Pick up a magazine and browse through tech blogs, or chat with your peers at industry conferences. You’ll quickly notice that almost everything from the tech world deals with AI or machine learning in one way or another. And the way they discuss AI, it almost sounds like they are promoting the idea that AI can solve all needs! Artificial intelligence is here to save humanity!
While it’s true that we can do a lot with AI technology, we don’t understand the full meaning of the word “intelligence. Intelligence means a system in which humans can have creative conversations – a system that has ideas and can develop new ones. The debate is about terminology. Today’s “artificial intelligence” usually describes some extension of human capabilities, such as object or voice recognition, but certainly not the full potential of human intelligence.
Thus, “artificial intelligence” may not be the best way to describe the “new” machine learning technologies we use today, but the train of change has already left. AI, meaning neural networks or deep learning as well as “classical” machine learning, will eventually become a standard part of the analytics toolkit as well.
Now that we have entered the AI revolution (or rather evolution), it is important to look at how the concept of AI has been absorbed, why it has been absorbed, and what it means in the future. In this article we take a deeper look at why AI and even some of its misunderstood versions are currently receiving such high levels of attention.
Why is artificial intelligence exploding now?
In current times, AI or machine learning is often described as a relatively new technology that has suddenly matured and only recently transitioned from the conceptual stage to application integration. It is widely believed that the birth of standalone machine learning products has only begun in the last few years. In fact, significant developments in AI are not new. Today’s AI is a continuation of the advances made in the last few decades. This change, and the reason we see AI appearing in so many places, is less about the AI technologies themselves and more about the technology surrounding them-namely, the development of data generation and processing capabilities.
At the end of the day, people may simply want to try different technologies on the same platform, without the limitations of certain software vendors or the inability to keep up with current advances in the field. That’s why open source platforms are leaders in this market; they allow practitioners to combine current state-of-the-art technologies with the latest cutting-edge developments.
Deep learning will become part of every data scientist’s toolbox as teams become aligned in their use of machine learning to achieve their goals and methods. For many tasks, adding deep learning methods to the mix will provide tremendous value. We’ll be able to leverage pre-trained AI systems, and we’ll be able to merge existing speech or voice recognition components. But eventually, we will realize that, just like classical machine learning before it, deep learning is really just another tool.
What’s next for artificial intelligence?
Just like 20 years ago, people have great difficulty in trying to understand what AI systems have learned and how they make predictions. This may not be important when predicting whether a customer will like a particular product. However, the problem arises when explaining why a system that interacts with humans will operate in an unexpected way. Humans are willing to accept “human failure,” but we will not accept the failure of an AI system, especially if we cannot explain why it failed (and correct it).
As we become more familiar with deep learning, we will realize, as we did with machine learning 20 years ago, that although the system is complex and the amount of data it is trained on is large, without knowledge of the domain, we will not be able to understand what is happening. But without knowledge of the domain, it is difficult to understand. Human speech recognition is so effective because we can usually fill in the gaps of lack of understanding through context.
Today’s AI systems do not have that deep understanding, and what we see now is superficial intelligence, the ability to mimic isolated human recognition abilities, sometimes even performing better than humans on these isolated tasks, and training systems with billions of examples simply requires having data and access to sufficient computational resources.
In the future, “true” artificial intelligence will certainly be a direction of research, but for now, people can use it to do their jobs faster and better.