Enumerating the question, What is AI? Artificial is something that is simulated, not real, but also not fake in the sense of being a scam. The best way to describe artificial is by thinking of the grass used to cover the stadium ground. That is artificial grass, but it functions similarly to real grass. It is a lot easier to care for, more resistant and is widely used in many sports. The point is, artificial objects can be used instead of the genuine things because the former is better in certain situations.
Intelligence is a complex term. You can say that it is the logic, self-awareness, learning, emotional knowledge, conscience, planning, creativity, the list goes on. Humans are the most intelligent species because we have something that all other animals lack. We can perceive our environment, comprehend, understand, and take actions based on what they learned. You can say that animals can do just the same to a certain degree.
However, our brain evolved exponentially until we invented language to communicate better with others, as well as tools to interact better with our environment. We also invented weapons so we can climb up the food chain until we are at the top as the dominant species. Both humans and animals have natural intelligence. We develop this naturally through the tedious evolutionary process.
While the existence of intelligence in plants is debatable, it is proven that they show their intelligence differently from humans or animals. Plants do not have a brain or any neural network, but they are still fully capable of reacting to their environment.
A perfect example of this is a study in which a flower is grown inside a dark box with only a small hole through which sunlight could go through. As the flower grows, it leans toward the light, indicating that plants too can react to their environment although they do not have any nervous system. Intelligence observed in plants is an intriguing topic of its own because it is not as easily observed as that in humans or animals.
Now, there is another type of intelligence known as artificial intelligence. Broadly speaking, it is the simulation of human intelligence process using machines. Such intelligence is found in computer systems. AI has some characteristics similar to human intelligence such as planning, problem-solving, knowledge representation, motion, learning, and many other things. There are two main types of AI: narrow AI, and general AI.
Narrow AI is found in computers. These AIs learn and are taught to how to carry out specific tasks without being programmed explicitly how to do so.
A perfect example is speech and language recognition capability in Siri, which is a virtual assistant in iPhones. Another example is the vision recognition systems in self-driving cars or the AIs that works behind the curtain that delivers advertisement on websites that are relevant to viewers based on their search history and internet activities. These AIs are called narrow AI because they can only learn or taught how to do a specific task.
Narrow AIs can do many things such as interpreting video feeds from surveillance drones or mundane tasks such as organizing personal and business documents. They are capable of responding to customer’s questions and coordinating with other AIs to book a hotel room at the right price and location. They have also been used in advanced applications such as spotting potential cancer tumors in X-rays, detect wear and tear in elevators or flagging inappropriate contents online.
On the other hand, general AI is an entirely different entity. It has the same adaptable intelligence found in humans, unlike narrow AI that can only learn to do one thing. This flexibility allows general AIs to build spreadsheets, give you a haircut, and drive without crashing into people. This is the sort of AIs displayed in Terminator the movie. So far, though, it does not exist yet and AI experts are still debating whether it will ever be a reality.
According to a survey conducted in 2013 by AI researchers, philosopher Nick Bostrom, Vincent C Müllet, as well as four other groups of experts, reported a solid 50% chance that AGI (Artificial General Intelligence) would be developed between 2040 or 2050.
Even if that is not the case, then the chances are up to 90% by 2075. According to the survey, the emergence of superintelligence is also predicted. According to Bostrom, superintelligence is an intellect that far exceeds human cognitive performance by a long shot, in all domains of interest. Its emergence is expected to be just 30 years after the development of AGI.
READ : Artificial Intelligence Integration
Still, those are just speculations and many AI experts are still skeptical about this projection. Their main reasoning is that we still do not understand our own brain. So, it is theoretically impossible to develop anything that imitates the human brain. AGI may still be a few centuries away from us. Whatever side you take, there is one event that put the AI’s capability in the spotlight for the general public.
Back in 2017, there was a demonstration of the capability of AI. OpenAI, an Elon Musk-backed startup, created AIs that could defeat the world’s best Dota 2 players. Dota 2 is an impossibly complex game in which two teams of five players compete to destroy each other’s base.
Because there are over 100 playable characters with their own unique abilities, not to mention that there are dozens of items with unique effects of their own, it is impossible to comprehend the complexity of the game because there are literally millions of different combinations that a player can choose from. Because of this complexity and competitiveness, it is one of the most famous e-sports in the world with annual tournaments held by Valve. In that year, there was a surprise segment in which finalists have to compete against a bot from OpenAI.
According to the engineers from OpenAI, the bots learned enough about the game to beat professional players although it took time. During that time, the bot amassed a lifetime of experience by utilizing a neural network, not to mention that it runs several instances of itself in accelerated game simulation to speed up the learning process. It all began with only a few instructions. The bot was told that dying is bad, doing damage to the enemy is good, etc.
Then, the bot learns to explore the surrounding, learn the in-depth mechanics of the game and eventually become competent enough to compete against professional human players with a relatively high degree of success. Again, the bot learned enough to match the skill of world-class gamers in just two weeks whereas the human players themselves took several years just to get to where they are today. Plus, it is no simple game of tic-tac-toe either. It is a game of strategy and planning with overwhelming complexity. This is a good enough example to demonstrate the frightening capability of AI through machine learning.
See Other Tech Stories on HEADNG
Leave a Reply