I gave a public talk about the math of artificial intelligence (AI) at MoMath on September 19, 2024. My goal was two-fold: first of all, to show that modern AI methods are based on elementary school mathematics such as addition and multiplication. Second, I wanted to give audience an idea about “convolution”, the key ingredient in convolutional neural networks (CNNs). It is precisely CNNs that enabled AI to crack the profound problem of understanding image content automatically.
Here you can download my talk slides: first part in Powerpoint (41MB) and second part in PDF (46MB).
Just for fun, I want to include below my suggestion for the simplest AI brain that understands at least something. It can recognize summer and winter days at my workplace, Kumpula Campus in Helsinki, Finland, based on temperature data.
