It was more than 20 years ago that a computer powered by artificial intelligence first beat humans at their own game. In a chess match that reverberated throughout the world against world champion Garry Kasparov, IBM’s Deep Blue successfully defeated its human opponent, making it crystal clear: Computer capabilities had surpassed humans in certain challenges.

As we enter 2018, the opportunities to utilize AI to augment human capabilities have never been greater. But while dominating a game of chess is one thing, the bigger question remains: Has AI come far enough to be entrusted with far more important, life-and-death decisions?

The autonomous car industry is a prime example of this dilemma. One key goal of the industry is to replace error-prone human drivers with technology-driven cars that will make far fewer mistakes, not only smoothing the flow of traffic but also potentially saving countless lives that are cut short every year due to human fallibility. To this end, self-driving cars are set to generate more than 4,000 GB of data per day. Nevertheless, the car industry is continuing to develop capabilities to better predict traffic, road hazards and safer routes.

But the even bigger dilemmas involving life-and-death decisions are far from being cracked by AI. Take a classic case: If someone on the sidewalk darts into the road and the car has the option of hitting the pedestrian or swerving but risking the driver, which does it choose? And what parameters does it take into account when making this split-second decision?

That is why the state of California, a bellwether of the autonomous driving industry, is already discussing legislation that will essentially mandate the necessity of a human taking control of an autonomous vehicle in certain perilous situations. To put it another way, we can have AI crunch numbers to steer our wheels, but putting our lives completely in its hands is not yet an option.

The health care industry is another sector AI is poised to radically transform. In fact, use of AI in the health sector is expected to have a compound annual growth rate of 40% between 2017 and 2025. The amount of data being accumulated by the industry is growing rapidly as records such as medical histories are being digitized and stored on the cloud. Since machine learning requires copious amounts of data to improve, AI will gain from the data accumulation and continue to play an increasingly significant role in everything from diagnostics to pharmacology and beyond.

But this, too, begs the question: Can AI be trusted to make life-and-death decisions when it comes to our health care?