Areas that should be studied if you want to create artificial Human Brains
Cognitive science, Evolutionary psychology, Information theory and Computer programming are the four
major areas of study for an AI programmer. People on the forums seem to agree that probability theory
(Bayes etc.), philosophy of mind and ethics, and cognitive science are very important. The same seems
to be true for evolutionary psychology, biology and algorithms, some neuroscience/anatomy, neural
networks, optimization, agent architectures and decision theory. The usefulness of logic, dynamical
systems theory and computational complexity theory seems to be more open for debate. Other more
traditional computer science and narrow AI disciplines were mostly dismissed.
MIT has the biggest AI lab in the world. It started the Mind Machine Project. Apparently they want to
"undo" the wrong turns AI has taken decades ago and "start fresh". This seemed pretty great to me,
since it shows that they are willing to think outside the box and apparently think that most of narrow AI
was not the way to go. Unfortunately, their goals seem pretty vague. Furthermore, I've seen people
express doubt about the fact that a lot of people heading the effort are the ones who made those wrong
turns in the first place.
On the other hand, a lot of talented people are working on AI at MIT, and their Human Intelligence
Enterprise group and Genesis project seem really interesting.
Stanford also has a big name in AI and an extensive AI lab where they seem to be working, amongst
other things, on the Icarus computational theory of the cognitive architecture and General Game
Playing. Although GGP isn't as general as we would like an AI to be, I wonder if the pursuit of generality
in an (intermediately) limited domain can nevertheless provide valuable insights. I also heard that Dr.
Noah Goodman is doing AGI-related work.
It seems that there are a fair number of universities that are working on what they call cognitive
architectures. I'm not sure if this is just a alternative term for AGI, or if there are more significant
differences. In case it's the same, or studying at these universities is can still be helpful for learning
about AGI, I'll list the projects and their universities anyway:
The Soar cognitive architecture seems to be most widely used, including by the universities of Michigan,
Hertfordshire, Southern California, Nottingham, Portsmouth and Carnegie Mellon. It seems that UMich's
Brain, Cognition and Action Lab has also worked on a cognitive architecture called "EPIC", but I'm not
sure if this project is still active, since all information I see seems to be at least 4 years old.
Carnegie Mellon also seems to be developing the ACT-R (Adaptive Control of Thought—Rational) and
4CAPS (Cortical Capacity-Constrained Concurrent Activation-based Production System) cognitive
architectures.
Rensselaer Polytechnic Institute is working on CLARION and their Human-Level Intelligence Laboratory's
goal is to achieve human-level computational intelligence and to explain the abilities of human
intelligence (based on the Polyscheme and Cognitive Substrate theories).
At the University of Memphis the Cognitive Computing Research Group's stated objective is "the design
and implementation of cognitive, sometimes "conscious," software agents, their computational
applications, and their use in cognitive modeling".
The Humboldt-University of Berlin and University of Osnabrück have been working on the MicroPsi
project, which is based on Dietrich Dörner's Psi (Principles of synthetic intelligence) theory.
__PEOPLE__
For the following institutions I didn't immediately see any projects that were (to me) clearly AGI related,
but they employ(ed) people who are pursuing AGI, which leads me to believe that they may be
(relatively) good environments to study this subject.
Dr. Wang Pei works at Temple University and seems to be quite busy with his Non-Axiomatic Reasoning
System (NARS). However, he has stated in the past that he is pretty much the only one working on AGI
in his department. I wonder if the situation at Indiana University is better, since Dr. Wang wrote his
dissertation there (for which I assume he had a supervisor) and the Center for Research on Concepts
and Cognition there is lead by the highly praised Dr. Douglas Hofstadter.
A couple of other people whose names I frequently see popping up (Dr. Schmidhuber, Dr. Hutter and Dr.
Legg) seem to do or have done AGI-related work at the Dalle Molle Institute for Artificial Intelligence
(IDSIA), so that seems interesting as well. Dr. Hutter has since moved to the Australian National
University, and Dr. Legg is now working at the Gatsby Computation Neuroscience Unit, which is part of
University College London. Dr. Hugo de Garis seems to have done a lot of AGI-related work at Xiamen
University. Finally, Dr. Itamar Arel runs the Machine Intelligence Lab at the University of Tennessee.
I'm sure I have failed to mention many important people (in academics) here. I don't mean to offend
anyone. I'm fairly new to the field and these are the people I happened to stumble upon.
http://prize.hutter1.net/
Next
Cognitive science, Evolutionary psychology, Information theory and Computer programming are the four
major areas of study for an AI programmer. People on the forums seem to agree that probability theory
(Bayes etc.), philosophy of mind and ethics, and cognitive science are very important. The same seems
to be true for evolutionary psychology, biology and algorithms, some neuroscience/anatomy, neural
networks, optimization, agent architectures and decision theory. The usefulness of logic, dynamical
systems theory and computational complexity theory seems to be more open for debate. Other more
traditional computer science and narrow AI disciplines were mostly dismissed.
MIT has the biggest AI lab in the world. It started the Mind Machine Project. Apparently they want to
"undo" the wrong turns AI has taken decades ago and "start fresh". This seemed pretty great to me,
since it shows that they are willing to think outside the box and apparently think that most of narrow AI
was not the way to go. Unfortunately, their goals seem pretty vague. Furthermore, I've seen people
express doubt about the fact that a lot of people heading the effort are the ones who made those wrong
turns in the first place.
On the other hand, a lot of talented people are working on AI at MIT, and their Human Intelligence
Enterprise group and Genesis project seem really interesting.
Stanford also has a big name in AI and an extensive AI lab where they seem to be working, amongst
other things, on the Icarus computational theory of the cognitive architecture and General Game
Playing. Although GGP isn't as general as we would like an AI to be, I wonder if the pursuit of generality
in an (intermediately) limited domain can nevertheless provide valuable insights. I also heard that Dr.
Noah Goodman is doing AGI-related work.
It seems that there are a fair number of universities that are working on what they call cognitive
architectures. I'm not sure if this is just a alternative term for AGI, or if there are more significant
differences. In case it's the same, or studying at these universities is can still be helpful for learning
about AGI, I'll list the projects and their universities anyway:
The Soar cognitive architecture seems to be most widely used, including by the universities of Michigan,
Hertfordshire, Southern California, Nottingham, Portsmouth and Carnegie Mellon. It seems that UMich's
Brain, Cognition and Action Lab has also worked on a cognitive architecture called "EPIC", but I'm not
sure if this project is still active, since all information I see seems to be at least 4 years old.
Carnegie Mellon also seems to be developing the ACT-R (Adaptive Control of Thought—Rational) and
4CAPS (Cortical Capacity-Constrained Concurrent Activation-based Production System) cognitive
architectures.
Rensselaer Polytechnic Institute is working on CLARION and their Human-Level Intelligence Laboratory's
goal is to achieve human-level computational intelligence and to explain the abilities of human
intelligence (based on the Polyscheme and Cognitive Substrate theories).
At the University of Memphis the Cognitive Computing Research Group's stated objective is "the design
and implementation of cognitive, sometimes "conscious," software agents, their computational
applications, and their use in cognitive modeling".
The Humboldt-University of Berlin and University of Osnabrück have been working on the MicroPsi
project, which is based on Dietrich Dörner's Psi (Principles of synthetic intelligence) theory.
__PEOPLE__
For the following institutions I didn't immediately see any projects that were (to me) clearly AGI related,
but they employ(ed) people who are pursuing AGI, which leads me to believe that they may be
(relatively) good environments to study this subject.
Dr. Wang Pei works at Temple University and seems to be quite busy with his Non-Axiomatic Reasoning
System (NARS). However, he has stated in the past that he is pretty much the only one working on AGI
in his department. I wonder if the situation at Indiana University is better, since Dr. Wang wrote his
dissertation there (for which I assume he had a supervisor) and the Center for Research on Concepts
and Cognition there is lead by the highly praised Dr. Douglas Hofstadter.
A couple of other people whose names I frequently see popping up (Dr. Schmidhuber, Dr. Hutter and Dr.
Legg) seem to do or have done AGI-related work at the Dalle Molle Institute for Artificial Intelligence
(IDSIA), so that seems interesting as well. Dr. Hutter has since moved to the Australian National
University, and Dr. Legg is now working at the Gatsby Computation Neuroscience Unit, which is part of
University College London. Dr. Hugo de Garis seems to have done a lot of AGI-related work at Xiamen
University. Finally, Dr. Itamar Arel runs the Machine Intelligence Lab at the University of Tennessee.
I'm sure I have failed to mention many important people (in academics) here. I don't mean to offend
anyone. I'm fairly new to the field and these are the people I happened to stumble upon.
http://prize.hutter1.net/
Next
 | |  | |  | |  |