Artificial General Intelligence (AGI) - or strong AI - refers to a domain-independent, machine-based system that approaches or exceeds human performance on any and all cognitive tasks. Estimates for the arrival of true AGI solutions range from last week (as in, we have one!) to decades, to infinity and beyond. As the general study of cybernetic systems and modern AI and cognitive computing capture the imagination of civic and business leaders, and fans of science fiction, it is important to be able to distinguish between progress and smoke & mirrors.
This webinar will present an overview of approaches to AGI, examples of promising research and commercial AGI activities, and show participants how to critically evaluate academic and vendor claims.
TrustArc Webinar - How to Build Consumer Trust Through Data Privacy
Smart Data Webinar: Artificial General Intelligence - When Can I Get It?
1. Artificial General Intelligence
When Can I get it?
Adrian Bowles, PhD
Founder, STORM Insights, Inc.
info@storminsights.com
Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
FEBRUARY 9, 2017
2. Foundations of AI & AGI
Games & AI/AGI
AGI Today
Overview of AGI Approaches
Interesting Research
Artificial vs Augmented General Intelligence
Evaluating Claims - Are We There Yet?
Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
AGENDA
3. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
FOUNDATIONS OF AI AND AGI
4. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
CONTEXT - HOW DID WE GET HERE? (AND WHERE ARE WE ANYWAY?)
AI Roots
AGI - Artificial General Intelligence
Focus on replicating intelligence by copying
brain functions and form/process
Natural Language Processing (NLP)
Learning and discovery
Heuristics, expert rules…
Logic - symbolic logic and
mechanical theorem proving
Strategy: Replace
Execution: Open concepts
Constraint: Processing
Modern AI
Focus on augmenting intelligence by
evidence-based interaction
Natural Language Processing (NLP)
Learning and discovery
Distributed ML driven by big data
Deep QA techniques
Strategy: Reinforce
Execution: Open code and data
Constraint: Data
5. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
IN THE BEGINNING
“We propose that a 2 month, 10 man study of artificial intelligence be carried out during the
summer of 1956 at Dartmouth College in Hanover, New Hampshire.
The study is to proceed on the basis of the conjecture that every aspect of learning or any other
feature of intelligence can in principle be so precisely described that a machine can be
made to simulate it. An attempt will be made to find how to make machines use language,
form abstractions and concepts, solve kinds of problems now reserved for humans, and
improve themselves.We think that a significant advance can be made in one or more of these
problems if a carefully selected group of scientists work on it together for a summer.”
A PROPOSAL FOR THE DARTMOUTH SUMMER RESEARCH PROJECT ON ARTIFICIAL INTELLIGENCE
J. McCarthy, Dartmouth College
M. L. Minsky, Harvard University
N. Rochester, I.B.M. Corporation
C.E. Shannon, Bell Telephone Laboratories
August 31, 1955
Emphasis added
6. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
FROM THE 1955 DARTMOUTH SUMMER AI RESEARCH PROPOSAL
The following are some aspects of the artificial intelligence problem:
1 Automatic Computers
If a machine can do a job, then an automatic calculator can be programmed to simulate the machine. The
speeds and memory capacities of present computers may be insufficient to simulate many of the higher
functions of the human brain, but the major obstacle is not lack of machine capacity, but our inability to write
programs taking full advantage of what we have.
2. How Can a Computer be Programmed to Use a Language
It may be speculated that a large part of human thought consists of manipulating words according to rules of
reasoning and rules of conjecture. From this point of view, forming a generalization consists of admitting a
new word and some rules whereby sentences containing it imply and are implied by others. This idea has
never been very precisely formulated nor have examples been worked out.
7. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
FROM THE 1955 DARTMOUTH SUMMER AI RESEARCH PROPOSAL
The following are some aspects of the artificial intelligence problem:
3. Neuron Nets
How can a set of (hypothetical) neurons be arranged so as to form concepts. Considerable theoretical and
experimental work has been done on this problem by Uttley, Rashevsky and his group, Farley and Clark, Pitts and
McCulloch, Minsky, Rochester and Holland, and others. Partial results have been obtained but the problem needs
more theoretical work.
4. Theory of the Size of a Calculation
If we are given a well-defined problem (one for which it is possible to test mechanically whether or not a proposed
answer is a valid answer) one way of solving it is to try all possible answers in order. This method is inefficient,
and to exclude it one must have some criterion for efficiency of calculation. Some consideration will show that to
get a measure of the efficiency of a calculation it is necessary to have on hand a method of measuring the
complexity of calculating devices which in turn can be done if one has a theory of the complexity of functions.
Some partial results on this problem have been obtained by Shannon, and also by McCarthy.
8. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
The following are some aspects of the artificial intelligence problem:
5. Self-lmprovement
Probably a truly intelligent machine will carry out activities which may best be described as self-
improvement. Some schemes for doing this have been proposed and are worth further study. It seems likely
that this question can be studied abstractly as well.
6. Abstractions
A number of types of ``abstraction'' can be distinctly defined and several others less distinctly. A direct
attempt to classify these and to describe machine methods of forming abstractions from sensory and other
data would seem worthwhile.
7. Randomness and Creativity
A fairly attractive and yet clearly incomplete conjecture is that the difference between creative thinking
and unimaginative competent thinking lies in the injection of a some randomness. The randomness must be
guided by intuition to be efficient. In other words, the educated guess or the hunch include controlled
randomness in otherwise orderly thinking.
FROM THE 1955 DARTMOUTH SUMMER AI RESEARCH PROPOSAL
9. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
WHERE DOES AGI FIT?
Learning Model
External Internal
Knowledge
Domain
Broad/
Unbounded
Narrow/
Constrained
AGI
10. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
PERCEPTION
UNDERSTANDING
LEARNING
PLANNING
Hardware
Software
Mimic Model
MOTIVATION PROBLEM-SOLVING
Classic
AI
CLASSIC IS NARROW, NOT AGI
NLP
11. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
Machine
Learning
Big
Data
Hardware
Software
Neuromorphic
TPUs
NPUs
GPUs
Mimic
GPUs
?
Model
HTM
MBR
Neural Nets
Classic
AI
#MODERNAI IS NARROW, NOT AGI
13. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
AI OR NOT AI?
14. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
GREAT EXPECTATIONS
8/9/2006
15. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
AI SPRING - VC ECOSYSTEMS
16. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
AI SPRING - VC ECOSYSTEMS
17. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
NOT SO FAST…
“At DeepMind, engineers have created programs based on neural
networks, modelled on the human brain. These systems make mis-
takes, but learn and improve over time. They can be set to play
other games and solve other tasks, so the intelligence is general,
not specific. This AI “thinks” like humans do.”
Financial Times, March 11, 2016. Dennis Hassabis, master of the new machine age.
(On Google’s AlphaGo)
18. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
RECOGNITION IS NOT UNDERSTANDING.
https://arxiv.org/abs/1112.6209
19. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
GAMES AND AI/AGI
20. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
AI OR NOT AI?
The LIFE Picture Collection/Gett
21. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
THE EDGE OF THE ENVELOPE IS ALWAYS MOVING
22. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
THE ROLE OF GAMES IN AI RESEARCH
2-Person
Perfect Information
Zero Sum
Checkers Chess Go
Arthur Samuel
IBM
1997 20161956
23. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
THE ROLE OF GAMES IN AI RESEARCH
3-Person
Imperfect Information
Zero Sum
Natural Language
Jeopardy! Poker
2-6-? —Person
Imperfect Information, Zero Sum
2011 2017
24. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
AI & THE BLUFF
25. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
AUGMENTED INTELLIGENCE FOR CHESS
26. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
AGI TODAY
27. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
IQ - THE GENERAL FACTOR (G)
IQ derived from a factor analysis of correlations between multiple tests.
Charles Spearman, 1904
General ability + narrow ability factors
There is no accepted g-factor for AI.
IBM True North Chips on a
SyNAPSE board.
28. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
Hearing (audioception)
~12,000 outer hair cells/ear
~3,500 inner hair cells
Vision (ophthalmoception)
Photoreceptors - Per Eye
~120,000,000 rod cells
(triggered by single photon)
~6,000,000 cone cells
(require more photons to trigger)
~ 60,000 photosensitive
ganglion cells
Touch (tactioception)
Thermoreceptors, mechanoreceptors,
chemoreceptors and nociceptors for touch, pressure, pain,
temperature, vibration
Smell (olfacoception)
Chemoreception
Taste (gustaoception)
Chemoreception
Human Cognition
~100,000,000,000 (100B) Neurons
~100-500,000,000,000,000 (100-500T) Synapses
AGI VS NATURAL GENERAL INTELLIGENCE
Learn
ModelReason
Understand
Plan
29. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
AGI MINIMUM REQUIREMENTS
or
Big Knowledge + Modest Processing
(Reasoning, KM…)
Big Processing + Big Data
(Reasoning, KM…)
With sufficient processing power, and
access to enough clean, validated data,
just in time knowledge acquisition.
Starting with sufficient knowledge
(includes the model with
assumptions) makes processing
requirements relatively modest to
accommodate incremental activities.
30. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
FUNDAMENTAL DESIGN CHOICE: SYMBOLS VS STATISTICS
Symbolic Logic
Representations
Reasoning
Concepts
Statistical Models
Mechanical Theorem Proving
31. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
REPRESENTATIVE AGI APPROACHES
Wikipedia contributors. "Cog (project)." Wikipedia, The Free Encyclopedia.
Wikipedia, The Free Encyclopedia, 10 Jul. 2016. Web. 8 Feb. 2017.
Focus on
human interaction
Focus on
machine learning
Focus on
capturing common knowledge
Focus on
brain-inspired architectures
Focus on representation,
philosophy and linguistics
32. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
OPENCOG: AN AGI FRAMEWORK
Knowledge represented in hypergraphs
(an edge can join n-vertices)
35. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
TRUE AGI CAN FUNCTION AS AUGMENTED GENERAL INTELLIGENCE
“I’m sorry Dave, I’m afraid I can’t do that…
This mission is too important for me to allow you to jeopardize it…
I know that you an Frank were planning to disconnect me
and I’m afraid that’s something I cannot allow to happen.”
HAL, 2001 A Space Odyssey
36. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
A fool with a tool is still a fool.
Collaborative
Evidence-Driven
Probabalistic
AGI TODAY = AUGMENTED GENERAL INTELLIGENCE
37. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
REVISITING THE 1955 DARTMOUTH SUMMER AI RESEARCH PROPOSAL
The following are some aspects of the artificial intelligence problem:
1 Automatic Computers
2. How Can a Computer be Programmed to Use a Language
3. Neuron Nets
4. Theory of the Size of a Calculation
5. Self-lmprovement
6. Abstractions
7. Randomness and Creativity
What does it mean to use vs understand?
The basis for modern machine learning.
In 60+ years, we have become adept at programming.
Well researched and documented progress
quantifying algorithmic complexity.
Partial credit, but much work remains to be done.
The next frontier?
Beyond ML techniques, this area is still full of open questions.
38. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
IS IT AGI? MY QUICK TEST
CAN I SEE IT?We
Have
AGI!
Show
Me!
DOES IT REQUIRE HUMAN INTERVENTION
TO LEARN ABOUT NEW DOMAINS?
CAN IT LEARN TO LEARN?
CAN IT COMMUNICATE ITS FINDINGS?
CAN IT ASK FOR HELP/MISSING DATA/KNOWLEDGE?
NO
YES
NO
NO
NO
No
39. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
KEEP IN TOUCH
adrian@storminsights.com
Twitter @ajbowles
Skype ajbowles
Upcoming 2017 Webinar Dates & Topics
March 9 Data Science and Business Analysis:
A Look at Best Practices for Roles, Skills, and Processes
April 13 Machine Learning - Moving Beyond Discovery to Understanding
May 11 Streaming Analytics for IoT-Oriented Applications
40. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
RESOURCES
http://bobkirby.info:8080/comparison.htmBob Kirby’s Knowledge Representation
Comparisons
https://www.theatlantic.com/technology/
archive/2012/11/noam-chomsky-on-where-
artificial-intelligence-went-wrong/261637/
Noam Chomsky on Where Artificial
Intelligence Went Wrong
http://opencog.orgThe OpenCog Foundation
http://www.businessinsider.com/
cycorp-ai-2014-7
Cyc
http://www.cyc.com
The AI Behind Watson http://www.aaai.org/Magazine/Watson/
watson.php
41. Copyright (c) 2017 by STORM Insights Inc. All Rights Reserved.
RESOURCES
https://www.cmu.edu/news/stories/archives/
2017/january/AI-tough-poker-player.html
CMU ARTIFICIAL INTELLIGENCE IS
TOUGH POKER PLAYER
https://www.theatlantic.com/technology/
archive/2016/03/the-invisible-opponent/
475611/
How Google's AlphaGo Beat a Go World
Champion