If you listen to Microsoft’s Satya Nadella, OpenAI’s Sam Altman, Google’s Sundar Pinchai or any other Tech CEOs you would probably hear the term “AI” used as a sort of mantra that starts and ends every sentence and sometimes in between. Besides these big guys, everybody else is also talking about AI nowadays, but few do know what AI really means and what types of AI are currently being developed. It’s hailed as the future of innovation, the force behind the next industrial revolution, a catalyst for unprecedented changes in various fields and a solution to solve all of the life’s problems. But is it really?
In this short article, I plan to give a simple overview what is AI, what types of AI exist and what it can and cannot do so far.
In a nutshell, what is AI?
Its a simulation of a human way of thinking or intelligence and behaviors by a machine or a computer. This is done by creating algorithms and models, training them with data and have them perform tasks that typically require human intelligence, such as recognizing speech, making decisions, solving problems, translating languages, drawing paintings, writing essays, automating processes, analyzing data and many other things. You can say that basically AI’s goal, if its not yet achieved, is to completely mimic all human behavior, knowledge and intelligence by a machine and enhance it by more precision, efficiency and scale.
What types of AI exist?
Narrow or Weak AI
its a type of AI that uses Machine Learning and similar models to learn and perform specific tasks based on historical data and clearly defined rules. Examples are: autonomous driving, image/text recognition, email spam filters, voice assistants (Siri, Alexa), etc.
Its task based and is chained to a certain limited scope, such as playing chess, driving a car, or translating text. Its called Narrow because it lacks generalization, meaning that, you cannot teach chess AI to play backgammon for example.
It is high performance compared to humans because it can process large amounts of data very quickly and provide very accurate results within a limited scope.
General or Strong AI (Artificial General Intelligence AGI)
Its a type of AI that can learn things across various tasks, just like a human being would and solve complex problems, apply creativity and understand abstract concepts that requires high intellect.
It has the ability to learn and adapt from experience and transfer knowledge from one task to another. For example if you teach AGI to learn to ride a bicycle, then it will use part of that knowledge to learn to ride a motorcycle quicker.
Furthermore it has the ability to be autonomous and all understanding, such as that it could understand the world on a deep level and make autonomous decisions without human intervention to manage all the processes entrusted to it.
Superintelligent AI or SAI
This is still a hypothetical concept where AI would surpass human intelligence in every way also adding things like emotional intelligence, wisdom and inspiration to machine’s abilities. At the moment, this is still a science fiction or fantasy type of concept that is very speculative and if ever possible.
Where are we now in the AI scale?
This is a speculative view, but if I may put my best guess estimate, we are now in the starting phase at the beginning of Narrow AI. In other words we are just starting the journey of AI possibilities. The AI that we are currently using is task based and is trained using past historical data and some experimental variables. So if something didn’t exist in the past or there is not enough good quality past data, then AI will not be able to produce a desirable result. The tools belonging to Narrow AI are ChatGPT, autonomous driving, voice assistant (Siri, Alexa), facial recognition, image generators, chess apps, language translators and others.
The Strong AI or AGI is now currently being developed in the labs of OpenAI and other pioneer research companies such as Microsoft and Google. Sam Altman claims that their new O1 model has reached AGI capabilities. However true that could be, its still a lab prototype with inconsistent user experience. The O3 model, which is the successor of O1, is the next release that has improved many flaws of O1, scoring about the same when compared to humans on visual reasoning tasks.
Super AI is still just a concept that could be seen in movies such as Matrix, where everything in daily life is controlled by an all-knowing super computer. However if that would the case, then “human” ethics would be a next big topic to be addressed in the world managed by machines.
Future Outlook
Given that we are in the beginning of Narrow AI phase, there is still a long road ahead on our AI journey. However that doesn’t mean that the long road is really long. It takes very quick time to achieve amazing results and make breakthroughs using AI. What was just simple impossible 5 years ago, is now a reality and the development is surging at a very fast pace. Sam Altman predicts that Super AI will be a reality in as soon as next decade. The fact is that most major IT companies like Microsoft, Google, Grok and others are pouring tons of focus and money on AI research – so many new features will soon be coming out. That is why the word “AI” became a mantra of today’s Tech world. So gear up and get ready for the AI age!