Artificial Intelligence

An early classic in computational neuroscience was a 1993 paper by Elman called "The Importance of Starting Small." The paper describes how initial limitations in a network's memory capacity could actually be beneficial to its learning of complex sentences, relative to networks that were "adult-like" from the start. This still seems like a beautiful idea - the cognitive limitations of children may somehow be adaptive for the learning they have yet to do. And Elman is not alone in proposing it; a number of other researchers have proposed that a lack of cognitive control or working memory…
What if training ourselves on one task yielded improvements in all other tasks we perform? This is the promise of the cognitive training movement, which is increasingly showing that such "far transfer" of training is indeed possible, while short of being "universal transfer." Interestingly, this phenomenon might be most likely to occur for some of the most abstract and challenging cognitive functions. New evidence for this claim comes from an in-press article at Psychological Science, by Persson & Reuter-Lorenz. The authors used several tasks which have been shown to engage the left…
Much evidence supports the idea that parietal cortex is involved in the simple maintenance of information, such as in object permanence paradigms (also here) and other tasks. This evidence is part of the justification for the "parietofrontal integration theory", which suggests that parietal areas work in concert with prefrontal regions of the brain to accomplish the maintenance and manipulation of information. Orthodoxy holds the prefrontal cortex is more involved than parietal cortex in information manipulation (eg). However, some have suggested that the spatial transformations…
To enhance any system, one first needs to identify its capacity-limiting factor(s). Human cognition is a highly complex and multiply constrained system, consisting of both independent and interdependent capacity-limitations. These "bottlenecks" in cognition are reviewed below as a coherent framework for understanding the plethora of cognitive training paradigms which are currently associated with enhancements of working memory, executive function and fluid intelligence (1,2, 3, 4, 5, 6, 7, 8, 9, 10, c.f. 11, 12, 13). By far, the most common complaint about limitations in cognition is…
Working memory - the ability to hold information "in mind" in the face of environmental interference - has traditionally been associated with the prefrontal cortices (PFC), based primarily on data from monkeys. High resolution functional imaging (such as fMRI) have revealed that PFC is just one part of a larger working memory network, notably including the parietal cortex, which has long been the focus of research in the visual domain, and is primarily thought to carry out spatial computations. What role might such spatial computations have in working memory? Wendelken, Bunge & Carter…
The organization of the human prefrontal cortex (PFC) is a lasting mystery in cognitive neuroscience, but not for lack of answers - the issue is deciding among them, since all seem to characterize prefrontal function in very different but apparently equally-valid ways. If this mystery were resolved, it could revolutionize cognitive neuroscience and neuropsychology as well as education and artificial intelligence - prefrontal cortex is crucial for the most high-level cognitive abilities we know of, including the executive functions, but no existing theory of intelligence specifies the exact…
Peter Hankins has written an excellent commentary criticizing the "positive comparisons" I make after contrasting brains with computers. Peter says: "... the concept of processing speed has no useful application in the brain rather than that it isn't fixed." While this statement may intuitively appeal to some philosophers, temporal limitations in neural processing are both critical for neuronal function and well accepted in both neuroscience and psychometrics. At the biological level, the membrane capacitance of neurons is important for regulating the firing rate of neurons, which itself has…
How does the human brain construct intelligent behavior? Computational models have proposed several mechanisms to accomplish this: the most well known is "Hebbian learning," a process mathematically similar to both principal components analysis and Bayesian statistics. But other neural learning algorithms must exist - how else could the brain disentangle mere correlations from true causation? Temporal precedence helps to some extent - and does seem to play a large role in Hebbian learning (e.g., spike-timing dependent plasticity). But the smell of rain does not actually cause rain -…
Phil Stearns has constructed a 45 "neuron" network of electronic parts which responds to lights and tones with a (rather cute) squealing sound. A picture of the components for this strange device: Each "neuron" consisted of analog electronics corresponding to each of 6 functions: Input, Summing, Threshold, "Offset," "Output," and "Structure" (not sure about those latter three). The connectivity was determined by hand. Phil states that the sculpture is not intelligent, but rather "some kind of squid baby." Neural networks have great potential for contributing to the arts. For example, JP…
One of the bottlenecks in human memory capacity is its "filtering efficiency" - irrelevant information in memory only detracts from an already-constrained memory span. New work by McNab & Klingberg images the neural structure directly responsible for such filtering, and shows it can predict behavioral measures of memory span. Impressively, the location of this "memory filter" is the globus pallidus, as predicted by a computational network model of cortex, but in contrast to that model, it shows functional correlations with parietal in addition to frontal areas. This work has immediate…
The world wide web can be understood as a giant matrix of associations (links) between various nodes (web pages). At an abstract level, this is similar to human memory, consisting of a matrix of associations (learned relationships, or neuronal connections) between various nodes (memories, or the distributed representations constituting them). In the new issue of Psych. Science, Griffiths et al. ask whether Google's famously accurate and fast PageRank algorithm for internet search might behave similarly to the brain's algorithm - whatever that might be - for searching human memory. About…
Speech recognition remains a daunting challenge for computer programmers partly because the continuous speech stream is highly under-determined. For example take coarticulation, which refers to the fact that the auditory frequencies corresponding to a given letter are strongly influenced by the letters both preceding and following it - sometimes interpreted to mean that there is no invariant set of purely auditory characteristics defining any given letter. Thus it's difficult to recover the words that a person is saying, since each part of that word is influenced by the words surrounding it…
A lack of clear definitions for terms like "intelligence" and "consciousness" plagues any serious discussion of those concepts. A recent article by Seth, Baars & Edelman argues for a core set of 17 properties that are characteristic of consciousness, and could be used in the "diagnosis" of consciousness in humans and other animals. Property 1: "Irregular" patterns of brain activity Electrical oscillations occuring between 20 and 70 times per second are common in awake humans, but epilepsy, sleep, anesthesia and some forms of brain damage are accompanied by the dominance of highly regular…
Among nature's most impressive feats of engineering is the remarkably flexible and self-optimizing quality of human cognition. People seem to dynamically determine whether speed or accuracy is of utmost importance in a certain task, or whether they should continue with a current approach or begin anew with another, or whether they should rely on logic or intuition to solve a certain problem. A topic of intense research in cognitive neuroscience is how cognition can be made so flexible. One possibility proposed by by Brown, Reynolds & Braver is that cognitive control is multi-faceted, in…
Very early in the history of artificial intelligence research, it was apparent that cognitive agents needed to be able to maximize reward by changing their behavior. But this leads to a "credit-assignment" problem: how does the agent know which of its actions led to the reward? An early solution was to select the behavior with the maximal predicted rewards, and to later adjust the likelihood of that behavior according to whether it ultimately led to the anticipated reward. These "temporal-difference" errors in reward prediction were first implemented in a 1950's checker-playing program,…
"To understand ourselves, we must embrace the alien." - PZ Meyers One difficulty in understanding consciousness is the fact that we know of only one species that certainly possesses it: humans. A new article by Jennifer Mather suggests that octopi may also possess consciousness, despite the vastly different architecture of their brain. If two very different neural architectures can both support forms of advanced cognition, then the similarities between them may help clarify the computational requirements for intelligent behavior. Octopus brains are striking different from those in primates…
"A good metaphor is something even the police should keep an eye on." - G.C. Lichtenberg Although the brain-computer metaphor has served cognitive psychology well, research in cognitive neuroscience has revealed many important differences between brains and computers. Appreciating these differences may be crucial to understanding the mechanisms of neural information processing, and ultimately for the creation of artificial intelligence. Below, I review the most important of these differences (and the consequences to cognitive psychology of failing to recognize them): similar ground is…
In the new issue of Seed, Douglas Hofstadter talks about "strange loops" - his term for patterns of level-crossing feedback inside some medium (such as neurons) - and their role in consciousness. Likewise, Gerald Edelman has talked about how a "reentrant dynamic core" of neural activity could tightly integrate large groups of neurons through positive feedback cycles. Similarly, many view interactions among neural oscillations as a candidate mechanism for the formation of consciousness - such oscillations can perform abstract computations (as in liquid state machines) and can interact with…
Whereas yesteryear's artificial neural networks models were focused on achieving basic biological plausibility, today's cutting edge networks are modeling cognitive phenomena at the level of neurotransmitters. In a great example of this development, McClure, Gilzenrat & Cohen have an article in Advances in Neural Information Processing Systems where they propose a role for both dopamine and norepinephrine in switching behavior between modes of "exploration" and "exploitation." First, a little background. In artificial intelligence circles, the "temporal difference" algorithm has been a…
As enigmatic as prefrontal function seems to be, the anterior portions of prefrontal cortex (aPFC) are even more mysterious. This results partly from the fact that aPFC is particularly difficult to access and study electrophysiologically in nonhuman primates, as Ramnani and Owen note in their 2004 Nature Reviews Neuroscience article, and so detailed neuroanatomical investigations of aPFC have been conducted only recently. The authors report how this work has led to a breakthrough in the understanding of aPFC's computations. Ramnani and Owen review Brodmann's early analysis of aPFC, also…