in
which conscious experience is more foundational than physical systems or
linguistic communications
Conscious
experiences are associated with patterns, and patterns are associated with
physical systems, but none of these is fully reducible to the other.
He makes the correct point that roughly-human-level AI’s will have
dramatically different strengths and weaknesses from human being, due to
different sensors and actuators and different physical infrastructures for their
cognitive dynamics. But he doesn’t even
touch the notion of self-modifying AI – the concept that once an AI gets smart
enough to modify its own code, it’s likely to get exponentially smarter and
smarter until it’s left us humans in the dust.
Совершенно не имеет отношения к теме, подход Хокинса легко масштабируется до сверх- и сверх-сверх-сверхчеловеческого интеллекта.
therefore if
AI closely enough emulates the human brain it won’t radically self-modify
either
Rather, I think the problem is that the
field of AI has come to focus on “narrow AI” – programs that solve
particularly, narrowly-defined problems – rather than “artificial general
intelligence” (AGI).
cognitive science, artificial general intelligence, philosophy
of mind and abstract mathematics
т.о. Гортзел признается, что вообще принимает и не считает нужным принимать нейронауку в расчет, т.е. опирается только на эмпирические представления о том, как работает сознание.
So what we’re doing
is creating commercial narrow AI programs, using the software framework that
we’re building out with our AGI design in mind.
и в этом его большое отличие от платформы Хокинса, которая имеет одинаковую структуру для всех ее применений
I tend to largely agree with his take on
the brain
I think he oversimplifies some things fairly seriously – giving
them very brief mention when they’re actually quite long and complicated stories. And some of these omissions, in my view, are
not mere “biological details” but are rather points of serious importance for
his program of abstracting principles from brain science and then
re-concretizing these principles in the context of digital software.
One point Hawkins doesn’t
really cover is how a mind/brain chooses which predictions to make, from among
the many possible predictions that exist.
Hawkins proposes that there are neurons or
neuronal groups that represent patterns as “tokens,” and that these tokens are
then incorporated along with other neurons or neuronal groups into larger
groupings representing more abstract patterns.
This seems clearly to be correct, but he doesn’t give much information
on how these tokens are supposed to be formed.
So, what’s wrong with
Hawkins’ picture of brain function?
Nothing’s exactly wrong with it, so far as I can tell.
But Edelman then takes the concept one step
further and talks about “neural maps” – assemblies of neuronal groups that
carry out particular perception, cognition or action functions. Neural maps, in essence, are sets of
neuronal groups that host attractors of neurodynamics. And Edelman then observes, astutely, that
the dynamics of the population of neuronal groups, over time, is likely to obey
a form of evolution by natural selection.
How fascinating if
the brain also operates in this way!
Hawkins argues that
creativity is essentially just metaphorical thinking, generalization based on
memory. While this is true in a grand
sense, it’s not a very penetrating statement.
Evolutionary learning is the most powerful
general search mechanism known to computer science, and is also hypothesized by
Edelman to underly neural intelligence.
This sort of idea, it seems to me, should be part of any synthetic
approach to brain function.
Hawkins mentions the notion, and observes
correctly that Hebbian learning in the brain is a lot subtler than the simple
version that Donald Hebb laid out in the late 40’s. But he largely portrays these variations as biological details,
and then shifts focus to the hierarchical architecture of the cortex.
Hawkins’ critique of AI, which in my view is
overly harsh. He dismisses work on
formal logic based reasoning as irrelevant to “real intelligence.”
So – to sum up – I think
Hawkins’ statements about brain function are pretty much correct
What he omits are, for instance,
The way the brain displays evolutionary learning
as a consequence of the dynamics of multiple attractors involving sets of
neural clusters
The way the brain may emergently give rise to
probabilistic reasoning via the statistical coordination of Hebbian learning
Learning of predictive patterns requires an explicit or implicit
search through a large space of predictive patterns; evolutionary learning
provides one approach to this problem, with computer science foundations and
plausible connections to brain function; again, Hawkins does not propose any
concrete alternative.
crucial question of how far one has to abstract away from
brain function, to get to something that can be re-specialized into efficient
computer software. My intuition is that
this will require a higher level of abstraction than Hawkins seems to
believe. But I stress that this is a matter
of intuitive judgment – neither of us really knows.
Of course, to interpret the
Novamente design as an “abstraction from the brain” is to interpret this phrase
in a fairly extreme sense – we’re abstracting general processes like
probabilistic inference and evolutionary learning and general properties
like hierarchical structure from the brain, rather than particular
algorithms.
Although I’m
(unsurprisingly) most psyched about the Novamente approach, I think it’s also
quite worthwhile to pursue AGI approaches that are closer to the brain level –
there’s a large space between detailed brain simulation and Novamente, including
neuron-level simulations, neural-cluster-level simulations, and so forth.
"Scientists are now able to switch the focus from particular regions of the brain to the connections between those regions," says Sherif Karama, a psychiatrist and a neuroscientist at McGill University's Montreal Neurological Institute.
A quantifiable "general intelligence factor," known as g, can be statistically extracted from scores on a battery of intelligence tests.
In 2001, Thompson showed that it is correlated with volume in the frontal cortex, a result consistent with a number of studies that have linked intelligence to overall brain size.
In 2007, Jung and Richard Haier, now professor emeritus of psychology at the University of California, Irvine, developed the first comprehensive theory drawn from neuroimaging of how the brain gives rise to intelligence.
As we "evolved from worms to humans," says George Bartzokis, a professor of psychiatry at UCLA, the number of non-neural cells in the brain increased 50 times more than the number of neurons. He adds, "My hypothesis has always been that what gives us our cognitive capacity is not actually the number of neurons, which can vary tremendously between human individuals, but rather the quality of our connections."
The type of MRI typically used for medical scans does not show the finer details of the brain's white matter. But with a technique called diffusion tensor imaging (DTI), which uses the scanner's magnet to track the movement of water molecules in the brain, scientists have developed ways to map out neural wiring in detail. While water moves randomly within most brain tissue, it flows along the insulated neural fibers like current through a wire.
Neuralangelo by NVIDIA: Neuralangelo, a new Artificial Intelligence model by NVIDIA Research for 3D reconstruction using neural networks (blogs.nvidia.com).
Miller and Cohen propose an Integrative Theory of Prefrontal Cortex Function. The two theorize that “cognitive control stems from the active maintenance of patterns of activity in the prefrontal cortex that represents goals and means to achieve them. They provide bias signals to other brain structures whose net effect is to guide the flow of activity along neural pathways that establish the proper mappings between inputs, internal states, and outputs needed to perform a given task” (Miller & Cohen, 2001). Essentially the two theorize that the prefrontal cortex guides the inputs and connections which allows for cognitive control of our actions.
Friston recently expanded on this to suggest an inversion method for hierarchical Bayesian dynamic models and to point out that the brain, in principle, has the infrastructure needed to invert hierarchical dynamic models [6].
In a recent review, Hegde and Felleman pointed out that the “Bayesian framework is not yet a neural model. [The Bayesian] framework currently helps explain the computations that underlie various brain functions, but not how the brain implements these computations” [2]. This paper is an attempt to fill this gap by deriving a computational model for cortical circuits based on the mathematics of Bayesian belief propagation in the context of a particular Bayesian framework called Hierarchical Temporal Memory (HTM).
This paper's other author, George, recognized that the Memory-Prediction framework could be formulated in Bayesian terms and given a proper mathematical foundation [8],[9].
Several researchers have proposed detailed models for cortical circuits [10]–[12].
Other researchers [4],[13] have proposed detailed mechanisms by which Bayesian belief propagation techniques can be implemented in neurons.
In 2007, Jung and Richard Haier, now professor emeritus of psychology at the University of California, Irvine, developed the first comprehensive theory drawn from neuroimaging of how the brain gives rise to intelligence.
Applying existing theories of how information flows in the brain, Jung and Haier hypothesized that neural signals travel from nodes near the back of the brain, where sensory data is collected and synthesized, to those in the frontal lobes, which are responsible for decision making and planning. The connections between these nodes, they argued, are just as critical as the nodes themselves.