustrating culture clash. But concluding this 50% threshold is the "most dangerous" era reveals a few blind spots.
First, friction does not equal fatality. A robotaxi freezing at a four-way stop is exasperating, but by pausing it has neutralized its kinetic threat. The actual danger is the impatient human swerving into oncoming traffic. Swapping out 50% of easily distracted drivers for 360-degree vigilant AVs removes massive systemic risk, even if it temporarily increases gridlock at some poorly designed intersections.
Second, the hypothesis treats the tech as static. By the time we hit 50% adoption, sensor suites will have commoditized into affordable factory options. When vehicles communicate via V2V telemetry, that four-way stop negotiation happens in milliseconds, bypassing human hand-waving entirely.
Finally, capping AVs is the exact wrong policy reflex. If the 50/50 mix is truly the most painful phase, why artificially trap our transit grid in it for decades? Instead of protecting the status quo, we should incentivize the transition. Look at Norway’s EV adoption: they built an incentive structure that made the legacy tech economically impractical.
Market pressures will soon do this for AVs. Skyrocketing insurance premiums for accident-prone humans will eventually collide with affordable AV subscriptions, fractional ownership, and cheap private autonomous options.
The transition will be bumpy, but the solution is to accelerate through the friction, not pull the emergency brake.
AV commentary is rife with experts making strong claims in public without strong evidence. I reread your column and see many examples. Here are two:
- “Robotaxi fleets, absent regulation, begin to deadhead—cruising neighborhoods while seeking rides, occupying road space without passengers. Each empty vehicle represents only a small inefficiency, but in aggregate they reshape traffic patterns, curb access, and street-level expectations.”
- “At the midpoint of AV road share, the system no longer risks mere strain, but failure.”
I chose pointed language because I think it is harmful to the quality of public conversation on an important topic to assert strong claims without strong evidence.
I am impatient with this. What 'strong evidence' would you have suggested I bring forward about events that have not yet happened, whether we choose to call them 'scenarios' or 'predictions'?
As for the rest of your concerns, by all means write your own essay and invite me to read it, rather than assuming you have standing to instruct me, or to grade mine against your personal standard.
When automated cars become common enough, I assume someone will develop a means of communication between vehicles, even from different makers. This will allow them to navigate road conflicts. In the case where human drivers are still common, maybe these can be usable by humans also. For example, a robotic hand on the dash board that can wave, beckon or give the finger or other readily apparent gestures to other vehicles .
Andrew, thanks for a thought-provoking piece. You nail the reality that mixing cautious software with human improvisation will create a frustrating culture clash. But concluding this 50% threshold is the "most dangerous" era reveals a few blind spots.
First, friction does not equal fatality. A robotaxi freezing at a four-way stop is exasperating, but by pausing it has neutralized its kinetic threat. The actual danger is the impatient human swerving into oncoming traffic. Swapping out 50% of easily distracted drivers for 360-degree vigilant AVs removes massive systemic risk, even if it temporarily increases gridlock at some poorly designed intersections.
Second, the hypothesis treats the tech as static. By the time we hit 50% adoption, sensor suites will have commoditized into affordable factory options. When vehicles communicate via V2V telemetry, that four-way stop negotiation happens in milliseconds, bypassing human hand-waving entirely.
Finally, capping AVs is the exact wrong policy reflex. If the 50/50 mix is truly the most painful phase, why artificially trap our transit grid in it for decades? Instead of protecting the status quo, we should incentivize the transition. Look at Norway’s EV adoption: they built an incentive structure that made the legacy tech economically impractical.
Market pressures will soon do this for AVs. Skyrocketing insurance premiums for accident-prone humans will eventually collide with affordable AV subscriptions, fractional ownership, and cheap private autonomous options.
The transition will be bumpy, but the solution is to accelerate through the friction, not pull the emergency brake.
Is it ‘50%’ where it will be a real problem, or will that start at more like 10%, when in some times and places, there will locally be 50%?
And those places will disproportionally impact the broader experience.
Great draft column on an important topic. Issues to fix before publishing ;-):
1. Phrase the scenarios as possibilities rather than predictions
2. Observe that both AVs and people will be learning continuously
3. Call out the Autopilot approach as having nasty failure modes unique to it
Is there a reason you phrased this so offensively?
Sorry I offended you.
AV commentary is rife with experts making strong claims in public without strong evidence. I reread your column and see many examples. Here are two:
- “Robotaxi fleets, absent regulation, begin to deadhead—cruising neighborhoods while seeking rides, occupying road space without passengers. Each empty vehicle represents only a small inefficiency, but in aggregate they reshape traffic patterns, curb access, and street-level expectations.”
- “At the midpoint of AV road share, the system no longer risks mere strain, but failure.”
I chose pointed language because I think it is harmful to the quality of public conversation on an important topic to assert strong claims without strong evidence.
I am impatient with this. What 'strong evidence' would you have suggested I bring forward about events that have not yet happened, whether we choose to call them 'scenarios' or 'predictions'?
As for the rest of your concerns, by all means write your own essay and invite me to read it, rather than assuming you have standing to instruct me, or to grade mine against your personal standard.