neurodudes

Soon machines will obtain higher IQ’s than humans in intelligence tests.
Traditionally the intelligence quotient has been considered the best indicator for scientifically evaluating natural intelligence.
Is this indeed the best way to measure this human capacity? Could a machine emulate a human being solving traditional intelligence tests? If so, could we affirm that a machine possesses an intelligence equivalent to that of a human?
KITBIT explores some of these possibilities.
The ability of KITBIT in symbolic logic problems, in those which verbal intelligence does not come into play, is comparable to that of humans.
On our web-page, TheIQChallenge.com, we challenge our visitors to put KITBIT to the test in solving numerical and logical problems which have the exact same format as traditional intelligence tests used by psychologists.
The KITBIT project develops research in diverse areas of artificial intelligence such as image recognition, creation of models, predictions and data mining.
KITBIT has been designed by a small team of engineers, mathematicians and programmers. Currently we hope to substantially enlarge this team to carry out our ongoing projects.
www.theIQChallenge.com
www.kitbit.com

[This sounds interesting... although it is fine to promote your personal projects here, we'd at least like to know a little bit about how your project is achieving its goal or how your specific algorithms make this different from similar AI endeavors. And please, please always put your name at the bottom of the post! -Neville]

What gives humans the unique ability to construct novel sentences from the building blocks of language? A recent article in Behavioral and Brain Sciences proposes a “neural blackboard architecture” is capable of just this.

From the article (doi: 10.1017/S0140525X06009022):

“This paper aims to show that neural “blackboard” architectures can provide an adequate theoretical basis for a neural instantiation of combinatorial cognitive structures. [...] We also discuss the similarities between the neural blackboard architecture of sentence structure and neural blackboard architectures of combinatorial structures in visual cognition and visual working memory [...]”

As with all main articles in Behavioral and Brain Sciences, this one is followed by extensive comment and criticism from colleagues, and finally a reply by the authors. This provides a very deep look at the article and the issues surrounding it.

An older, but freely available, version of the article is available here.

The New York Times has an article discussing the current state of consumer-grade robots and a look ahead into the future.

Among many hopeful new directions that are being taken, the article reports that Sony is discontinuing the Aibo. Sad.

New Scientist has the scoop on a state-of-the-art walking robot that navigates uneven terrain, slopes, and significant kicks.

The movie has to be seen to be believed. The robot is called BigDog and it looks like a CGI special effect. But don’t let the film industry desensitize you from the accomplishment here. The robotics engineering required to do this on real terrain is extraordinary. The engineers over at Boston Dynamics really deserve kudos for this.

Looking at static pictures of people running versus pictures of people standing still “evokes a delayed response in an area that overlaps with motionsensitive cortex (hMT+)”. Past studies have indicated a similar response for images depicting a falling cup versus a cup resting on a table.

The paper discusses the role of top-down influence from the temporal lobe as a possible cause for the response. How could this kind of brain activity be influencing our ability to recognize objects in scenes? Is this evidence of the activation of a distributed cortical representation of a moving object?

Should the field of AI be trying to figure out how to replicate a similar top-down influence in next-generation object recognition algorithms?

Abstract from the Journal of Cognitive Neuroscience is available here.

Although Bayle and I are always surprised when we see how many people are actually reading Neurodudes every day (”you really like us! you really do!”), I think we realized we had hit a new milestone when Ray Kurzweil’s book agent called to give us an advance copy of his new book. Let me be clear here: We will gladly review any AI-/neuro-related books you send us. Free books are great! (Heck, we’ll even do an occasional historical biography, if you send us one.)

There’s a lot to say about Kurzweil’’s new book, The Singularity is Near (book website; book on Amazon). This book is similar to his previous books (Age of Intelligent Machines, Age of Spiritual Machines) in style and research but the thesis here is that we are on the precipice of a major change in human civilization: We are soon going to create entities of superior intelligence in all aspects to our own selves. This is the Singularity.

Full book review after the jump (more…)

I skimmed this paper very briefly and it looks cool. Of course, I’m not a computer vision expert so I can’t really tell how state-of-the-art the results are.

Murphy-Chutorian, E., Aboutalib, S., Triesch, J.(2005). Analysis of a Biologically-Inspired System for Real-time Object Recognition. Cognitive Science Online, 3.2, pp. 1-14. http://cogsci-online.ucsd.edu/3/3-3.pdf

“We present a biologically-inspired system for real-time, feed-forward object recognition in cluttered scenes. Our system utilizes a vocabulary of very sparse features that are shared between and within different object models. To detect objects in a novel scene, these features are located in the image, and each detected feature votes for all objects that are consistent with its presence. Due to the sharing of features between object models our approach is more scalable to large object databases than traditional methods. To demonstrate the utility of this approach, we train our system to recognize any of 50 objects in everyday cluttered scenes with substantial occlusion. Without further optimization we also demonstrate near-perfect recognition on a standard 3-D recognition problem. Our system has an interpretation as a sparsely connected feed-forward neural network, making it a viable model for fast, feed-forward object recognition in the primate visual system.”

Here’s a database of ideas from biology that might be useful for engineering.

http://www.bath.ac.uk/~ensab/TRIZ/

(about)

For those with theoretical interests with respect to machine learning flavored AI, the ML Theory blog run by John Langford is highly recommended. Though recently started, Langford and others have so far been doing an excellent job of commenting on both the science and culture of theoretical learning research.

Jeff Hawkins is starting a new company, Numenta, to apply insights from neuroscience into developing better artificial intelligence. We’ll definitely be keeping an eye on this company as we get more details on what kind of AI applications they will be targeting. Undoubtedly, some of Jeff’s ideas in On Intelligence will probably be a large part of it.

Read on for the NY Times article by Markoff…
(more…)