The supposed argument for autonomous vehicles is that they will cut back on fatalities stemming from car crashes by over 90 percent. That cheery-eyed view masks an important point: It’ll be a long time before—if ever—that roads will be totally dominated by robot cars. For a while, it’ll be a blend of manually-operated and autonomous vehicles. And from the Governors Highway Safety Association, that’s a problem.

“The research and media attention given to autonomous vehicles often overlooks the safety implications that a mix of driver-operated and autonomous vehicles will bring,” said Dr. James Hedlund, a former senior official with the National Highway Traffic Safety Administration and co-author of the study, in a statement.

“Unfortunately, ignoring the driver side of the equation may negate many of the expected safety benefits.”

The study is making a point that’s often missed in coverage of autonomous vehicles: There’s a ton of hype about what the technology can potentially accomplish, and not much of a measured approach to how we’ll get there. The study brought to mind a report from a couple years back about the cost of adding autonomous technology to a vehicle. That said, by 2025, self-driving gadgets will add $7,000-$10,000 to a car’s sticker price; a decade later, it’d drop to about $3,000. Still a decent chunk a change. The point? It’s going to be sometime before a significant chunk of society is tooling around town in a robot car.

In 2009, screenwriter Charlie Kaufman spoke to a reporter about the excessive expectations that…

And that doubles back to this new study. Because the reality is, there’s going to be a blend of manually-operated and autonomous vehicles on the road for years to come. So the governor’s highway association suggests a few recommendations to deal with it:

– States should develop education campaigns on the benefits and risks of AVs, how to operate vehicles with some autonomous features safely, and how to share the road with AVs. – States should wait until model laws and regulations have been developed to encourage a common structure and prevent a patchwork of inconsistent laws and regulations that may delay AV implementation. – States must identify vehicle automation levels in their registration, driver licensing and crash information systems. Police crash reports should be designed to help facilitate comprehensive and accurate data collection.– States should include law enforcement in their planning, as AVs raise many issues for law enforcement, including officer safety, enforcement procedures and vehicle identification.

“Drivers are often forgotten as we discuss autonomous vehicles, but cars driven by humans will be on the road for at least another generation,” said the association’s executive director, Jonathan Adkins, in a statement. “As human drivers begin to share the road with different levels of autonomous vehicles, states will need to stay informed, be patient and be flexible.”

Great points—especially for writers covering AV tech to take into consideration. For now, human driving still matters.