You may also like … – Huh?

Introduction

‘You may also like ..’ is about artificial intelligence to a non-programmer. We have seen this phrase ‘you may also like…’ from our accounts with Netflix, Amazon, Facebook, grocery stores, insurance companies and even banks! How do they know what to suggest to us, on what we may like, or potentially buy etc. .

These companies are using Artificial Intelligence or AI, to suggest what you might like. This article does not purport that I am saying they use this method. I have not worked for any of these companies and it is just my theory that they use a similarity algorithm in their AI programs.

What is AI? And how does my watching one movie suggest that I may like the others. We shall start with defining intelligence, artificial intelligence and then delve into how a similarity algorithm is used.

Intelligence?

Before we get into meaning of artificial intelligence. Let us focus on what type of intelligence we are interested in this article.

Wikipedia defines it as: “Intelligence has been defined in many ways: the capacity for logic, understanding, self-awareness, learning, emotional knowledge, reasoning, planning, creativity, critical thinking, and problem-solving. More generally, it can be described as the ability to perceive or infer information, and to retain it as knowledge to be applied towards adaptive behaviors within an environment or context.”

The key word in the definition is ‘adaptive’ as that defines that intelligent beings adapt themselves after gaining knowledge. Humans are an intelligent species and they can do a lot of stuff other animals cannot do, however other animals are as good as humans or maybe better in certain aspects. For example, crows can memorize faces, dolphins speack to each other, elephants do have incredible memory.

However, when dealing with Artificial Intelligence (AI) we will focus on this definition of intelligence to this subset: “…the following components of intelligence: learning, reasoning, problem solving, perception, and using language.” ¹

So with that in mind, let us talk about how these web sites suggest to you what you may also like..

Artificial Intelligence

Artificial intelligence is defined as “Artificial intelligence enables computers and machines to mimic the perception, learning, problem-solving, and decision-making capabilities of the human mind.“²

So simply put, AI is enabling machines to be ‘smart’ to learn and solve problems. This is accomplished by computer software programs. So in case you are not aware even computer hardware has programs to function e.g. the clock on your laptop is a program that is in the chip of your laptop.

An example of artificial intelligence in use is in the robot vacuum. The robot vacuum when it starts to do its job, it hits a wall or furniture and moves on realizing there is an obstruction in its path by maneuvering itself by a slight angle and moves again. After a few sessions of vacuuming the same room, the robot (bot) stores the room schema in its memory so that it knows where the furniture and walls are and does not bump into them.

This intelligence of the bot is by simple repetition and storing, which is learning. Now if you get a new piece of furniture, it will update its schema in a few tries realizing this is a permanent fixture.

How is ‘you may like this…‘ done

Here are two sentences

1. He likes Avengers’ movies
2. She likes Avengers’ movies

At first glance we can see that the two sentences differ by he and she and all other words are the same. In AI, the similarity score is computed and it comes as 75%! Not even close to 90%, even though both sentences differ my just one word!

(Please take my word that the computed using cosine similarity method the score is 75% and not higher. For those who want to know more about similarity score, specifically cosine similarity score, I will talk about in a future article.)

Let us now take a similar example but a quick synopsis of two movies of Robert Downey Jr, viz, Iron Man and the other Sherlock Holmes. If we were to look at the story line of the two movies on imdb.com we will get the following:

1. Detective Sherlock Holmes and his stalwart partner Watson engage in a battle of wits and brawn with a nemesis whose plot is a threat to all of England. 
2. After being held captive in an Afghan cave, billionaire engineer Tony Stark creates a unique weaponized suit of armor to fight evil.

These two story lines have a similarity score of 24% which is not good enough for us to even consider that if we watched one of them; then how can the distributor recommend the other movie. Actually, they don’t take the story line alone, they also add other parameters like:

• actors in the movie
• your profile which has your likes and dislikes on movies
• others who watched sherlock holmes and watched Iron man or avenger movies
• your reviews and others reviews similarity score
• settings of the movie
• your age group

and others parameters. When these are all concatenated they come with a different score. For instance, by concatenating a few words like ‘mystery, robert downey, adventure’ to the above texts and then computing the score I get 33%.

Before suggesting the movie ‘sherlock holmes’ since you watched ‘Iron Man’, a threshold is set say at 32% and if you get a score more than 32% then suggest the AI program suggests to watch ‘Sherlock Holmes’.

The more information I have about you, your likes, ,your tastes in movies, books, clothes and your personal data (data points) the more accurate I can predict what you may like as the similarity program gives a higher score. The companies do not suggest to you what you may like unless it meets a certain threshold. If they have more data points about you, then the AI is written in such a way to move the threshold depending on your data points it has about you. The threshold keeps moving up so that by the time they get to a similarity score of your likes as about say 70% then they will show you only those which are higher than that as you watch more movies and shows.

The programs are written to adjust and adapt depending on the number of data points etc. If you change your taste from avengers to say rom-com movies then in the beginning the AI may not show you a lot of relevant movies and after a few more data points, it will adjust and show both genre suggestions.

Summary

Artificial intelligence enables computers and machines to mimic the perception, learning, problem-solving, and decision-making capabilities of the human mind.

In this article, we define how major companies use artificial intelligence in their websites, apps to suggest what we may like or buy in future. They possibly use a similarity algorithm (the cosine similarity algorithm) to suggest your likes in their market space.

To get a similarity score, they use your data, others’ data who also are in your age group, gender, location, income level etc. and compute a score. The more data points they have about you the higher the score and more accurate their predictions.

References

1. https://www.britannica.com/technology/artificial-intelligence#ref219078
2. https://www.ibm.com/cloud/learn/what-is-artificial-intelligence
3. https://www.bbc.com/news/technology-54522442 for further reading on how AI is making us buy stuff
4. https://www.americanbazaaronline.com/2020/09/06/the-world-of-artificial-intelligence/ – Sood gives a non-technical history of AI in his article