The first car appeared in late 19th century Europe. It would soon revolutionize transportation and society as people knew it, offering new independence and economic possibility. It wasn’t until the 1939 World’s Fair that people envisioned a futuristic society in which humans didn’t have to always be in the driver’s seat, literally and metaphorically. Cars that could drive themselves were a fantasy that would signify a new age.

While it seemed like the Twilight Zone back then, that future is closer than we may think. The self-driving car has been in stages of research, testing and production for the past several years. In 2017, the hype for driverless cars reached a peak with companies like Google, Uber and more starting to produce them. If the idea comes to fruition, connected and automated vehicles (CAVs) could dramatically reduce the 1.3 million people per year who die in automobile-related incidents. They could also improve traffic conditions in congested metropolitan areas, reduce stress, and free up humans to spend time outside the confines of the vehicle.

However, excitement from these technological advancements have met some roadblocks, no pun intended. In March 2018, two tragic accidents alarmed everyone in the industry. A self-driving Uber struck and killed a pedestrian, and a Tesla autopilot Model X car killed a customer when it crashed into a lane divider.

These incidents poured cold water on a burgeoning idea. Google’s Waymo, which didn’t have any such tragedies, slowed its rollout of vehicles in showrooms. Manufacturers need to conquer the remaining technical challenges before our roads consist of empty driver seats. Now, experts are predicting that self-driving cars may not enter the market for several more decades. Automation poses a challenge to self-driving cars succeeding, but it is humans who are, ironically, the real hindrance.

Artificial intelligence shines brightest in the self-driving car. So what’s missing? Human intelligence in situations like avoiding a jaywalking pedestrian. What’s more, the computer systems that CAVs rely on are at risk for hacking. Fiat Chrysler recently recalled 1.4 million Jeep Cherokees after discovering they were easily hackable for functions like acceleration and radio. The CAV would rely 100% on similar computer systems, requiring much more robust security. Another challenge is ensuring sensors read signs correctly after suffering  harm by things such as  vandalism or weather damage.

How can the automobile industry overcome these challenges and persevere with the self-driving car? Finding a way for humans and computers to mesh well on the road is key to breaking through barriers to this technology entering the market. Until then, fully automated vehicles pose great risks to humans, whether they are in the vehicle or on the sidewalks as pedestrians. Emergency vehicles like police and ambulances will likely still need human drivers who can communicate with the automated vehicles on the road.

At the moment, driverless cars remain in controlled testing environments where structures like lane-changing systems, post-accident braking and crash avoidance systems are being improved. The University of Michigan is leading the effort with its autonomous city. There may be a long way to go, but the future is bright with the combination of seamless technology and the power of the human mind.