Paper Summary: Online learning and generalization of parts-based image representations by non-negative sparse autoencoders

This paper presents an online autoencoder that produces non-negative sparse embedding.

Paper Summary: Universal Intelligence: A Definition of Machine Intelligence

This paper defines intelligence as an agent's ability to achieve goals in a wide range of environments. And authors present elegant mathematical formulation based on the concept of rewards and Kolmogorov complexity.

Lecture Summary: AI-Horizons Colloquium by Yoshua Bengio

This lecture summarizies the current challenge of Deep Learning and a few approaches by Yoshua Bengio.

Paper Summary: An Immersive System for Browsing and Visualizing Surveillance Video

This paper presents HouseFly, which is an immersive video analytics platform where 3 years of videos from multiple camera can be shown as immersive video.

Paper Summary: A cortical sparse distributed coding model linking mini- and macrocolumn-scale functionality

This paper presents Sparsey model which uses sparse distributed coding or representation (SDR) to build a hierarchical classifier. The main idea is using familarity to control the randomness of representation. But the simplication poses limited applicability.

Paper Summary: AGI Preschool: A Framework for Evaluating Early-Stage Human-like AGIs

The main idea is using an environment that resembles a preschool. I think it is an improvement over other tests such as Turing test, College Student test, or 8th grader test. But it still does not answer how the agent acquire communication skill using human language.

Paper Summary: Mapping the Landscape of Human-Level Artificial General Intelligence

This paper shows an overview of AGI capabilities from developmental psychology, mathematical, physiological, and information processing perspectives. Also it discusses how to evaluate AGI using environments, tasks, and scenarios.

Observational Learning: How I would build AGI

Below I summarize my plan for building artificial general intelligence (AGI). I start with the test for AGI, and then goes into environments and models for it.

Olfactory learning

Intelligence is the rules to actuate actuators according to sensory input to improve one's well-being. Learning is acquiring new rules after birth. Learning requires sensors, actuators, and memory. Olfactory learning is the first learning that can give insight to the minimum parts for learning.

My intellectual journey

I was lazy and news-addict that I wanted news story that suits my interest flows towards me. I started with recommender system but soon found that it cannot deal with the delicacy of human language. I studied visual analytics to combine human intelligence with machine learning for the text analysis. But still found it unsatisfactory. Now I am studying grounded language which lead to artificial general intelligence research.