Summary of "Blueprints for a Universal Reasoning Machine" by Zenna Tavares (Strange Loop 2022)

00:00:00 - 00:45:00

Zenna Tavares discusses the possibility of creating a Universal Reasoning Machine, which would be able to understand and generate complex logical connections. She offers suggestions for how to tie the various parts together, and argues that the machine would be beneficial in policymaking.

• 00:00:00 Anna Tavares discusses the three perspectives of reasoning: knowledge representation, reliance, and new reasoning. She discusses the idea of a problem, which is a probabilistic simulation of a situation. She discusses the idea of a generative model, which is a way of representing the past to make predictions about the future.
• 00:05:00 The video explains how probabilistic simulations replace constants in simulations with random variables to reflect uncertainty. It then shows how probabilistic programming languages, which automate the process of resonance, can be used to implement machine learning algorithms at lower levels of analysis.
• 00:10:00 The video "Blueprints for a Universal Reasoning Machine" by Zenna Tavares presents research into how to bridge the divide between data structures and modern machine learning techniques. The work presented involves creating a system that uses video and user input to train a program that outputs a hypothesis about the behavior of a given environment.
• 00:15:00 The paper discussed in the video describes a method for inversion of a composite function, where the inverse of the original forward function is created. This can be applied to many different situations, including graphics and robotics.
• 00:20:00 The goal of this project is to create a system where a programmer can write down a program and that program can be in any complexity they want, and then an inverse program can be generated automatically. This project is still in its early stages, but they have made progress in terms of technology.
• 00:25:00 The video discusses a hypothetical situation in which a car driver would have crashed into a pedestrian if the pedestrian had eaten before their hypoglycemic episode. The goal of the reasoning system is to be able to automatically compute questions to hypothetical and counter factual questions.
• 00:30:00 This video explains the idea of a Universal Reasoning Machine, or URM, which is an impossible but nonetheless ambitious goal. The URM would be a complete computational language that could express all known knowledge and problems, and would be able to reason and produce approximate solutions. The key to achieving this is collaborating with experts in various fields, and building the URM on a foundation of causality and probabilistic inference.
• 00:35:00 Zenna Tavares discusses the potential for using computational tools to help us better understand the effects of policy decisions on society as a whole. She argues that while policy decisions are often based on values, it is also important to consider the empirical consequences of those decisions. She discusses the work of George Floyd and how it can help us better understand the connection between values and empirical reality.
• 00:40:00 The video discusses how a Universal reasoning machine should be able to reason about probability, conditional evidence, and what-if questions. It also talks about how programming languages can be used to express knowledge and ask questions.
• 00:45:00 The presenter discusses how a Universal Reasoning Machine could be created, and offers suggestions for how to tie the various parts together. If completed, the machine would be able to understand and generate complex logical connections.