Investors tend to put a lot of faith in technology. Two technologies in particular have captured the imagination of a lot of people this decade: electric vehicles (EVs) and artificial intelligence (AI). Both areas have gotten a ton of attention in the media, and both seem to promise to fix much that’s wrong with the world.

And, of course, the two areas are connected as auto companies race to develop autonomous – self-driving – EVs. Self-driving cars depend on AI to tell them when to stop, when to go, when to speed, when to swerve around an obstacle.

If you believe all you read, you may assume that EVs will soon be taking over the world’s roads, helping save the planet from the ravages of global warming. Many countries have set ambitious goals for EVs. Both China and India, for instance, have announced that between 2030 and 2040, EVs will account for all new car sales.

Wildly Inflated Projections

The U.S. hasn’t set any goals for EV production and sales. Still, in addition to Tesla, every major U.S. automaker is hard at work developing its own EV models. And just this week, General Motors announced plans for a self-driving EV by the end of the decade.

The reality, though, is that for different reasons, the rosy projections for both EVs and AI are wildly inflated. EVs are unlikely to account for even close to 100 percent of new car sales in any major country by 2040. A more realistic target date might be 2050 or even 2060. As for self-driving EVs, shortcomings in AI technology make relying on AI to drive your car akin to trusting your eight-year-old child behind the wheel.

Investing in Resource Shortages

This doesn’t mean investors should ignore EVs. They will be an important industry, and investors can benefit in a big way. But the biggest gains won’t come from investing in EV companies themselves. You won’t get far by putting your faith and investment dollars in Tesla.

The problem for EVs, one that could ultimately place an upper limit on their production in coming decades, will be chronic – possibly acute – shortages of the critical materials EVs require. There’s no guarantee these shortages ever will be remedied.

The U.S. is particularly low in these resources. This leaves us dependent on other countries, particularly China, which has a monopoly on critical ores. Equally important, China is dominant in supply chains. Obtaining the raw ores is just the first step in a complicated multi-step process. The ore has to be refined and then turned into the finished products EVs require. China has the greatest expertise and capability here, too.

The critical materials needed for EVs are:

• lithium
• cobalt
• graphite
• assorted rare earth elements

While China produces the bulk of these materials, there are a few non-Chinese companies that produce them as well and that are open to investors. This is the sweet spot for U.S. investors looking to benefit from the push for EVs. The interest, both in the private sector and by many governments, in promoting EVs will ensure a frantic effort to obtain the resources needed to build the cars. The shortages in materials and competition to obtain them will ensure rising prices.

In our latest issue of our publication Real World Investing (RWI) we offer five particularly compelling companies in the forefront of producing the materials EVs depend upon.

Does a Turtle Look Like a Gun?

As for the recent announcement by General Motors that it will have autonomous EVs on the road perhaps by the end of 2019, this defies everything we currently know about the capabilities of artificial intelligence. Just google the phrase “adversarial examples.” It will produce numerous examples of how easy it is to fool AI.

A recent article in New Scientist, for example, is headlined: “Visual Trick has AI Mistake Turtle for Gun.” It reported that when the pattern on the shell of a model turtle was subtly altered, AI misidentified it as a rifle. The article noted:

Previous studies have shown that changing just a few pixels in an image – alterations that are imperceptible to a human – can thrown an AI off its game, making it identify a picture of a horse as a car, or a plane as a dog. The model turtle now shows that an AI can be made to misidentify an object even from multiple angles.

Not Ready for Prime Time

I recently spoke with someone who says a major player in the autonomous field, a company that GM could be partnering with, has been testing cars in a small city in the Midwest. The cars, it turns out, are nowhere near ready for prime time. That’s in a pristine environment – one with freshly painted lane lines and stop signs that hadn’t been tilted. It was found that some of the self-driving cars being tested would wait for a minute or so at a stop sign when no other cars were waiting. Then it would pull into the intersection and stop.

It’s even less likely that autonomous cars will be reliable under real-world conditions. If GM has developed stunningly better algorithms, it would be a game changer, and we’ll be happily writing about it. But so far we’ve seen no evidence.

Meanwhile, you don’t have to sit on the sidelines while waiting for the technology to improve. If you put your investment dollars in the right resource companies, you should drive away a happy camper.