Can AI do anything? If not, what are the things AI can do today? (Part 1)

Reading Time: 3 minutes

Can AI do everything? If not, what are all the things AI can do today?

Recently, a train accident occurred, and I noticed the following LinkedIn post floating around

Well, I completely understand the angst and emotions behind this post. However, I would like to clarify things about the capabilities of AI in this post.

About six months ago, I was a non-technical person who knew little about AI. It seemed like some kind of super technology that could do things no normal human could.

And then to stay on top of the latest trends, I decided to learn AI. 

With the advent of ChatGPT, we all have heard of AI. There are also many videos showing what AI is capable of – from writing codes to generating videos to producing images of people who don’t exist.

All of these things give us the impression that AI can do anything.

But is it the truth? Let’s find out.

Then, how does AI work?

AI is model based.

For doing any activity, we need an AI model. For ex – there is an ai model to convert speech to text, there is an ai model behind ChatGPT, there is an ai model to generate an image from text, so and so forth.

So, if there is no AI model for a particular activity, we can’t do that activity with the help of AI.

Now, how AI models are built?

AI models are built by training them on existing data. 

An AI model finds some patterns in old existing data and then it applies that learning to the new data.

This thing is the same for all AI models – be it a traditional model for predicting the price of a house or the state-of-the-art ChatGPT.

In both cases, we need thousands of data (if not millions) for an AI model to learn some patterns.

So, if there is no previous data then there is no AI model. It’s as simple as that.

Drawing the analogy of train accident with House example:

AI can predict the price of a house provided we have 1000s of existing house sales data.

But AI can’t predict whether a house will collapse suddenly. In order to predict that, AI again needs 1000s of house collapsing data.

Let’s say, you get this data somehow.

However, AI still will not be able to predict a fire incident that happened in that house because of human negligence. 

The reason again is simple, we need enough data on human negligence for models to learn and predict 

Coming back to the train example.

AI can’t predict train accidents due to human or technical errors. In order to predict this, we need to have thousands of such data. 

To cut a long story short – It is difficult to build an AI model that can predict accidents due to human negligence on a real time basis.

So, instead of focusing on what AI can’t do it’s easier to understand what exactly AI can do. Let’s do that in Part 2 of this article

Thanks for choosing to read it completely. I can try to simplify any AI concept for you.

In case you want me to simplify any AI topic, please let me know on LinkedIn – https://www.linkedin.com/in/nikhileshtayal/

Let’s learn to build a basic AI/ML model in 4 minutes (Part 1)

Master AI/ML from scratch with these free courses. You need to opt for “Audit the course” in order to get it for free.