AI vs machine learning: what are the differences?

Where does machine learning end and artificial intelligence begin? We asked Dr. Amir HajiRassouliha – UneeQ’s Senior ML developer.

May 8, 2020
AI vs machine learning: what are the differences?
Where does machine learning end and artificial intelligence begin? We asked Dr. Amir HajiRassouliha – UneeQ’s Senior Computer Vision and Machine Learning developer – to unpack the basic differences between AI and ML.

AI versus machine learning. You may be familiar with the adversarial-sounding headline. But the reality is that AI and machine learning are perhaps just as well understood through their similarities as their differences.

Both are fields in computer science. Both try to help machines mimic human intelligence and responses. Both rely on sophisticated algorithms to complete tasks. And both are changing the way we do business and interact with machines every day.

So, where do they diverge?

The differences between AI and machine learning

The large overlap between AI and machine learning is why you’ll often hear the two terms used interchangeably. It’s an understandable mistake but worth correcting.

AI, or artificial intelligence, is the overarching concept of using computers that can imitate human intelligence. Program a machine to make decisions, solve problems and perform actions based on its environment or inputs, and you’ve got AI. Standard chatbots are a good example; they’re often designed to recognize and respond to customer questions using a set of pre-defined rules. Even rudimentary chatbots could be defined as AI of a type.

Machine learning is a bit more complex. It’s a subset of AI covering machines that can learn on their own without needing explicit programming to do so. They continually improve and adapt their responses based on new data. This is the kind of behavior you’ll see in only the best AI chatbots and virtual assistants. You talk or type, they don’t just listen and respond. They have learnt, not memorized, the answer.

A way to visualize the difference between AI and machine learning is by imagining a set of Russian nesting dolls. AI would be the larger Russian doll and machine learning would be a smaller one, fitting entirely inside it. In other words, all machine learning is AI, but not all AI is machine learning.

If you wanted to carry the metaphor further, deep learning would be an even smaller doll that fits inside machine learning, as it’s considered a subset of this field. But that’s a topic for another day!

The history of AI and machine learning

The idea of smart machines is hardly a modern phenomenon. In Greek mythology, the god of blacksmithing, Hephaestus, is said to have created Talos, a giant bronze automaton who protected the island of Crete by throwing rocks at pirates and invaders.

That’s just one example. But whether it’s religion, science fiction or a clever hoax, our cultural fascination with intelligent machines far predates our scientific one, and the coining of the term ‘artificial intelligence’ altogether.

But to find the inflection point of technological progress in AI, let’s start there – the 1950s.

English mathematician and legendary war-time code breaker Alan Turing wrote his seminal ‘Computing Machinery and Intelligence’ Paper in 1950. Six years later, the Dartmouth Summer Research Project on Artificial Intelligence took place. The workshop was attended by some of the world’s brightest minds in mathematics and computer science. It’s widely considered the birth of AI as a field.

The 1950s were also when early pioneering research into machine learning began. Things gained momentum in the 1960s with the introduction of computers that could potentially update their predictions based on new information. The theory was sound, but the technology was a lifetime away.

Steady progress continued in these fields over the decades. But the 2010s is when technological advances, theoretical innovation and consumer demand reached a tipping point that allowed AI and machine learning to flourish.

How AI and machine learning is influencing chatbots and virtual assistants

We’ve already touched very briefly on how AI and machine learning is used in chatbots and virtual assistants, but let’s dig a little deeper. Currently, around 75% of all customer interactions are expected to be through chatbots by 2025, and 70% of businesses believe voice assistants are revolutionary.

It’s not hard to see why. These technologies can provide instant answers and information, they’re available 24/7 and can ease the burden on customer services teams. These are just the standard benefits, and machine learning allows AI chatbots and virtual assistants to do so much more.

For example, natural language processing (NLP) and natural language understanding (NLU), another machine learning subset, means computers can analyze, understand and generate human language (most commonly in English). Alexa, Siri, Google Assistant and various other chatbots and voice assistants are smart enough to do this.

Conversational AI goes a step further, with machines capable of having almost human-like dialogues. Alexa has learned to carry context over from one conversation to the next, much like we do when answering follow-up questions. However, machine learning is still largely in its infancy, and chatbots and virtual assistants are yet to reach the true potential of conversational AI and beyond. What happens when they do?

The benefits of AI and machine learning for digital humans

One of the main drawbacks of chatbots is that consumers have low expectations of the customer experience (CX) they can provide. Only 43% of people assume they’ll receive a good CX while using these services and less than a third find them friendly and approachable.

Meanwhile, virtual assistants are considered easier, more convenient and faster at performing simple everyday tasks than a human. But for more serious issues, such as money management, people tend to go with what they trust – other people.

Digital humans can help bridge these gaps, and AI and machine learning may play a vital role. Our own research at UneeQ shows that digital human interaction can drastically improve user experience. Yes, digital humans solve problems and answer queries like chatbots or voice assistants, but they also do it with a smile and a personality, building rapport and trust, as well as adding a more human touch to brand experiences.

UneeQ’s Cognitive Architect, Piers, has written deeply about the studies and research into why people trust AI when it has a face, if you’d like to delve into it more. Though suffice to say, the emergence of machine learning as a means to create more sophisticated, lasting connections with people only emphasizes the need to give the interface similar sophistication.

And there’s no better, more time-tested way to communicate than via the human face.