Hey fam 🤗 Was wondering if anyone is working on something similar to the title of this email?
My take on why GA hasn't given us real world scale of evolution was that it's bound to the complexity of the simulation one confines the algorithm within. I.e. there's no entropy source, game worlds don't have sun shining down on them! Large trained models of today capture a lot more of the entropy of the world outside the computers. Here's what I'm thinking (10 minutes musings, very uncooked), add to a traditional GA the following: - Entropy source: chromosome in each generation isn't a fixed length binary number, it's variable length and evolving data structure, output of a large model - Quantum mutations: the mutation step is sourced from a quantum number generator. Many of these exist today with free APIs, e.g. https://qrng.anu.edu.au - Optional step: also evolve the world that the population is embedded within, in each generation, from a genAI With this setup, I'm trying to question a fundamental assumption of conventional GAs, that complexity can arise from a simple set of rules. Any recent related work that you know of? I know Ben is working on a similar approach, his entropy source is MeTTa agents interfacing with endpoints to bring in what I refer to as entropy here. Any other research ongoing on this? ❤️ K ------------------------------------------ Artificial General Intelligence List: AGI Permalink: https://agi.topicbox.com/groups/agi/Te9337bb274956116-M78c0a2537ffc6a044bbe68b0 Delivery options: https://agi.topicbox.com/groups/agi/subscription
