The story began in science fiction: machines that can speak, machines that can think, machines that can feel. Although the last part (feeling) may be impossible without sparking widespread debate about the existence of consciousness, recent scholars have been able to make strides within the first two parts.
Over the past years, we’ve been hearing a lot about artificial intelligence, machine learning, and deep learning. However, how do we distinguish between these three rather vague terms, and how do they relate to each other? We’ll explore all that in this article.
Starting with Some Definitions
Artificial Intelligence (AI) is the general field that covers everything related to imparting “intelligence” to machines to emulate the unique human reasoning capabilities.
Machine learning is a category within the broader field of artificial intelligence that specializes in giving machines the power to “learn”. This is achieved by using algorithms that can detect patterns, generate insights from the data presented to them to apply them to future decision-making processes and predictions, a process that avoids the need to program algorithms in a way that is custom-built for each possible action on its own.
On the other hand, deep learning is a subset of machine learning: it is the most advanced branch of artificial intelligence, which brings AI closer than ever before to the goal of enabling machines to learn and think as much as a human being and even surpasses humans in some areas. In short, deep learning is a subset of machine learning, where machine learning belongs to artificial intelligence.
In the next sections, we’ll present some general historical information to better explain the differences between the three branches, and how each discovery and every development paved the way for what was achieved next.
Philosophers have attempted to understand human intelligence for millennia now, and this idea led to the emergence of the term “artificial intelligence” in 1956. Philosophy is still believed to have an important role in the development of artificial intelligence to this day. David Deutch, a physicist at Oxford University, wrote in an article about his belief that philosophy still holds the key to achieving general artificial intelligence (AGI), a level of machine intelligence comparable to that of a human brain, even though there is no brain or system on earth now that is close to knowing what our brains do to achieve any of its functions or how to replicate it.
Machine learning is nothing but an approach to achieve artificial intelligence, eliminating the need (or significantly reducing) to write code explicitly dealing with all eventualities and branching paths opting for a more holistic approach. Throughout the period from 1949 to the 1960s, American electrical engineer Arthur Samuel worked hard to develop artificial intelligence from initially being only pattern recognition to learning from experience, which made him a pioneer in this field. He used the game of checkers in his research while he was working with IBM, and this later influenced the programming of the first IBM computers.
When we dig into higher and ever more sophisticated levels of machine learning, this is where deep learning comes into play. Deep learning requires a complex architecture that mimics the neural networks of the human brain with the goal of understanding patterns, despite the noise, missing details, and other sources of interference. Although the capabilities of deep learning are very wide, its requirements are also many to get it to work successfully. You need a large amount of data and enormous mathematical capabilities. If you wanted to implement it in your own projects, you’ll likely need data science consulting.