Expert Input Importance in Artificial Intelligence

The Importance of the Subject Matter Expert in AI

We tend to think we and our computers alone could solve the problems of the “world”.

What do we think about Artificial Intelligence (AI)?

When we think about AI, we think about it as a magic box that is about to guess what needs to be accomplished and just perform at the highest level of performance, better than a human would do.

We think that all we need to do is to provide a clever supercomputer with a couple of details and automatically it will learn from it and come out with amazing insight we would not be able to come out on our own.

Information we need

When we want to learn something, we must get access to as much information as we can about the subject to search, compare, manipulate, extract insights, and analyse.

Then we would not be happy with our first conclusions, and we would do it all again. We would add another source of information, exclude another one and tune a third one. We know that the process could be improved overtime, and we would compromise to a minimal valued product (MVP) at first to have something to monetize on until we have the time and resources to further improve our process (model).

Information the machine needs

AI is not much different from the above. For a machine to learn something as we would, it needs to have access to the same information and expertise to process it. And it needs to be trained too.

In the same way, it is not enough to give books to another person and expect them to perform a task only based on that, we need to teach the machine and train it.

So why would we need AI if we must do all the effort training the machine?

In the work environment when a person needs to do a task they do not know how, it is better to give them all the necessary information, teach and train them on how to do it themselves. This way we will spend a bit more time now, but it will save us time in the future so we can focus on other tasks we need complete the most.

When a machine performs an intelligent task

One thing is to have a mathematical formula that gets inputs from some variables, is processed, and returns a result. It is another thing when the number of variables is large, and the result may vary as it relies on historical information.

If the machine can give us a response, that normally would require a person with the right knowledge to measure, weigh, think of the consequences and influence; and if this response from the machine is correct (acceptable). Then you have an intelligent response coming from a machine, you have artificial intelligence or a machine that learned (machine learning).

When Machine Learning pays off

Imagine that you have reasonable responses for a machine that has been training with data and input from the experts that perform better analysing your data?

Imagine that the same consistent analysis could be performed by different employees or even automatically as part of a process.

When you get to these scenarios, it is when AI and Machine Learning can be scaled in a way that would be difficult to achieve in other conventional ways.

Conclusion

Does not matter how much we want to take ownership of an AI model and deliver. We think we have the data; we know what needs to be done, and we have the computer resources. However, at the end of the day, to have AI trained models that perform at their best we still need to have those specialists give their input to create or improve an already existing algorithm.

Showcase

To highlight the importance of an expert involvement in the process I have created an ML (Machine Learning) model which I document in a YouTube video demonstrating the use of AI/ML to process real data from a Football Fantasy website and how to process its data to predict player scoring round by round for the Brazilian Premier League 2021 using Azure Automated Machine Learning.

Then I compare the results with a real friend’s competition to check if the model performed well.

Additional details are available in the Article “Azure Artificial Intelligence and Fantasy Football Prediction”