Artificial intelligence (AI) has long captured the human imagination, but it has been relegated to the theoretical realm until recently. Thanks to fast processes and growing volumes of data, the technology is becoming a reality—and is yielding real benefits. What AI trends can we expect in the next few years?
Articles in the press often talk about AI getting too smart for our own good. Elon Musk worries that it could destroy people, while the late Stephen Hawking was concerned that it could replace us altogether. But not everyone worries that the sky will fall tomorrow. Bill Gates has voiced concerns about the technology, too, but says that “people shouldn’t feel the problem is imminent.”
Positing a future in which AI is as smart as humans—or smarter—assumes that we will solve one of the technology’s most long-standing problems: its ability to think like people. “Hard” or “broad” AI would consist of computers that are able to learn anything and develop reasoning capabilities just like humans—only at lightning speed. Every AI-powered villain you’ve seen on the silver screen is based on this, but this doesn’t actually exist today. We don’t yet understand enough about how humans think, let alone have the computing power or software capabilities to duplicate it.
Today’s AI is “soft” or “narrow.” It trains statistical analysis algorithms on large volumes of data to detect patterns, and then to spot these patterns in new data. The result is software that doesn’t think at all, but automatically completes narrowly defined tasks.
This AI can be highly useful in certain situations. Andrew Ng, former head of the Stanford Artificial Intelligence Laboratory, says that if a task takes a person less than a second, AI can probably automate it. That’s great for analyzing simple inputs and quickly producing an output, such as transcribing audio, recognizing faces, approving loans, or deciding if a piece of equipment will fail.
A great source of information on future AI trends is Rodney Brooks, professor of robotics at MIT and cofounder of iRobot (the creator of the well-known robotic Roomba vacuum).
Brooks warns that everything takes longer than we expect, and that successful engineering doesn’t mean immediate adoption. One example is the driverless car. These are on the road today, and he predicts that there will be dedicated lanes for them on public freeways by 2021, but he doesn’t foresee them really starting to change society by making massive inroads into American cities until at least 2027, and maybe not before 2031.
When it comes to robots, some of the most basic tasks that we do today without thinking will likely stump AI for a long time to come. Don’t expect to see robots that can deliver packages from a vehicle to inside your front door until at least 2028, says Brooks. And we shouldn’t expect to see affordable devices that can move easily around U.S. homes on their own before 2035.
But what about conversational bots, like the ones currently driving online interactions in call centers? Brooks is more optimistic here. He predicts that we’ll see bots on the market that are able to understand long-term context and not fall easily into replicable patterns by 2025. As for a robot that has any idea of its own existence and understands the existence of people? Brooks doesn’t think that’ll happen before 2050—so we clearly have a long way to go.
While we wait for some of these predictions of future AI trends to play out, we can still make great strides with the technology as long as we put it in context. We must also take the right strategic approach and ensure that we have the appropriate resources.
Companies can use technologies like machine learning (the most commercially popular branch of AI) to automate those simple human tasks and save money. They can also make gains in customer service and experience by processing data and queries more effectively.
Machine learning is a data-hungry technology, though. To feed AI algorithms, companies must create organizational and technological structures that enable employees to easily share and aggregate that information while preserving security and privacy. This is a tricky problem that needs senior management buy-in.
Data alone is not enough. The skills to manipulate it are in high demand. Employing data scientists that understand your industry will be another important part of the puzzle. You must also be prepared to integrate these AI improvements into your business workflows, which can require a significant investment in technology.
When all this is done, companies can gain commercial advantages over the competition. However, these first forays into AI will create a platform for a much longer journey. It seems that AI technology is the gift that will keep on giving. It will just do so on a slow, steady basis.
Learn more about what CEOs need to know about AI strategies.