I was asked the question: “What is Artificial Intelligence?” 🙄
My sister was back in town for the weekend and decided to ask me about this buzzword she’s hearing more and more often. After starting explaining what was Artificial Intelligence (AI) I understood that even I did not have a clear and simple definition. Therefore, I decided to write about it in a clear and simple manner so that she and others who work in unrelated areas can make sense of what AI is and what is used for.
First thing, memorize this, Artificial Intelligence is the Father or Mother (as you prefer) of all branches you will listen in the future such as Machine Learning (ML), Deep Learning (DL), Computer Vision, Natural Language Processing (NLP), etc. This important mainly when it comes to ML, since most times people tend to confuse both terms. AI is a broad branch of Computer Science.
Like a great number of things in the world, the new and cool is in fact old… It all started back in the 1950s, when Marvin Minsky and John McCarthy, known as the fathers of the field, coined the term to refer to the heady aspiration of realizing in software and hardware an entity possessing human-level intelligence.
“Any task performed by a program or a machine that, if a human carried out the same activity, we would say the human had to apply intelligence to accomplish the task.”
The easiest way to think of AI is in the context of a Human, since we are the most intelligent creatures we know off. The goal of this field is to develop a system that can work intelligently and independently. So, I guess we can say that AI is a technique which enables machines to mimic human behavior. At the same time is that “magic” frontier of the computer doing something that amazes us and makes us think that in some way is alive!
Therefore, it makes sense that most AI systems will usually try to mimic, to some extent, one of the following behaviors associated with Human intelligence:
- Problem solving,
- Knowledge representation,
- Social intelligence and creativity.
Nevertheless, most AI systems can only perform a single task! Unfortunately, we are still a long way to assume that this single focused applications can add up to a more general intelligence or that they learn like Humans, which is not the case. If you think about it, when an algorithm gets the correct answer you find amazing how it was able to see things as we do. However, it is only when it fails in recognizing that pattern that we see how far it is from processing information the same way we do.
We Humans are very broad in our understanding of the world, our surroundings and capturing information from our senses. AI is trying to replicate this but it has still not reached the amazing capabilities which we humans took millions of years to develop.
A simple example, from the ML field, are systems which can translate the word “chair” into 37 different languages. However, the system is unable to understand what that object is used for. It is only capable of performing a single tasks.
AI is only as good as the content that goes through it.
Now, that you understand what AI is and its goals, let’s see how new AI technologies are trying to mimic human behavior.
- Speech Recognition = Humans can speak and listen, through language, to communicate with others;
- Natural Language Processing = Humans can write, read and find patterns in text on a language;
- Computer Vision = Humans can see with their eyes and process what they see;
- Robotics = Humans can understand what surrounds them and their environment, making decisions based on this
- Pattern Recognition|Machine Learning = Humans have the ability to identify patterns on what they see
- Deep Learning = If we can mimic our algorithms to work as Human brains, then we can achieve cognitive capabilities in machines.
The two groups of AI
At a very high-level Artificial Intelligence can be divided into two broad types : Narrow AI and General AI.
- Narrow AI
Narrow AI is the type you’re more familiarized with. It is what we see all around us in computers; systems which have been taught or learned to perform specific tasks without being explicitly programmed on how to execute them.
You can usually see this type of system in YouTube or Netflix’s recommendation systems, virtual assistants such as Siri or Cortana, in the vision recognition of self-driving cars, Google Translate, among others. As described previously, these systems are only capable of performing a single task which is why this area is called Narrow AI.
- General AI
This is the type of intelligence that most resembles human’s own one.
Like these known Artificial Intelligence robots, Hollywood was so kind to present us with over the years, General AI exhibits the type of adaptable intellect, a flexible form of intelligence capable of carrying out different types of tasks such as Human. From controlling the doors, temperature and route of your spaceship as well as carrying you out of that planet when a giant tsunami threatens to kill you.
This type of AI is the one which is furthest from the near future. According to a survey conducted among four groups of experts in 2012/2013 by AI researchers Vincent C Müller and philosopher Nick Bostrom, around 50% believe that General AI will be developed between 2040 and 2045, rising to 90% by 2075. The group went further on their predictions, stating that the expected “superintelligence” we usually see in movies related to robot invasion is expected only 30 years after the achievement of General AI.
Nevertheless, there are still others who believe we are centuries of distance from that day since we are very far from understanding our own brain.
If you liked it, follow me for more publications and don’t forget, please, give it an applause!
Ok ok, I know I introduced some new words that might have raised some new questions… Don’t worry I’m writing more posts for you!