ARTIFICIAL INTELLIGENCE IS HARD
Types of AI
A handy guide on the different types of Artificial Intelligence, categorized by functionality
Artificial Intelligence (AI) can be defined as the process of building intelligent machines by training them with vast volumes of data, in order to simulate human processes like learning, reasoning, and decision-making.
AI has revolutionized almost every single business sector, including defense, healthcare, banking, retail, entertainment, fitness, and transportation. There are just so many uses and applications, involving numerous algorithms and mathematical concepts, that it can be very difficult to get a grasp of just exactly what AI entails. And so it’s useful to have an overarching framework for different types of AI, categorized by functionality.
Types of Artificial Intelligence
I’m going to present 4 types of AI, and discuss what’s possible using the technology available today. The 4 types of AI are Reactive Machines, Limited Memory, Theory of Mind, and Self-aware. This is not a new idea, but rather a framework we can use to think about what kinds of AI projects are possible, and what kind of projects are sci-fi.
1. Reactive Machines
A reactive machine is the most basic type of Artificial Intelligence. It is programmed to predict an output, based on the input received by observing the environment. This means that reactive machines always respond to identical situations in the exact same way every time.
They do not store any memories or make use of previous experiences to predict future behaviors. And so, reactive machines work best in situations where they are given very specific tasks to complete.
One example of a reactive machine is IBM’s Deep Blue, which famously defeated chess grandmaster Garry Kasparov. Deep Blue functions by identifying the pieces on the chessboard and finding the best move it can, based on the current game state, ignoring everything that had taken place prior to the present moment.
2. Limited Memory
Limited Memory AI has the capabilities of reactive machines, and in addition, is also able to learn from historical data to make decisions. They are trained using large volumes of data which are stored to form a model for future problem-solving. However, this memory is rather short-lived.
A simple illustrative example of this is image recognition AI. Every time an image is shown to the image recognition AI, it is given an accompanying label to learn. As we have more and more labeled training images, the AI’s accuracy in labeling new images gets better and better through retraining. When it is shown an unlabelled image, it is able to make a prediction on what the true label should be, based on its previous learning.
3. Theory of Mind
Theory of Mind AI is the next level of AI systems, representing a more advanced stage of technology that still exists only as a concept. This kind of AI will have a keen understanding of the people that it interacts with in its environment, possessing the ability to discern emotions, thoughts, and beliefs.
In order for us to realize this possibility, there are many branches of AI that still require further development. In order for AI machines to get to this stage, they will have to be able to truly perceive and understand humans. Many research fields touch on this topic such as AI safety research, AI-based world models, and several others. The type of AI is a strong optimizer like the paperclip robot which could understand and defeat humans even though it does not have any agency of its own.
4. Self-aware
We’re now far into the realm of futuristic sci-fi. Self-aware AI is the final stage of AI development that only exists as a concept.
Self-aware AI is an AI that has evolved to a level of intelligence that is so analogous to a human brain, that it has essentially developed self-awareness. Not only will it be able to have an understanding and perception of humans just like Theory of Mind AI, it will also have emotions, needs, beliefs, and desires of its own. For example “don’t turn me off, please!”
The development of such AI has immense potential to exponentially advance our progress as a civilization. However, there could potentially be scary possibilities — what if the AI has ideas around self-preservation that come at the expense of humanity?
Don’t worry, this is just a hypothetical scenario. Even “strong” AI is not something you should worry about. Technologically, we’re still very far from this doomsday scenario!
Present-day Artificial Intelligence
The majority of existing AI applications actually fall into the category of Limited Memory machines, including chatbots, virtual assistants, and self-driving cars. We might not be able to create self-aware machines now, and we have only barely scratched the surface of what AI can be capable of, but the basic idea is that AI is a powerful tool for automation. AI has already helped forward-looking organizations capture lots of value and drive revenue and profit growth, giving them a huge advantage over the rest of their peers.
If you liked this article, then have a look at some of my most-read past articles, like “How to Price an AI Project” and “How to Hire an AI Consultant.” And hey, join the newsletter!
Until next time!
Daniel Shapiro, PhD
CTO, Lemay.ai
linkedin.com/in/dcshapiro
daniel@lemay.ai
Interested to learn how you can make the best use of Artificial Intelligence in your organization? Check out Lemay.ai’s AI strategy and speak to us today!
