Google Tech Talk
July 30, 2009

ABSTRACT

Google Faculty Summit 2009: Global Connection: Inform and Empower

Statistical Machine Translation, presented by Franz Och.

Each year Google hosts leading academics from universities across the globe though our Faculty Summit program. Faculty Summits are designed to provide researchers with a chance to learn more about what Google does in each region and how we support university programs, as well as provide valuable networking time for academics and engineers.

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

OpenCog: An Open Source Software Framework & A Design and Vision for Advanced AGI
Presented by Ben Goertzel, co-founder of the OpenCog project.

More on OpenCog: http://opencog.org/about/

Speaker Bio: http://wp.goertzel.org/?page_id=58

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Probabilistic Programs: A New Language for AI
Presented by Noah Goodman, Stanford University

ABSTRACT

How can logical and probabilistic approaches to understanding
intelligence be reconciled? I will argue that probabilistic
programming is the best way to merge logic and probability, providing
a new set of tools for thinking about representation and inference in
systems with human-like intelligence. I will illustrate these ideas by
introducing the probabilistic programming language Church (a
stochastic LISP), describing two universal inference algorithms (i.e.
algorithms that can perform probabilistic inference for any Church
program), and giving a series of examples. These examples, drawn from
cognitive science and AI, will include multi-agent reasoning and concept learning.

About Noah Goodman: http://stanford.edu/~ngoodman/

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Self-Programming Workshop:

Self-Programming = Learning about Intelligence-Critical System Features.
Presented by Ben Goertzel.
http://www.iiim.is/wp/wp-content/uploads/2011/05/goertzel-agisp-2011.pdf

An Implemented Architecture for Feature Creation and General Reinforcement Learning.
Presented by Brandon Rohrer.
http://www.iiim.is/wp/wp-content/uploads/2011/05/rohrer-agisp-2011.pdf

Behavioral Self-Programming by Reasoning.
Presented by Pei Wang.
http://www.iiim.is/wp/wp-content/uploads/2011/05/wang-agisp-2011.pdf

Heuristic Search in Program Space for the AGINAO Cognitive Architecture.
Presented by Wojciech Skaba.
http://www.iiim.is/wp/wp-content/uploads/2011/05/skaba-agisp-2011.pdf

Emergent inference, or how can a program become a self-programming AGI system?
Presented by Sergio Pissanetzky.
http://www.iiim.is/wp/wp-content/uploads/2011/05/pissanetzky-agisp-2011.pdf

Self-Programming through Imitation.
Presented by J. Storrs Hall.
http://www.iiim.is/wp/wp-content/uploads/2011/05/hall-agisp-2011.pdf

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

The Future of AGI Workshop Part 1 - Ethics of Advanced AGI

Workshop intro: Ben Goertzel

Steve Omohundro, Design Principles for a Safe and Beneficial AGI Infrastructure
http://agi-conf.org/2011/abstract-stephen-omohundro

Anna Salamon, Can Whole Brain Emulation help us build safe AGI?
http://agi-conf.org/2011/anna-salamon-abstract

Anna Salamon, Risk-averse preferences as AGI safety technique
http://agi-conf.org/2011/carl-shulman-abstract

Mark Waser, Rational Universal Benevolence: Simpler, Safer, and Wiser than "Friendly AI"
http://becominggaia.wordpress.com/papers/rational-universal-benevolence-simpler-safer-and-wiser-than-%E2%80%9Cfriendly-ai%E2%80%9D-agi-11/

Itamar Arel, Reward Driven Learning and the Risk of an Adversarial Artificial General Intelligence
http://agi-conf.org/2011/abstract-itamar-arel/

Ahmed Abdel-Fattah & Kai-Uwe Kuehnberger, Remarks on the Feasibility and the Ethical Challenges of a Next Milestone in AGI
http://agi-conf.org/2011/abstract-ahmed-abdel-fattah-and-kai-uwe-kuehnberger/

Matt Chapman, Maximizing The Power of Open-Source for AGI
http://agi-conf.org/2011/matt-chapman-abstract/

Ben Goertzel and Joel Pitt, Nine Ways to Bias Open-Source AGI Toward Friendliness
http://agi-conf.org/2011/wp-content/uploads/2009/06/NineWaysToBias.pdf

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

AGI and Neuroscience Session

Randal A. Koene: AGI and Neuroscience: Open Sourcing the Brain
http://rak.minduploading.org/publications/publications/koene.AGI-2011.position-statement.pdf

Serge Thill: Considerations for a neuroscience-inspired approach to the design of artificial intelligent systems
http://agi-conf.org/2011/wp-content/uploads/2009/06/CowardNeuroscienceTrackAGI2011.pdf

Andrew Coward: Brain anatomy and artificial intelligence
http://agi-conf.org/2011/wp-content/uploads/2009/06/CowardNeuroscienceTrackAGI2011.pdf

Janelle Szary, Bryan Kerster and Christopher Kello: What Makes a Brain Smart? Reservoir Computing as an Approach for General Intelligence
http://cogmech.ucmerced.edu/~jszary/SzaryKersterKello_AGI.pdf

Murray Shanahan: Artificial General Intelligence Requires Consciousness

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Architectures Session, Part I

Javier Snaider, Ryan Mccall and Stan Franklin: The LIDA Framework as a General Tool for AGI
http://ccrg.cs.memphis.edu/assets/papers/2011/The%20LIDA%20Framework%20as%20a%20General%20Tool%20for%20AGI%20final.pdf

Paul Rosenbloom. From Memory to Problem Solving: Mechanism Reuse in a Graphical Cognitive Architecture
http://agi-conf.org/2011/wp-content/uploads/2009/06/AGI-2011.pptx

Joscha Bach. A Motivational System for Cognitive AI
http://micropsi.com/publications/Bach_AGI_2011.pdf

Zhenhua Cai, Ben Goertzel and Nil Geisweiller. OpenPsi: Realizing Dorner's "Psi" Cognitive Model in the OpenCog Integrative AGI Architecture
http://goertzel.org/OpenPsi_agi_11.pdf

Matthew Ikle and Ben Goertzel. Nonlinear-Dynamical Attention Allocation via Information Geometry
http://faculty.adams.edu/~moikle/Site/Research_files/ECAN%20with%20Info%20Geom_agi-11_v3.pdf

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Friday Morning Keynote Sessions

Aaron Sloman: The biological bases of mathematical competences: a challenge for AGI
http://www.cs.bham.ac.uk/research/projects/cogaff/misc/agi-2011-talk-sloman.html

Ed Boyden: Technologies for Understanding How Brain Circuits Perform Computations

Read more about this video…

The Fourth Conference on
Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

The Future of AGI Workshop, Part II

* Florin Popescu, What is Dexterity?
http://agi-conf.org/2011/florin-popescu-abstract/

* Alessandro Oltramari, Visual Intelligence Beyond Representation
http://agi-conf.org/2011/abstract-alessandro-oltramari/

* J. Storrs Hall, The Economics of Singularity
http://agi-conf.org/2011/wp-content/uploads/2009/06/econ.pdf

* Bill Hibbard, The End of Rough Equality of Intelligence
http://agi-conf.org/2011/bill-hibbard-abstract/

* John Smart, The Brain Preservation Prize: The Mouse Connectome as a Path to Natural and Artificial Intelligence

Read more about this video…

The Fourth Conference on
Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Session 5

Samuel Epstein and Margrit Betke. An Information Theoretic Representation of Agent Dynamics as Set Intersections

Sergio Pissanetzky. Structural Emergence in Partially Ordered Sets is the Key to Intelligence

Florin Popescu. Wagging the dog: human vs. machine inference of causality in visual sequences.

Helmar Gust, Ulf Krumnack, Maricarmen Martinez, Ahmed Abdel-Fattah, Martin Schmidt and Kai-Uwe Kuenburger. Rationality and General Intelligence

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Session 3, Architectures Part II

Haris Dindo, Antonio Chella, Giuseppe La Tona, Monica Vitali, Eric Nivel and Kristinn Thórisson. Learning Problem Solving Skills from Demonstration: An Architectural Approach
http://roboticslab.dinfo.unipa.it/papers/agi2011-dindo.pdf

Alessandro Oltramari and Christian Lebiere. Extending Cognitive Architectures with Semantic Resources
http://www.psy.cmu.edu/people/Oltramari-Lebiere-2011.pdf

Unmesh Kurup, Christian Lebiere and Anthony Stentz. Integrating Perception and Action for AGI
http://www.ukurup.com/media/files/AGI-11-Preprint.pdf

Benjamin Johnston.The Collection of Physical Knowledge and its Application in Intelligent Systems
http://www.benjaminjohnston.com.au/papers/collecting.pdf

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Session 4

Tobias Glasmachers and Jürgen Schmidhuber. Optimal Direct Policy Search

Yi Sun, Faustino Gomez and Juergen Schmidhuber. Planning to Be Surprised: Optimal Bayesian Exploration in Dynamic Environments

Tom Schaul, Leo Pape, Tobias Glasmachers, Vincent Graziano and Juergen Schmidhuber. Coherence Progress: A Measure of Interestingness Based on Fixed Compressors

Juergen Schmidhuber, Dan Ciresan, Ueli Meier, Jonathan Masci and Alex Graves. On Fast Deep Nets for AGI Vision

Linus Gisslen, Matt Luciw, Vincent Graziano and Jürgen Schmidhuber. Sequential Constant Size Compressors and Reinforcement Learning http://www.idsia.ch/~gisslen/gisslen_agi2011.pdf

Derek Monner and James Reggia. Systematically Grounding Language through Vision in a Deep, Recurrent Neural Network http://www.cs.umd.edu/~dmonner/papers/agi11.pdf

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Saturday Keynote #4

Presented by Zhongzhi Shi
AGI Research Progress in Intelligent Sciences Lab at Chinese Academy of Sciences

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Thursday Evening Keynote
Dynamic Vision as a Key Element for AGI
Presented by Ernst D. Dickmanns

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Thursday Evening Session
Opening Remarks by Moshe Looks and Peter Norvig

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Thursday Evening Session
Mapping the Landscape of AGI - Ideas and Conclusions from the 2009 AGI Roadmap Workshop
Presented by Ben Goertzel

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Saturday Session 2

Laurent Orseau and Mark Ring. Self-Modification and Mortality in Artificial Agents
http://www.agroparistech.fr/mmip/maths/laurent_orseau/papers/orseau-ring-AGI-2011-mortal.pdf

Mark Ring and Laurent Orseau. Delusion, Survival, and Intelligent Agents
http://www.agroparistech.fr/mmip/maths/laurent_orseau/papers/ring-orseau-AGI-2011-delusion.pdf

David L. Dowe, Jose Hernandez-Orallo and Paramjit K. Das. Compression and intelligence: social environments and communication
http://users.dsic.upv.es/proy/anynt/paper5-compression.pdf

Jose Hernandez-Orallo, David L. Dowe, Sergio España-Cubillo, M.Victoria Hernandez-Lloreda and Javier Insa-Cabrera. On more realistic environment distributions for defining, evaluating and developing intelligence
http://users.dsic.upv.es/proy/anynt/paper3-environment-complexity.pdf

Javier Insa-Cabrera, David L. Dowe, Sergio España-Cubillo, M.Victoria Hernandez-Lloreda and Jose Hernandez-Orallo. Comparing humans and AI agents
http://users.dsic.upv.es/proy/anynt/paper1-comparing.pdf

Read more about this video…

The Fourth Conference on Artificial General Intelligence
Mountain View, California, USA
August 3-6, 2011

Self-Programming Workshop II
Panel Q&A Session

Read more about this video…

Google Tech Talk
February 14, 2012

Presented by Ron Garret.

ABSTRACT

The Remote Agent Experiment: Debugging Code from 60 Million Miles Away

The Remote Agent Experiment (RAX) was an autonomous control system for an unmanned interplanetary spacecraft called New Millennium Deep Space 1 (DS1). In May, 1999, control of the DS1 spacecraft, a $150-million asset, was handed over to the Remote Agent software for three days. It was the first -- and, to date, the last -- time that an interplanetary spacecraft has been under fully autonomous control. RAX was a resounding technological success, but a political disaster. Instead of paving the way for future autonomous missions, RAX is the reason that NASA has not flown an autonomous mission since. This talk is about the lessons learned from an ambitious but ultimately failed attempt to introduce technological change into a large, bureaucratic organization, the limitations of static code analysis, and the unique challenges of debugging code when the round-trip ping time is 45 minutes.

Slides available at http://www.flownet.com/ron/RAX2.pdf

Dr. Ron Garret is a software engineer turned entrepreneur and angel investor. He has co-founded three startups and invested in a dozen others. In a previous life he was an AI and robotics researcher at the NASA Jet Propulsion Lab where he led the development of one of the four major components of the Remote Agent. In 2000 he went to work for what was at the time an obscure little Silicon Valley startup called Google, where he was the lead engineer on the first release of AdWords, and the author of the Google Translation Console. He is currently working on launching a new startup company.

Read more about this video…

Google Tech Talks
July 30, 2009

Read more about this video…

Abstract
Take advantage of user data to create intelligent applications. This talk will focus on data mining to create complex application behavior and gain insight into the patterns and habits of your users. Examples of these techniques can be seen with recommendation systems like those created by Amazon, Netflix, last.fm, and others. Additional examples include spam filtering systems for email or comment filtering provided by Akismet.

We will focus on techniques for gathering data, specific gems and plugins for performing various data mining and machine learning tasks, and performance issues like how to distribute the work to separate servers. Theory in this talk will be light and the specific algorithms will only get a mention by name. We’ll be looking at real world Ruby and Rails code examples for building recommendation, ranking, and classification systems.

Bio
Paul Dix has been working with Ruby and Rails since 2005 when he first started attending meetings at NYC.rb. Paul was a speaker at GoRuCo 2007 where he presented on the topic of document classification. He is also the author of the Basset Ruby Gem, which provides an API for various machine learning tasks. Paul has worked in companies large and small as a consultant, developer, network engineer, and software tester. The big names include Google, Microsoft, McAfee, and Air Force Space Command. He currently works for Mint Digital where he contributes to various Rails applications for clients and Mint’s Rails application platform. In addition to his consulting, Paul is working on a new startup that makes use of collective intelligence techniques. He is also currently a student at Columbia University where he studies topics in machine learning, natural language processing, information retrieval, and search.

Read more about this video…

Abstract
Archaeopteryx generates MIDI bytecodes which produce music via prosumer applications such as Propellerhead Reason. If you’ve ever wondered whether or not use cases really exist for Ruby features like lambda, eval, continuations, and higher-order functions, Archaeopteryx demonstrates not just the usefulness but the power and flexibility of these techniques. Archaeopteryx also uses statistical AI methods - specifically, probability matrices, similar to those used in naive Bayesian classifiers – to define templates for possible grooves.

An experiment in progress, Archaeopteryx currently generates techno drum rhythms and New-Agey ambient soundscapes with ease. Developing additional features by the time of GoRuCo is very, very likely. Following the pattern of all successful technology projects, Archaeopteryx targets low-hanging fruit with superior technology, and scales up from there. This presentation will feature no slides at all, and focus instead on demonstrations, explaining existing code, and writing new code live.

Bio
Giles Bowkett’s programming blog is well-read in the Ruby community and occasionally gets 10,000-visit days. He has worked with Perl, PHP, Python, Java, JavaScript, ActionScript – all the usual suspects – and now spends nearly all his time coding Ruby. He’s worked on the streaming video site Hulu and contributed a patch to Rubinius, he has a recipe in the upcoming “Advanced Rails Recipes” book, he’s speaking at five conferences in three countries in 2008, he regularly checks in code before 8am on Saturdays, he kind of needs to get a life, and between December and February 2007-2008 he released 9 open source projects (with more coming). Giles studies acting with the same feverish intensity he applies to programming. He also practices meditation, which, when you consider everything else, is probably a good idea.

Read more about this video…

Google Tech Talks
November, 29 2007

In the 1980's, new learning algorithms for neural networks promised to solve difficult classification tasks, like speech or object recognition, by learning many layers of non-linear features. The results were disappointing for two reasons: There was never enough labeled data to learn millions of complicated features and the learning was much too slow in deep neural networks with many layers of features. These problems can now be overcome by learning one layer of features at a time and by changing the goal of learning. Instead of trying to predict the labels, the learning algorithm tries to create a generative model that produces data which looks just like the unlabeled training data. These new neural networks outperform other machine learning methods when labeled data is scarce but unlabeled data is plentiful. An application to very fast
document retrieval will be described.

Speaker: Geoffrey Hinton
Geoffrey Hinton received his BA in experimental psychology from Cambridge in 1970 and his PhD in Artificial Intelligence from Edinburgh in 1978. He did postdoctoral work at Sussex University and the University of California San Diego and spent five years as a faculty member in the Computer Science department at Carnegie-Mellon University. He then became a fellow of the Canadian Institute for Advanced Research and moved to the Department of Computer Science at the University of Toronto. He spent three years from 1998 until 2001 setting up the Gatsby Computational Neuroscience Unit at University College London and then returned to the University of Toronto where he is a
University Professor. He holds a Canada Research Chair in Machine Learning. He is the director of the program on "Neural Computation and Adaptive Perception" which is funded by the Canadian Institute for Advanced Research.

Geoffrey Hinton is a fellow of the Royal Society, the Royal Society of Canada, and the Association for the Advancement of Artificial Intelligence. He is an honorary foreign member of the American Academy of Arts and Sciences, and a former president of the Cognitive Science Society. He received an honorary doctorate from the University of Edinburgh in 2001. He was awarded the first David E. Rumelhart prize (2001), the IJCAI award for research excellence (2005), the IEEE Neural Network Pioneer award (1998) and the ITAC/NSERC award for contributions to information technology (1992).

A simple introduction to Geoffrey Hinton's research can be found in his articles in Scientific American in September 1992 and October 1993. He investigates ways of using neural networks for learning, memory, perception and symbol processing and has over 200 publications in these areas. He was one of the researchers who introduced the back-propagation algorithm that has been widely used for practical applications. His other contributions to neural network research include Boltzmann machines, distributed representations, time-delay neural nets, mixtures of experts, Helmholtz machines and products of experts. His current main interest is in unsupervised learning procedures for neural networks with rich sensory input.

Read more about this video…

Google Tech Talks
November, 8 2007

ABSTRACT

This presentation is about a potential shortcut to artificial intelligence by trading mind-design for world-design using artificial evolution. Evolutionary algorithms are a pump for turning CPU cycles into brain designs. With exponentially increasing CPU cycles while our understanding of intelligence is almost a flat-line, the evolutionary route to AI is a centerpiece of most Kurzweilian singularity scenarios. This talk introduces the Polyworld artificial life simulator as well as results from our ongoing attempt to evolve artificial intelligence and further the Singularity.

Polyworld is the brain child of Apple Computer Distinguished Scientist Larry Yaeger, who remains the primary developer of Polyworld:

http://www.beanblossom.in.us/larryy/Polyworld.html

Speaker: Virgil Griffith
Virgil Griffith is a first year graduate student in Computation and Neural Systems at the California Institute of Technology. On weekdays he studies evolution, computational neuroscience, and artificial life. He did computer security work until his first year of university when his work got him sued for sedition and espionage. He then decided that security was probably not safest field to be in and he turned his life to science.

Read more about this video…