Levels of AI Explained

 

Narrow AI

Narrow AI basically means the artificial intelligence systems that are specified to handle a singular or limited number of tasks. The Narrow AI is also referred to as Weak AI. This kind of AI pulls the data from a specific-dataset so because of limited data fed it can only perform single task. They are not allowed to perform multiple tasks as the data is limited.

Every machines intelligence that surrounds us are the result of narrow AI. For example:- Google Assistant, Apple’s Siri and other natural language processing language tools that are used in this decade. If we talk about common man perspective on this, then these technologies are not considered weak because of the their human to machine interaction. But, if we talk about it, these are not technologies we think about because they are not nearly as intelligent as human-mind.

When we speak to these devices, there is a lack of consciousness, self-awareness and it is a machine which enters your question in search engine(Google) and returns the answer for your question. We cannot converse with the device it is more like a question and answer situation.

This why we lack in conversing and asking about ourselves and not getting any genuine response. It will just give you answer on the basis of what you have written about yourself on websites or articles not a response that is original given by itself. From this we can interpret that it is not capable doing the real conversation that two humans normally do with each other.

This explains why we get vague answers that does not make sense or we get the links to existing articles when we ask some questions like meaning of life or approach the personal life question to Siri or Google assistant.

As a human, when we converse to other person we know the surroundings, sentiment and emotional stage of the person we are talking to. But if we talk about technology there is no flexibility or fluidity or emotion in any condition to think like humans.

So, from this we can conclude that these are the innovation of narrow AI who can do limited tasks through limited data.

General AI

General AI, also referred to as ‘strong AI’, is the concept of a machine with general intelligence that mimics human intelligence and behaviour with the ability to learn and apply its intelligence to solve any problem. It can think, understand and act in a way that is indistinguishable from that of human in any given situation.

But defining General AI is very difficult because ‘General AI’ already implies that it’s very broad term, and even if we consider human intelligence as a base line, not all humans are equally intelligent.

The subsets of Machine learning are good enough for facial and object recognition, recommendation systems, chatbots etc. But does deep learning solve the general AI? Certainly not. Deep learning and neural networks have so far proven to be good at spatial and temporal consistency in data. But they are really very poor at generalizing capability  and reasoning about the things in the world like human do.

A well-trained neural network can detect the cricket bat, ball and player in the video given above. But we cannot expect the neural network to determine the behaviour and relation of the different objects in the video. It will also break under the novel situation such as viewing an object from the different angle as given below.

 

 Super AI 

 For Super AI, I can only say that it is a type of AI which surpasses human intelligence in every aspect of life. It is also called Superintelligence. It best and better at everything like maths, science, medicine, hobbies etc. you name it. Even   the best and brightest human minds cannot defeat Super AI. 

There are no limits to Super AI that what it could do. It is impossible to even  imagine about what world it will be where AI is more intelligent and bigger deal than human mind.

It could be a threat or it could be the most beautiful gift to the human-kind from technology.