AI, ML, DL, and several other two-letter words are on everyone’s mind these days. These three things are closely related in the space of big data and emerging technology. While similar, they aren’t quite the same thing. Given the confusion and our perspective on it, we thought we’d start with a basic primer on artificial intelligence, machine learning, and deep learning.
Artificial Intelligence (AI) deals with the concept of machines being able to perform tasks that we would consider ‘smart.’ Typically this would include reasoning, problem solving, or perception; qualities we would normally associate with human abilities.
There are several examples of what we might call artificial intelligence in today’s world. The most prominent would be the advent of voice assistants such as Alexa or Siri.
To the average consumer, Alexa and Siri may seem pretty smart! However, true ‘artificial intelligence’ does not really exist just yet. That is, a machine that appropriates human intelligence.
Siri and Alexa are, at best, a form of behavior-based machine learning with what some would refer to as ‘augmented intelligence.’ They mimic the abilities of an AI system and of human behavior. However, they aren’t truly learning or solving problems on their own outside of a predetermined framework.
Systems such as Alexa or Siri are designed to enhance or aid human knowledge and intelligence rather than replace it. Much the same way an augmented reality program enhances the environment, or the experience of the person in that environment, rather than replacing the environment itself with something different.
Machine learning is the concept of machines ‘learning’ how to perform tasks without being explicitly programmed to do them. It is a subset of artificial intelligence but it does not require that the machine perform human-like tasks or behaviors.
Rather, it only requires that the machine can acquire new skills without someone programming it (in a traditional sense) to do so. The behaviors aren’t simply a function of executing a program written by a human. Guidelines are set for performing a task but the execution may change as more data becomes available.
Machine learning focuses on giving a computer the ability to learn to do new things over time when given more data to work with. Consider a navigational app. The system knows you are trying to get from point A to point B and that there are multiple ways to get there. As more data becomes available such as weather or traffic reports, tt is able to suggest alternative routes. The end goal is the same: point A to point B. The process for getting there has just changed based on the available data to become more efficient.
Another example of this might be the SPAM filter on your email which identifies junk email. As you and other users report SPAM, the system learns to identify it better. As time progresses, it becomes better at it.
Deep learning is a subset of machine learning that functions in a different way. For example, Netflix and Hulu employ a form of machine learning. When they make recommendations to you based on your viewing habits, that’s learning.
However, these recommendations follow a set of pre-programmed conditions. In theory, it allows it to make better recommendations as more data becomes available.
If the recommendations being made are poor, there’s a problem with the algorithm being used and it will need to be adjusted.
In a deep learning model, the system is capable of determining on its own whether the algorithm is functioning properly. If it determines that it needs tweaking, it may adjust itself. This process is built to follow a logical pathway, similar to the way the human brain functions.
The most famous example of deep learning might be Google’s AlphaGo. Google created a program to play the board game ‘Go’ and it was able to defeat exceptionally skilled human players.
In a standard machine learning model, the system would have been instructed (programmed) when to make specific moves. Instead, AlphaGo learned how to play at a high level by playing against human players. It learned how and when to make moves based on experience rather than a set of pre-programmed conditions.
In practical application, machine learning and deep learning are subsets of what we would call artificial intelligence. AI is in itself a broad term for ‘anytime a computer does something smart.”
However, as continued work is done in these areas, parsing out the differences is essential. It also helps push us towards building the type of artificial intelligence we’re all looking for.
We fully believe that bringing multiple capabilities together from this category of tools is really where things get interesting. Check back soon as we dig deeper on specifics and share some of what’s surprised us along the way!