Automakers are waking up to the realities of autonomous vehicles. In 2016, Ford said that it’d be mass producing vehicles capable of driving without the need for a steering wheel, pedals or a human — Level 5 Autonomous — by 2021. At a recent meeting at the Detroit Economics Club, CEO Jim Hackett walked that back, saying that while the 2021 goal was still in place, but the application would be narrow because of the complexities involved.

Getting such an admission out of Tesla CEO Elon Musk might prove difficult, but a future where our personal vehicle is a robo-chauffeur is not necessarily imminent. At the recent AutoSens conference in Detroit, experts agreed that midway through 2019, the timeline looks a lot different from estimates of five years ago.

“Building an autonomous car is incredibly hard,” May Mobility CEO Edwin Olson said during a panel covering autonomous expectations. Human drivers, on average, drive 100 million miles between fatalities. “For a car to drive that far without making a serious error is a huge leap from where the technology is now.”

Using Moore’s Law as a guiding point, Olson estimated that even with autonomous performance doubling every 16 months, it’ll still take 16 years for self-driving vehicles to match that of consumers.

“We’re still a factor of 10,000 short in the performance of the best autonomous vehicles today and human drivers,” he said.

The problem is, if the automakers and Silicon Valley startups knew how hard it’d be to get to full autonomy, they weren’t telling us. March’s fatal Tesla crash in Florida happened within ten seconds of the driver engaging Autopilot, according to an NTSB investigation. Autopilot is the Level 2 autonomous system Musk estimated would allow drivers to sleep at the wheel, safely, within two year’s time. That was 2017.

“There was a lot of buzz about things happening really quickly,” Robert Stead, managing director of AutoSens organizer Sense Media Group, said. Some of that comes from pressures to be first.

After years of hearing about autonomous taxis and fleets each January at CES, the fatalities and wait-time has had an effect on the public.

“The consumers have this perception of speed to market,” he said. “It’s hard for the automakers to resist the temptation to tap into that.”

To Stead, as the understanding of how difficult the path to Level 4 or Level 5 has sunk in — namely, the transference of control from car to driver — projections have become more realistic.

In addition to Sense Media, Stead is also the chair of P2020 working group of the IEEE Standards Association, which is working on a standard for automotive camera systems. He’s been in discussions about developing a Level 2.5, focused features like emergency reverse braking, and incremental safety systems that will slowly get us to Level 5.

“It’s better than rushing into technology that either hasn’t been fully tested, or we’re not ready to do,” Stead said. “I don’t think we need to worry if we don’t have fully autonomous vehicles zooming around every city in the next five years.”