Artificial intelligence (AI) gets more advanced every year, but there are still some major limitations keeping us from seeing a futuristic reality that includes robot butlers and near-complete societal automation.
Fortunately, some of these limitations are on the verge of being overcome, and if you watch and plan carefully, you’ll be able to take advantage of those improvements for your business.
Here are some of the biggest hurdles we may overcome as early as 2020:
1. Unsupervised learning
Right now, most AI systems “learn” new information through a kind of structured force-feeding, relying on information given to those systems by humans. However, this form of “supervised learning” isn’t scalable, and doesn’t mimic the way that human beings naturally learn.
In fact, we humans are immersed in our environments, perceiving pretty much everything that crosses our path and naturally filtering out what’s unimportant. We also experiment to learn how objects interact and how the world works.
Currently, we’re still a few years away from machines that can learn this way, but when we get there, we’ll have the ability to use them to generate or augment ideas, and produce concepts we couldn’t come up with on our own.
2. Creativity and abstract thinking
Humans tend to think of ideas — and solutions to problems — in terms of abstractions. For example, think about a horse. Chances are, you aren’t thinking about a very specific example of a horse, and you don’t need to list out all the required “ingredients” that constitute a horse. Instead, you conceptualize the general idea of what a “horse” is.
A modern computer, on the other hand, would need thousands of examples of horses to understand what “horse” means, and even then, it would have to define this conceptual equine concretely and completely to use that idea in any application. If we want machines that can take raw data and turn it into intuitive concepts that can be grasped, we’ll need to create machines that can think abstractly.
Read the source article at Entrepreneur.