INTRODUCTION
Artificial Intelligence is the ability for machines to demonstrate similar (or equal) intelligence to humans. It has long been the challenge to create AI that is able to make decisions, think, act, or solve problems like a human being. Artificial Intelligence is a broad field of computer science, but there are specific categories of what AI can do, as well as the abilities it possesses. The goal of this basic material is to break down the three main natures of AI, making the structure of AI progression clearer.
The three main natures of Artificial Intelligence are: Artificial Narrow Intelligence (ANI), Artificial General Intelligence (AGI), and Artificial Super Intelligence (ASI). Qualifiers are assigned to describe the nature of AI. A qualifier like “weak” or “strong” can identify the complexity of problems the AI is meant to solve or what goal the AI is trying to achieve. The model below shows the structure of these types in increasing sophistication:
Figure Adapted From Alex Castrounis' "AI For People & Business"
ARTIFICIAL NARROW INTELLIGENCE
Artificial Narrow Intelligence (ANI) is often referred to as “weak” or “narrow” AI. This technology can carry out a single, specific, and narrow task highly well, but is weak in the fact that it cannot demonstrate cognition or understanding. ANI is not socially aware or conscious, placing it far below the capabilities of superintelligent AI. On the human intelligence scale, ANI is still less than the human brain, for it cannot perform multiple tasks or make intelligent decisions without human input.
ANI qualifiers include “shallow”, “deep”, and “applied” AI. “Shallow” and “deep” AI describe the number of hidden layers in a neural network.
Artificial Neural Network (ANN)- computer system inspired by and partially modeled on the biological neural networks (brain connections) of humans or animals. ANNs are part of the Machine Learning (ML) field and can be used to generate AI learning applications.
“Shallow” AI refers to a neural network with a single hidden layer while “deep” AI refers to a neural network with more than one hidden layer.
“Deep” AI is generally more complex in structure, as it associates with Deep Learning. “Applied” AI is the application of AI to real world problems. “Applied” AI includes “smart” hardware or software to find solutions in any industry. This qualifier of AI is still considered to be in the narrow category, but will likely advance into AGI. As of today, ANI is the only existing AI. Examples of ANI tech include:
Siri by Apple, Alexa by Amazon, Google Home Assistant (other virtual assistants)
Email Spam Filters
Facial Recognition
Speech Recognition/Translation Software
Google Search
IBM Watson
Content Recommendation Systems (Netflix, Youtube, Amazon, online shopping)
Self-Driving Cars
Drone Robots
ARTIFICIAL GENERAL INTELLIGENCE
Artificial General Intelligence (AGI) is often referred to as “strong” AI because it is able to carry out cognitive processes. On the human intelligence scale, AGI demonstrates intelligence that is equal to humans. AGI machines are able to perform any task that a human can, and are unlimited in the ability to solve multiple problems rather than a single one. AGI qualifiers include being “AI-complete” or “AI-hard”. This means that such AI problems are highly complex, for it is extremely difficult to create machine intelligence that is equal to that of a human.
If we reflect on ourselves, humans are unique and complex beings that are capable of insurmountable abilities. We possess the senses of sight, hearing, taste, sound, and touch. We have muscles that extend and flex for our movement. We can use our hands to grasp items, feet to kick a ball, or our mouths to speak.
The human brain is the most impressive wonder of human existence. It is often used as a model to try to replicate human learning for AI. The brain itself is a complex algorithm, and there are still vast amounts of information not understood about it (making it even more difficult to use it as a model for AI). What scientists do know about the human brain is that it is a neural network, processing 86 billion neurons of sensory information. It can store memories, solve problems, recognize patterns, cognate behaviour, etc. Training the human brain begins from the time we are born and ends when we die. By learning through trial and error, all of our information is continuously stored and developed.
For these reasons, it is nearly impossible to make AI that is equal to human nature. How could you ever replicate such powers in a machine? Therefore, the creation of AGI is extremely far off. However, scientific experts in the AI community have predicted AGI to emerge (on average) by the year 2060, so it is possible that we could still live to see it. To imagine more of what AGI could be like, here are some examples from media and pop culture:
Ironman’s Jarvis
Her
HAL from A Space Odyssey
AI programs from The Matrix
The Terminator
Data from Star Trek
ARTIFICIAL SUPER INTELLIGENCE
Artificial Super Intelligence (ASI) is AI that evolves to become self-improving and can even surpass human intelligence and technological advancement. This type of AI is often referred to as “Technological Singularity” or “Superintelligence”. Technological Singularity is often talked about by philosophers and futurists as the point in which technology can no longer be controlled by humans, but assumes its own domination… and possibly the domination over humans. The “Origins” host of National Geographic, Jason Silva, believes (along with many others in the tech and business industries) that such a singularity is already occurring:
In the unending debate of AI ethics, this Artificial Superintelligence has been predicted to take over humankind, possibly wiping out our biological existence. Although it is likely hundreds of years away from being developed, it is still a major fear that this technology could be our ruin.
CONCLUSION
Overall, the progression of one nature of AI into another is becoming more probable with time. Seen in the progress of new discoveries and milestones, ANI has been becoming more advanced at a faster rate. There are ethical debates of how far humankind is able to pursue AI without losing control of it (morphing into ASI), but for now, such technology is well beyond our current years.
Comentários