Introduction: What is Intelligence, ML and AI?

I have started some formal training to learn more about the emerging technology, artificial intelligence (AI). It’s a helpful learning tool for me to explain what I have learned, so I figured writing it in a blog post would be great. If someone finds this helpful, that would just be an added bonus.

Over the next few posts, I will be discussing various subjects related to AI. In this post, I am going to provide an overview of the basic characteristics that define AI, explain what machine learning (ML) is and how it applies to AI, and provide some basic concepts that are used in ML.

What is Intelligence?

The characteristics of intelligence include the ability to judge, discern, and learn from past experiences. AI is the machine-based “thinking” that simulates human intelligence based on algorithms, programming and massive amounts of data. AI can be trained on data and used to generate new content or solve various problems. It’s not exactly what we might think of when we conjure thoughts of sci-fi movies of sentient machines who look like humans. It much less mystical and involves more of complex mathematics, programming, and terabytes of data.

Machine Learning (ML)

Machine learning (ML) and AI are often confused as one of the same. However, AI is actually a subset of ML. ML is a program that builds a predictive model based on input data. This predictive model can then be used to make predictions on new data of the same type of data used to build the model.

How does AI learn?

AI can use a trial and error approach to learn and make better decisions. For example, if a program is learning to spot the difference between cats and dogs, a large number of images of cats and dogs will be used as the input data, teaching the model the difference between the two. Then if the program mistake’s a picture of a cat for a dog, it can then be fine tuned to focus on specific characteristics of each animal.

Inductive and Deductive Reasoning

Inductive reasoning uses logic to arrive to a general conclusion based on specific characteristics of the input data. Inductive Reasoning would look at an image of a cat, seeing that the animal has whiskers, small ears, a small head, and a small snout to provide some confirmation based on this evidence, that the animal is a cat, and that there is a higher probability that the animal is a cat rather than a dog. If an image of a cheetah is introduced, inductive reasoning may conclude that although it is not a cat or a dog, the evidence supports the conclusion that the cheetah is closer to a cat than a dog based on its characteristics (a general conclusion).

Deductive reasoning uses logic to arrive to a specific conclusion based on general characteristics. In our example of cats and dogs, deductive reasoning would conclude that since it is not a cat, it must be a dog. Although this is a specific and certain conclusion, if the cheetah image is introduced, the prediction will fail.

Deductive reasoning can be relied on for certain conclusions while inductive reasoning is more of a best guess. Both can be valuable for problem solving, but it depends on the specific application and what type of prediction is needed.

Thanks for reading

In the next post I will discuss expert systems and AI reasoning.