Danko Nikolic says: "I see the no free lunch theorem striking every day. Every time we pick one ML architecture for one type of problem and another architecture for another type of problem, it is the No Free Lunch Theorem dictating the fact that we have to make thos chices and are not able to have one the same architecture for all kinds of problems."
Matt Mahoney says: "There is no simple, universal prediction algorithm. Suppose you have one. Then I can create a simple sequence that you can't predict. My program runs a copy of your program and outputs the opposite of your prediction." "The best compressors have lots of code to handle lots of rare, special cases. It's not because of the no free lunch theorem. It's because you can always find something that your program can't compress, and you have to add yet another special case." Jim Bromer says: "If there was no such thing as a free lunch all progress would have been impossible. By falling into your presumptions you lock yourself into them." ---------------------------------------------------------------------------------------- My response: Progress, advances, and evolution are actually Survival/ Immortality. Lots of stuff happens in our universe, but each changes and each is lost, and each is meaningless or just as good ex. a rock or hydrogen atom or kid or yellow or chair are all equally acceptable. To resist change is to repair and clone oneself, resisting death. Our physics has only one way it works. Yeah there's a lot of diversity, but resisting "change" is a pretty fundamental thing, unlike any other thing! At least, it is here if not. Yeah that is free lunch, the change occurs on Earth and well, less change occurs now cuz we better battle death and loss of information. Matt is right about that, nice idea, but obviously we are able to make "advances", we are getting better at predicting our world, which means we can better survive longer, the physics we model doesn't say "hey, let me compare your machine and give you opposite predictions/physics!", so Matt is wrong there, physics doesn't change its laws/predictableness, we can learn how to predict those laws. Danko is wrong too, there is general purpose patterns that solve many problems/patterns fast, our AI algorithms already do that, so do tools like TVs, vacuums, etc. And yes, there is still many rare cases, which add up to quite a lot of unexpected problems cuz there's many of them even though each is rare, and these need their own code, but no one said they all don't share a pattern, they actually all are a similar problem still, and I will show just that soon. Take for example two patterns below, they deal with recognizing the mentioning of a child and parent and last name given, or recognizing that one word is similar to another word - a recognition of a recognition: "Xio Yong has a son named Jex _" ? There is many problems for AI to answer, and many of them are rare and never before seen. This pattern above is the Last Names pattern, it's pretty rare. Another: "bird moon eagle, book rock guide, house wind home, cave broom hole, football bathroom _" ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Ta433301e9ac5fb42-M602b75ae09c7a807b6f5c163 Delivery options: https://agi.topicbox.com/groups/agi/subscription
