ML & AI teach us a lot about the way we see things

The perception of how we see problems & situations can influence the solutions we come up with. Working with Machine Learning & Artificial Intelligence projects makes this evident with much more clarity. If the input data is noisy, you cannot expect good results / predictions / generated data. How we format that training data is a crucial step. Have we been missing this in our usual approach?

When a challenge surfaces, do we jump to finding an answer or do we try and prepare the challenge in a way that makes more sense to us? Would you prefer to see the challenge as an FFT, or as a waveform? Would MEL Spectrogram give us a better visualisation of the right features or should we just see the audio as an array of numbers?

Image credits: https://medium.com/analytics-vidhya/understanding-the-mel-spectrogram-fca2afa2ce53

We put so much effort into "choosing the right angle of perception" in ML before getting into making any further studies. ML & AI will teach us to ask the right questions – something that we always knew was crucial but underestimated it until the machines forced us to pay attention.

What are your thoughts on this? I'm eager to discuss more. Connect with me using the Contact Me page. Feel free to share the article with anyone who might be interested in such analogies.