Lectures on Capsule Networks, AI and Finance
In the second half of 2018, I gave several lectures in the areas of Capsule Networks, AI and Finance. I’ve not really spoken much in public since the ‘80s, when I was a keen debater and my self-aggrandizing urges got me reasonably high rankings in the World Debating Championship and an invitation to judge. The same egotistical drives are probably behind my recent activity but an extra one was to get my head around Capsule Nets; the motivation being, that to lecture on a subject requires you to work harder to understand it.
To some extent, the approach was successful in that I’ve now mastered the basics of Capsule Networks and built what I believe to be the first MNIST digit classifier in Excel (more on this later). The speaking engagements also lead to various commissioned projects such as developing neural networks to derive individual consumer credit risk scores based on mobile phone metadata. The data here was a way too big for Excel, using 300 million individual mobile connections and certainly ramped up my Python skills. Other commissions also came in media, advertising, and art—all of which employed my new found superpower in deep learning with Neural Networks.
This first video from August was at a CodeUp event sponsored by a payment platform. In this, I speak about the path to human-level intelligence and the advances that Capsule Nets offer. If you’re familiar with the subject, you see that much of my intuition is infantile, but hey, this was 4 months ago.
This second video is an interview I was asked to do on MBAs, Finance, and AI for a group who assist promising Indian MBA students.
If you make it through either of these I’d welcome any feedback you may have no matter how critical or abusive. It’s always nice to hear from outside your own echo chamber.