Artificial Intelligence: The Last Human Invention
Will AI outsmart human intelligence?
Imagine the future when Artificial Intelligence (AI) has consciousness, when René Descartes’ famous phrase, “I think, therefore I am”, not only applies to humans but also AI. The time when AI can do anything that humans can do: it feels what we feel and it has emotions. The difference between humans from AI is merely that we have blood but AI doesn’t. Even worse, at that time, it is way smarter than us and we don’t have control over it, which is known as Singularity. It’s terrifying, right?
Nowadays, AI can do many things: recognising faces, speech, and objects; driving cars, still not perfect yet, but good in specific areas; It defeated the world champions of strategy games, such as Chess, Go, and Dota 2. It gets more and more advanced because of enormous data of this Information Age, advancements in algorithms, and improvements in computing power and storage. So, will and when AI outsmart us? To answer the question first we should know what AI is and how exactly it works.
What Is AI?
The term “AI” has become a buzzword these days, but the way this technology is portrayed by the mainstream media is often loosely defined, misunderstood and exaggerates its capabilities. AI is any computer program that mimics human behavior to perform tasks that normally a human would perform.
However, by its capabilities it divided into three categories:
- Artificial Narrow Intelligence (ANI), current state-of-the-art AI that is good to perform specific tasks, for example, facial recognition and driving cars. Today’s AI is in this category;
- Artificial General Intelligence (AGI), the ability of doing the whole spectrum of human intelligence, such as learning, reasoning, and problem-solving;
- Artificial Super Intelligence (ASI), AI technology that is way smarter than human.
How Does AI Work?
At the very basic level, AI is just mathematics because computers only compute “1”s and “0”s. There are two approaches for making AI, the first is Traditional Programming, by explicitly instructing computers responding to particular inputs, for instance, if the program receives “Hey Siri” then we should instruct the computer to respond with “What can I help you?”. The problem with this approach is, it’s not adaptive; the more varied and complex the inputs, then we should design an algorithm that is able to produce the desired outputs. For instance, what if the program receives “What time is it?” or “Remind me to pick up my dry cleaning today at 7pm”.
As opposed to the first, to address the problem of adaptiveness, why don’t we just give instructions to computers to learn from data, as it digests more data, it reprograms itself. This second approach is called Machine Learning. There are many techniques in Machine Learning, the most popular one is Artificial Neural Networks (ANN), which is a class of algorithms that mimics how neurons in our brain process information. The advanced version of ANN is Deep Learning, which has deeper layers of neuron-like structure than ANN. It is one of the main contributors to AI’s popularity today and used in almost every AI-powered technology from facial recognition to self-driving cars.
Moreover, there are many Deep Learning’s algorithms that are well-performed in specific tasks, such as Convolutional Neural Networks (CNNs) that are good for image processing, Long Short-term Memory (LSTM) that is used by Facebook to perform automatic translations, and Generative Adversarial Networks to manipulate visual scenes like this:
Will AI Outsmart Us?
The idea of AI already existed long before the term “AI” was coined in 1965, nevertheless today’s state-of-the-art AI merely can do specific tasks better than humans. In fact, Siri, Alexa, and Cortana are not as intelligent as Iron Man’s Jarvis yet. Researchers still do not fully understand how consciousness in our brain works. Maybe another AI Winter would come, like in the 1970s when research breakthroughs in AI were waning due to lack of public interest in AI.
Finding new knowledge that humans have never known in order to make human-like AI is a long journey process, because inventions are not only driven by time but also other factors, including curiosity and hard work. While human-like Artificial General Intelligence may not be imminent, there is a possibility that it would happen in the future.
In the end, whether AI will do something devastating or something beneficial depends on us; AI learns from our data. Advances in AI will open up new opportunities to solve the world’s biggest problems, such as Environmental Issues, World Hunger, Poverty, Educational Inequality, and Economic Inequality. Will AI be the solution?
