Posts

2 million $ to the person you love or 1$ anonymously to 200 million poor?

Here is an ethics thought experiment that explores personal connection in altruism, similar to Peter Singer's thought experiments. Suppose you were given these two choices Choice A: You can give any single person you know (other than yourself) 2 million $ openly. Choice B: You can give 200 million of the poorest people on the earth 1 $ each, anonymously.  If one were a sensing and feeling type of person, then the choice would be obviously A. You could give your loved one 2 million dollars and that would be enough to make them financially independent for the rest of their life (assuming they use the money judiciously). You would be interacting with them regularly, you know them well for a long time and will continue seeing them be happy for the rest of your life. This is similar to the ending in The Last of Us. If one were logical, intuitive and truly selfless then choice B is clearly the better choice. Poor people across the globe would benefit many orders of magnitude more than in

Autofunctions

Autofunctions: Exploring Recursive Utility in Everyday Objects Introduction In a world teeming with multifaceted objects and tools, a curious phenomenon often goes unnoticed: the ability of some objects to perform their primary function on others of their kind. I've coined a term for this fascinating concept: "Autofunctions." This blog post delves into the realm of autofunctions, exploring their presence in our daily lives and their implications in various fields. Defining Autofunctions An autofunction is when an object or tool applies its primary function to another of its kind. Think of a bag carrying other bags, a crane lifting another crane, or a software debugging tool debugging itself. This phenomenon transcends simple utility, reflecting a self-referential or recursive nature in the design and function of these objects. Everyday Examples To truly grasp the concept of autofunctions, let's look at some commonplace examples: Containers Holding Containers: This inc

Informative YouTube channels

  YouTube (YT) can be a double edged sword for our long term well being. One the one hand it contains many informative, useful videos that improve our understanding of the world, on the other hand one could easily be sucked into mindlessly watching useless videos. Understanding how YT works will help us use it better. One thing we have to be wary about is YTs recommender system. This can hook us on and make us crave for more videos . I would recommend extensions like Distraction Free YT to avoid seeing YT recommendations next to current video. Instead of recommendations we can choose based on channels. Channel list Below I have shared a list of channels/playlist which I believe are more substance than style.   General 3Blue1Brown - YouTube Maths visualizations. Asianometry Semiconductors, history, general. Crash Course World History This channel has multiple courses, but the original history one was really interesting. Daniel Patton - YouTube Book reviews. DW Documentary - YouTube

Media franchise lifecycle in a capitalistic world

One pattern I see in the western world is that if any media franchise/IP becomes popular, then it is milked till people stop liking it. e.g. Star wars, Game of Thrones TV series, Simpsons, Spongebob etc. Big corporations are hesitant to create new IPs, rather they try to play it safe by continuing with existing franchises or formulae. However if a new IP is created and it does become popular then it is extremely valuable as now the studios can start milking it till it dies. Thus a pattern we see is that every franchise will always end in a flop. Because in a capitalistic world otherwise it would be like leaving money on the table.  Only in cases where the franchise is owned by non-greedy entity who appreciates the art more than the money does the franchise not end in a flop. It takes generosity and respect for the creation for the owner to say "I know I can make some more money if I make a new movie from this, but I don't have any novel ideas, so I'm not going to make it.&

Social Networks, Truth and Trust

Originally posted in https://ideafair.bearblog.dev/social-network-and-trust/ One of the main issues with the current social networks is the issue of favoring appeal over truth. Posts that appear good get likes, regardless of whether they are actually true or not. Thus they are susceptible to people's biases. This leads to the problem of false news travelling faster than true news. Our current recommender systems promote posts that get a higher number of likes from people and thus this gets amplified. Instead we need to change our recommender systems to value truth over likings. Perhaps we can associate each user with a truth score based on how true their past predictions were and then promote posts from truthful people. This is similar to the ideas in the book "Signal and Noise" by Nate Silver. We can have a system where people can make testable predictions, something like https://longbets.org/ and develop a truth score for each user. Then the social network should promot

Why we enjoy music

Originally published in https://ideafair.bearblog.dev/why-we-enjoy-music/ Why we enjoy music 10 Jan, 2021 M usic and ML I think the reason why we enjoy music can be explained in terms of Machine Learning (ML). Music can be seen as an audio signal that has multiple levels of patterns in it. When we listen to music our brain is trying to learn and predict this pattern. We enjoy music when the brain is finally successful at this task. There must be some innate reward function for being able to learn and predict patterns successfully in our brain. Confidence and Accuracy To go bit deeper into when exactly we enjoy music the most, we can look at our brain's confidence & accuracy of the music prediction. At first both the confidence and accuracy are low, but after listening to the tune several times our accuracy increases. At this point we are able to predict with higher accuracy, but the confidence is still low. This is my guess when the enjoyment is highest as we are building confi

ML system design interview

 This was originally posted on https://ideafair.bearblog.dev/ml-system-design/ on 1 Jan 2021 ML System design interview 01 Jan, 2021 This post details how a typical Machine Learning (ML) system design interview is conducted. It is quite similar to a software engineering system design interview, but there are many differences which I will list here. The candidate is given a problem and asked how they would go about designing the solution for the problem. Some example problems are: plagiarism detector for a class, YouTube video recommender, grammar correction service etc. The interviewer then expects the candidate to lead most of the discussion, occasionally bringing in what-ifs, asking clarifying questions, digging down on the details etc. My approach for solving ML design is to divide it to following sections. Sections of interview Problem clarification It is good to ask as many questions as possible, state out all of your assumptions clearly. Not doing any one of the above and jumping