The Three Types of AI

Narrow AI

Narrow AI, which is also called limited or weak AI, is the AI that we are using today. To provide good to excellent outputs narrow AI must be restricted to specific subjects, and the user must understand that it can be very powerful in many applications. It can process very fast, vast amounts of data at a capacity that neither humans nor traditional machines have.

However, narrow AI has no cognitive capacity and no common sense as we understand it, nor the ability to judge the quality of the outputs provided. The user must keep that in mind. We will discuss this further when talking about Machine Learning, when these limitations will become clearer in regard to the consequences of wrong or mistaken outputs.

Narrow AI drives many applications that we use in our daily lives like SIRI, Alexa, Facial and Bio Recognition, Search Engines, Chatbots, and autonomous driving cars. You may be questioning how come we call SIRI a limited or narrow AI when it answers all of your questions? Well, SIRI does not think or create anything. All it does is recognise your language and question, and search in search engines like Google or Bing for the most common answer used to your question. There is no subjectivity in the answer and if you try that, the answer will not be satisfactory or beyond what the search engine would have provided to you.

Despite claims to the contrary, Narrow AI is the one that we are using today. It must be restricted to specific subjects to be efficient. It is very powerful, as it can process fast and learn from huge amounts of data that no human or traditional machine can, but again we stress that it has no cognitive capacity or common sense.

General AI

General AI (sometimes also called Artificial General Intelligence as opposed to Artificial Generative Intelligence) is still a theoretical concept and not a reality. The concept of General AI is an AI that will have all of the capacity of Narrow AI in terms of rapidly processing vast amounts of data, looking for the most common patterns, and providing a solution. A Gen AI machine will in theory understand subtle emotions, and tone of voice, and will have cognitive capacity, common sense, and the ability to judge the output quality. It is expected that the General AI will have the capacity to learn independently.

The main objective of the General AI is to perform tasks that it was not specifically trained for and to learn new skills and have autonomous self-control and self-understanding. It should be able to help in solving complex problems in scenarios and contexts that were not considered at the time of developing the AI models.

Let’s think of General AI as a computer system that has human cognitive and comprehension capacities unlike the AI that we have today that depends on training on large data sets to solve specific tasks in a specific domain. In summary, General AI will have the capacity to abstract thinking, transfer learning, common sense and think beyond the presented problem, creating new solutions and hypotheses, helping human cognitive capacity. Today General AI is a theory, but scientists believe that will be a reality in 20 to 30 years’ time, but we can’t know for sure.

Super Intelligent AI

What is Super Intelligent AI? It is when AI will surpass human intelligence in almost all the economic relevant areas, including in creativity, problem solving, and social abilities. It is when the AI will not mimic human intelligence but will surpass it. It is when the AI will create the “Replicators from the Terminator”!

Due to its potential capacities, it is only logical that when it exists, Super Intelligent AI will have a significant impact on society. Some unthinkable improvements and innovations in the fields of medicine, engineering, economy and science could become a reality and a day-to-day event, as the capacity for learning and extrapolating would be endless and fast. However, all this “power” raises important ethical and security questions and potentially the survival of the human race as we know it. If this type of AI is mismanaged the consequences could be catastrophic.

Such machines could find the cure for cancer, which has puzzled scientists for years, or generate economic models that could prevent future economic or financial crises.

While Artificial Superintelligence can lead to revolutionary discoveries and breakthroughs, it also presents significant risks. The central issue is that if a Super Intelligent AI chooses a course of action that is not aligned with human interests, it could be very difficult, if not impossible, to stop it. Also, Super Intelligent AI can increase inequality in the world, making the gap between nations and individuals even bigger, creating significant barriers to the development and improvement of individuals and nations, as the operational, energetic requirements, technology and the formation of individuals to deal with it can carry a cost impossible to certain nations. It may result in having a few nations or institutions controlling such technology. Super AI can also be used as a weapon and could for example lead to an uncontrollable escalation of conflict.

For today though, Super Intelligent AI is science fiction! However, some scientists believe that once we get General AI, Super Intelligent will not be far away. Let’s wait and see.