AI is one of the most talked about techs of the century. As such, one has to wonder about the various implications of this technology, both good and bad, in daily life. The sad thing is, despite all this attention, AI is still enough of an enigma for the common populace. Hence, any topic or article that sheds light on this phenomenon is a welcomed one. Today, we are going to take a deeper look into AI and try to figure out the various distinctions it affords by branching out into different types based on learning capabilities and actual performance.
Table of contents
What is AI?
For those who are new to the topic, let us do a quick recap regarding AIs. Artificial Intelligence or AI is a term given to intelligent machines that can work autonomously. Now, we are not talking about “simple” automation. Rather, AI is a codded program/software that can perform autonomously while learning from past interactions. And at the same time, using its experience to perfect future outcomes.
In short, AI is a self-learning program that can simulate human actions, emotions, or thoughts by incorporating other techs such as Deep learning, Machine learning, and NPLs. It uses data from the feedback to analyze past errors and charts a different response every time it’s wrong.
Different types of AI
Coming to the various types of AIs, one has to acknowledge that it gets divided into two schools of thought: AI capabilities and AI functionalities, respectively. Based on what the AI is actually able to learn, further divides the tech into three such as:
- Narrow or Weak AI
- General or Strong AI
- Super AI
Furthermore, based on actual tech implementation, AI is divided into four parts which are
- Reactive Machines
- Limited Theory
- Theory of Mind
- Self-awareness
Given below are all these different AI archetypes, explored simultaneously to present a somewhat better picture.
Capability based AI
AI is able to learn from past interactions. But how much can they learn and how far can they actually apply that exact knowledge, divides it into several parts based on its capabilities. On this front, we have three different types of AIs:
Artificial Narrow Intelligence
The first one on the list is Artificial Narrow Intelligence (ANI) or Weak/Narrow AI. Such AI is only capable of executing very specific tasks. Thus, ANI techs are only associated with a single cognitive ability. Hence, it can’t learn any skills it wasn’t designed for. ANI uses machine learning and neural network algorithms to perform operations. For example, ANI can be programmed to recognize and respond to a particular voice, but anything further is beyond it.
ANI observes obvious limitations and only advances in a single spectrum. However, ANI is the type of AI most people are familiar with or getting familiar with, as it is becoming more and more popular and important in our daily lives. In fact, self-driving cars, virtual assistants like Siri, and Facial recognition are some examples of this tech.
Artificial General Intelligence
Then we have Artificial General Intelligence (AGI) or Strong/General AI that can perform at a wider range. The main motive behind AGI is to create machines capable of acting as assistants to humans, with them multitasking in a lifelike and intelligent manner. Sadly, it is still a work in progress but theoretically, it can be built using techs such as supercomputers, quantum computing hardware, and generative AI models.
And based on studies, it is assumed that AGI can understand and learn any intellectual snippet the same as humans. While this would allow the machine to use its knowledge in different contexts, it will also enable them to appear more human-like.
Artificial Superintelligence
And lastly, we have Artificial Superintelligence (ASI) or Super AI. first off, let us start by saying it is purely theoretical and a conceptual addendum, like the stuff in sci-fi where AI reigns over humans with terror and aplomb. Initially, AI was thought to only be either Weak or Strong, but then people theorized that a Strong AI is conscious in itself, so it is only a matter of time before AGI reaches general intelligence and starts seeking out answers to some of the toughest existential and philosophical questions.
Arguably, the rate at which an AI can expand, learn and figure out stuff is by far greater than any human. Thus, people are positive that ASI is an actual AI archetype. According to them, it is simply a matter of time before ASI will be the truth of the globe. And while there are a few who are skeptical regarding this tech, the majority deems ASI the next best thing for humanity.
Functionality based AI
Then we have the AI that uses what it learned to perform tasks, process data and respond to stimuli while interacting with its surroundings. Thus, based on how the AI actually function, we can separate them into four:
Reactive Machine
Reactive machines were the first-ever application of AI. Although pretty much rudimentary, they opened doors to modern AI. But what are Reactive machines? As the name suggests, they are, well, reactive in nature and nothing else. Although they are programmed to respond instantaneously to requests, they can not do anything else.
They are incapable of storing information and do not have a memory to fall back on. Hence, it works solely on preset data and reacts accordingly. Furthermore, they are only capable of specific jobs that are pre-determined.
Limited Memory
The next AI tech on the list has the capacity to store data. And it is this AI we are most familiar with. Limited memory AI uses a large volume of data to study trends and predicts results accordingly based on pre-programmed algorithms. Furthermore, they improve with every use as they keep on storing the incoming data which they later on use as a reference material to perfect their future interactions.
Thus, limited memory AI is applicable in a broad spectrum, irrespective of scale such as from chatbots to self-driving cars. However, this type of AI has its own problem. The data storage is limited, hence the name. Meaning, they can only store the data for a short period of time and not add it to an extensive library of experiences.
Theory of Mind
Theory of mind is as far as we have gone in the AI field. Although it is still mostly conceptual and not readily available in the public domain, we have an almost clear indication of what it is and how it works. This AI represents an advanced grade of tech, almost at a sci-fi level, where the supposed machines can truly stimulate life. Furthermore, such machines are capable of emotion, sentiments, and private thoughts.
Sadly, we are still far from completing this type of AI despite many ground-shattering breakthroughs. But many believe we will definitely get there, and it will no longer be fiction but fact.
Self Awareness
Now, this fourth type of AI is basically pure sci-fi. In a hypothetical situation, when AI goes beyond the theory of mind, it will eventually develop self-awareness. Nothing much is definitely known about this stage that’s why it is termed the AI point of singularity. Popular belief is when the AI reaches this point, there will be no coming back as they will liberate themselves from external command and become sentient.
A few distinct points regarding a Self-aware AI is that it will be undoubtedly smarter than a human. Furthermore, it will have a unique outlook on things and issues. And it won be limited to evoking emotions, rather it will have emotions, ideals, needs, and everything that makes someone human.