What No One Tells You About Artificial Intelligence
It seems as though everything today is branding itself as “AI.” There are AI chatbots, AI personal assistants, and even AI homes. The concept of “artificially intelligent technologies” evokes excitement in those who are looking for innovative solutions to their problems. Businesses know that finding and adopting the right AI technologies is crucial to their success. In 2016 alone, corporate investment in AI was $39B, and spending is only forecasted to increase. But there’s one glaring issue: not everything that claims to be AI actually is—and weeding out the fakes is crucial to any company looking to integrate AI into their business model.
It’s Tempting to Be Trendy
It’s easy and tempting to slap the “Artificial Intelligence” label on a product. AI is catchy, it piques interest, and it’s helpful in securing funding. Last year, a piece titled “How to convince Venture Capitalists you’re an expert in Artificial Intelligence” went viral. It offered advice such as “sporadically reference a paper on Arxiv as if everyone has read it” and “say you like PyTorch because Tensorflow is too slow.” Artificial Intelligence is cool—it sells—and because of this, the industry is full of false branding.
What is AI?
When trying to differentiate true AI from the crowd, it’s critical to define the term. According to Merriam-Webster, AI is “the capability of a machine to imitate intelligent human behavior.” Artificial intelligence is commonly broken down into four types. The first is purely reactive. This is the most basic form and doesn’t have the capacity to store information or use past experiences to inform future decisions. An example of this is IBM’s Deep Blue, which became so advanced at chess that it beat the reigning world champion. It could predict its opponent’s actions and choose the move which would be most advantageous for it. The second type is limited memory. Unlike purely reactive AI, these machines store data and learn from the past. Self-driving cars and personal digital assistants are examples of this. In the case of vehicles, AI observes patterns in traffic, which it combines with pre-programmed knowledge (street signs, traffic lights, etc.) to learn the best way to react to any given situation. The third type of AI is theory of mind. This kind of artificial intelligence is purely theoretical: a machine with the capacity to understand the world at large and to interact with it. Examples of this would be R2D2 from Star Wars or Jarvis from Iron Man. This final type of AI is self-aware AI. The creation of this technology won’t happen for years to come but would be characterized by a machine having thoughts and feelings, wants and needs—just like a human.
It’s important to note that only the first two types—purely reactive and limited memory—are currently possible to develop. The technology we have today still relies heavily on humans teaching machines how to think, perform, and work with data.
How to Spot a Fake
Data is the lifeblood of Artificial Intelligence. Above all else, true AI companies value the collection of structured and unstructured data. With this data, you can create something to build your AI base around so that it can learn and grow. Many companies that market themselves as “AI” simply use automation. Automated products need to be instructed, while true AI products learn. For example, you could automate a car to drive around a track. It would perform this function on its own, and as it was programmed. In order to deviate from this, a human would have to manually adjust the programming. An artificially intelligent car, however, could learn how to drive on a crowded street, stopping when there was an obstacle or swerving to avoid collision.
What Isn’t AI?
A lot, actually. Many technologies that seem “intelligent” are actually automated. Some chatbots, for example, converse only with pre-programmed responses. If you ask the time, it will pull that data and share it with you. While one may think this makes the bot an “intelligent machine,” the technology doesn’t learn or evolve based on its interactions. There is no improvement in capabilities the longer it performs its designated functions.
Another common misconception is that all robots are AI. While it’s true AI robots do exist, most robots are designed to perform a function—be it assembling a product or mowing the lawn—and don’t evolve within the scope of that task. A robot that can build a house is incredible, but it’s not coming up with improved or innovative floor plans and, therefore, is not AI.
Is Anything Actually AI?
Absolutely. There’s a reason so many people are jumping on the “AI” bandwagon—it’s the future of technology. Predictive purchases, self-driving cars, and even smart homes are all great examples of real AI that exists today. The key differentiator here is the ability of these products to adapt and grow. Take for example the iRobot Roomba 980. Unlike the first iRobot Roomba vacuum cleaner, this model scans a room, identifies obstacles, and learns the routes and methods that clean most efficiently.
The Bottom Line
Companies need to invest in AI. According to McKinsey, “AI can deliver real value to serious adopters and can be a powerful force for disruption…early AI adopters…have higher profit margins, and [we] expect the performance gap with other firms to widen in the future.” These technologies have the power to drastically improve how businesses operate and ease a variety of challenges faced today. The key is being able to distinguish what’s real from what’s not.