Artificial intelligence is a buzzword in 2017, and you can see it in the news and all over social media–especially with sci-fi sounding projects like Elon Musk’s new company, Neuralink. Then again, so are “machine learning,” “deep learning,” and “neural network.” There are a lot of different terms for the way non-human entities are gathering and utilizing information, but are they all interchangeable?
Not exactly. There are a few key differences between all of these terms as well as others like “machine vision.”
So, what do they each mean?
This is the big, broad field. If machine learning and deep learning were generalized pistachio and Talenti Sicilian Pistachio Gelato, then artificial intelligence is ice cream and similar creamy, frozen treats. They are a deepening rabbit hole of concepts that can be held within one another. Just the same, machine learning and deep learning are not the whole of AI, just as pistachio is not the only flavor of frozen treat.
But that doesn’t really tell us what artificial intelligence is, does it?
To create an artificial intelligence, one must understand what “intelligence” implies. Without getting into a philosophical loopty-loop of a semantic argument, let’s instead revert to the commonly accepted definition of artificial intelligence.
That in itself can be difficult, because what people perceive as AI tapers off as that smart solution becomes accepted mainstream as a normal form of computation.
For general purposes, artificial intelligence involves a machine that exhibits any of the characteristics of learning, perceiving, and using the knowledge they’ve gathered toward a useful and intentional application, or, even more broadly, doing something “smartly.”
But how can a machine do that? That’s where the other terms come into play.
Machine Intelligence is generally accepted to be the same thing as artificial intelligence, except that “machine intelligence” is more widely used in England, according to AI Business.
This is not to be confused with the difference between artificial intelligence and machine learning,which are not quite interchangeable.
According to Bernard Marr at Forbes, “Machine Learning is a current application of AI based around the idea that we should really just be able to give machines access to data and let them learn for themselves.”
Machine learning is what a lot of people think of when they think of artificial intelligence, but it is certainly not the whole. It is, however, one of the most promising AI technologies on the market.
Machine learning is involved in systems where machines transmute data into usable information and train themselves depending on what data they were given. It is one step up from deep learning, a subset of machine learning.
As previously mentioned, deep learning is a subset of machine learning (which is a subset of artificial intelligence). Two major characteristics of deep learning that set it apart from other forms of machine learning and AI are that:
- It uses massive data sets and
- It often uses learning techniques that mimic human decision making called “neural networks” which are based on the human brain, and they utilize large ones
These are the big bad beasts of the artificial intelligence world right now like DeepMind’s AlphaGo and the very popular DeepDream.
“Deep” isn’t a random term, and there isn’t much wiggle room on this one, unlike many terms in the artificial intelligence world. Rather, it is a scientific term that references the number of hidden layers in a neural network; deep is any neural network with more than one hidden layer.
Neural networks (Artificial Neural Network – ANN & Deep Neural Network – DNN)
Neural networks are based on the way biological nervous systems such as the human brain process information. Well, really, the human brain is a neural network–a network of neurons– and what most people in the AI community are actually referencing when they say “neural network” is an artificial neural network (ANN).
Artificial neural networks are what mimic the processing methods of the human brain and are often employed in machine learning. In this system, artificial neurons act as nodes that transfer information in much the same way our own neurons interact with their surrounding neighbors to communicate information.
Recently, Stanford University has even created computing techniques that function physically like the biological human brain (vs. programmatically like most neural networks) using silver nanoclusters implanted in nonconductive film that are placed over an electrode on each end as we discussed in this article.
Deep neural networks (DNNs), as mentioned above, are artificial neural networks that have multiple hidden layers between the input and output layers.
These aren’t the only types of neural networks. You also have convolutional deep neural networks (CNNs), recurrent neural networks, etc., but these are the two most commonly referenced in the media.
Supervised vs. Unsupervised Learning
The difference between these two types of learning are the difference between an elementary school student vs. a collegiate student researcher.
In elementary school, the answers are given, eventually, and so the student is trained in how to find the correct answer. This is supervised learning in the field of AI, where the machine is given the desired outputs.
On the contrary, the university researcher is experimenting and interpreting data. There may be no one to offer an answer or there may be no correct answer, they can only learn from what they are given. This is like unsupervised learning in artificial intelligence, where the machine is not provided a desired outcome.
Natural Language Processing
Although this is a very complex field in artificial intelligence with many facets as well as one that is highly relevant today, natural language processing (NLP) can be described as the way a machine interprets and understands human speech. Human speech is what is considered a “natural language” in comparison to an artificial language, such as a computer programming language.
NLP is what allows common AI applications like Siri, Alexa, and Cortana to understand what users want when they ask.
Machine Vision vs. Computer Vision
Computer vision involves all of the technologies that allow machines to see the world around them, and also the process of creating machines that are able to see and process visual information better than humans.
Machine vision, on the other hand, is used for specific industrial or practical applications, and is commonly known for such tasks as inspections.
Are there any other terms surrounding AI that you see crop up a lot in the media? Let us know so we can help others understand what all is going on in the wild and wonderful world of artificial intelligence!