Diahuman AI level of progress
Diahuman AI, according to J. Storrs Hall, is the phase of AI development similar to humans who don't
learn well and who function competently only at simple jobs for which they must be trained.
The core of the diahuman phase will be the development of autogenous learning.
In later phases, AI will be more autonomous and learn what it needs to know and deciding what it needs
to learn.
Will we have inexpensive robots with diahuman intelligence or will AI advance beyond that stage by the
time robots are inexpensive enough to put on production lines where humans now serve?
parahuman AI
Cyhthia Breazeal's sociable robots are forerunners of robots which will interact with humans. The
Japanese have robots which serve as receptionists. They speak many languages, which is a plus over
most human receptionists.
According to Hall, a parahuman AI should act like a lawyer, doctor or accountant with deep knowledge
and endless patience. Once AI and cognitive science is sufficiently advanced AI teachers will be built
that will model in detail how each individual student absorbs material and will optimize the presentation
so as to maximize understanding as well as motivate and stimulate the student.
The downside of parahuman AI is that it can be used to manipulate humans through advertising or being
an individual con man.
http://www.nsf.gov/discoveries/disc_summ.jsp?cntn_id=112947&org=NSF
Synthetic Brains
Researchers study the feasibility of brains made from carbon nanotubes
Researchers are building mathematical models that accurately reflect neuron connections.
January 27, 2009
Synthetic brains are a long way from reality, but researchers at the University of Southern California,
funded by the National Science Foundation, are taking the first steps to build neurons from carbon
nanotubes that emulate human brain function.
"At this point we still don't know if building a synthetic brain is feasible," said Alice Parker, professor of
electrical engineering. "It may take decades to realize anything close to the human brain but emulating
pieces of the brain, such as a synthetic vision system or synthetic cochlea that interface successfully
with a real brain may be available quite soon, and synthetic parts of the brain's cortex within decades."
The challenges to creating a synthetic brain are staggering. Unlike computer software that simulates
brain function, a synthetic brain will include hardware that emulates brain cells, their amazingly complex
connectivity and a concept Parker calls "plasticity," which allows the artificial neurons to learn through
experience and adapt to changes in their environment the way real neurons do.
There is also the matter of scale. By 2022, with conventional technology, if the team could construct a
synthetic brain that emulated real brain function, even crudely, it would take 100 billion artificial neurons
and a very a large room to hold them.
"Obviously the technology will have to be downsized to aid a human being or be feasible as a robot
brain," Parker said. Power is another consideration. The power requirements for a synthetic brain are
staggering because a human brain never turns off. "In a transistor things are on or off so it's a
black-or-white situation, but in the brain there are also many shades of gray and power is continuously
being consumed," Parker noted.
But before the researchers can tackle concerns of power and scale, they are building mathematical
models that accurately reflect the Byzantine connections of all the neurons and demonstrate how the
connections allow neurons to communicate with each other.
Each neuron in the cortex--a part of the brain that contributes significantly to conscious thought and
intelligence--is connected to tens of thousands of other neurons. The researchers are also
implementing the complex computations carried out by each neuron on all the inputs it receives from
other neurons.
"It's a nonlinear phenomenon and almost impossible to model but that's what we're attempting to do,"
Parker said.
The researchers have shown that portions of a neuron can be modeled electronically using carbon
nanotube circuit models and have performed detailed simulations of the circuit models. A single
archetypical neuron, including excitatory and inhibitory synapses, has been modeled electronically and
simulated. Parker and her co-researcher Chongwu Zhou are in the process of combining these circuit
models of neurons to create a functional carbon nanotube circuit model of a small network of neurons.
This small network of interconnected neurons will be simulated using the carbon nanotube models. This
network demonstrates an interesting neural circuit that detects moving edges in a selected direction.
Parker believes carbon nanotubes are an ideal material to emulate brain function because their
three-dimensional structure allows connectivity in all directions on all planes and because a
carbon-based prosthesis is less likely to be rejected by the human body than one made from inorganic
materials. But their invasive nature could result in them invading surrounding tissue and prompting
lesions and cancers.
"It's a possibility and something else that needs to be addressed for the technology to be feasible,"
Parker said.
As the researchers move ahead with their mathematical modeling and neuron construction, beginning
with a single synapse, they ponder "plasticity," neuroscientists' term for the brain's ability to learn and
adapt to change. "Our brains can grow new neurons and the synapses between them in an hour--a
remarkable biological feature that is difficult to emulate from an engineering perspective," Parker said.
Emulating such plasticity in a synthetic brain will require a major leap in technology, similar to the leap
from cathode ray tubes to transistors. "We don't know what the new technology will look like yet, but it
will be a technology that can self-assemble and reshape itself.
As we work in the lab building neurons or constructing mathematical models, we must consider the
requirement of plasticity, even if we don't yet know what it looks like."
Aside from the daunting technological challenges, a synthetic brain or brain components will also raise
ethical and environmental issues. The role of emotions in learning are just beginning to be understood,
and it appears they are incredibly important to brain function.
"Based on what I know right now, emotions would have to be included for a synthetic brain to be able to
learn," Parker said. "It's important to understand their cause and effect."
-- Diane E. Banegas, (703) 292-8070 dbanegas@nsf.gov
Investigators
Alice Parker
Chongwu Zhou
Next
Diahuman AI, according to J. Storrs Hall, is the phase of AI development similar to humans who don't
learn well and who function competently only at simple jobs for which they must be trained.
The core of the diahuman phase will be the development of autogenous learning.
In later phases, AI will be more autonomous and learn what it needs to know and deciding what it needs
to learn.
Will we have inexpensive robots with diahuman intelligence or will AI advance beyond that stage by the
time robots are inexpensive enough to put on production lines where humans now serve?
parahuman AI
Cyhthia Breazeal's sociable robots are forerunners of robots which will interact with humans. The
Japanese have robots which serve as receptionists. They speak many languages, which is a plus over
most human receptionists.
According to Hall, a parahuman AI should act like a lawyer, doctor or accountant with deep knowledge
and endless patience. Once AI and cognitive science is sufficiently advanced AI teachers will be built
that will model in detail how each individual student absorbs material and will optimize the presentation
so as to maximize understanding as well as motivate and stimulate the student.
The downside of parahuman AI is that it can be used to manipulate humans through advertising or being
an individual con man.
http://www.nsf.gov/discoveries/disc_summ.jsp?cntn_id=112947&org=NSF
Synthetic Brains
Researchers study the feasibility of brains made from carbon nanotubes
Researchers are building mathematical models that accurately reflect neuron connections.
January 27, 2009
Synthetic brains are a long way from reality, but researchers at the University of Southern California,
funded by the National Science Foundation, are taking the first steps to build neurons from carbon
nanotubes that emulate human brain function.
"At this point we still don't know if building a synthetic brain is feasible," said Alice Parker, professor of
electrical engineering. "It may take decades to realize anything close to the human brain but emulating
pieces of the brain, such as a synthetic vision system or synthetic cochlea that interface successfully
with a real brain may be available quite soon, and synthetic parts of the brain's cortex within decades."
The challenges to creating a synthetic brain are staggering. Unlike computer software that simulates
brain function, a synthetic brain will include hardware that emulates brain cells, their amazingly complex
connectivity and a concept Parker calls "plasticity," which allows the artificial neurons to learn through
experience and adapt to changes in their environment the way real neurons do.
There is also the matter of scale. By 2022, with conventional technology, if the team could construct a
synthetic brain that emulated real brain function, even crudely, it would take 100 billion artificial neurons
and a very a large room to hold them.
"Obviously the technology will have to be downsized to aid a human being or be feasible as a robot
brain," Parker said. Power is another consideration. The power requirements for a synthetic brain are
staggering because a human brain never turns off. "In a transistor things are on or off so it's a
black-or-white situation, but in the brain there are also many shades of gray and power is continuously
being consumed," Parker noted.
But before the researchers can tackle concerns of power and scale, they are building mathematical
models that accurately reflect the Byzantine connections of all the neurons and demonstrate how the
connections allow neurons to communicate with each other.
Each neuron in the cortex--a part of the brain that contributes significantly to conscious thought and
intelligence--is connected to tens of thousands of other neurons. The researchers are also
implementing the complex computations carried out by each neuron on all the inputs it receives from
other neurons.
"It's a nonlinear phenomenon and almost impossible to model but that's what we're attempting to do,"
Parker said.
The researchers have shown that portions of a neuron can be modeled electronically using carbon
nanotube circuit models and have performed detailed simulations of the circuit models. A single
archetypical neuron, including excitatory and inhibitory synapses, has been modeled electronically and
simulated. Parker and her co-researcher Chongwu Zhou are in the process of combining these circuit
models of neurons to create a functional carbon nanotube circuit model of a small network of neurons.
This small network of interconnected neurons will be simulated using the carbon nanotube models. This
network demonstrates an interesting neural circuit that detects moving edges in a selected direction.
Parker believes carbon nanotubes are an ideal material to emulate brain function because their
three-dimensional structure allows connectivity in all directions on all planes and because a
carbon-based prosthesis is less likely to be rejected by the human body than one made from inorganic
materials. But their invasive nature could result in them invading surrounding tissue and prompting
lesions and cancers.
"It's a possibility and something else that needs to be addressed for the technology to be feasible,"
Parker said.
As the researchers move ahead with their mathematical modeling and neuron construction, beginning
with a single synapse, they ponder "plasticity," neuroscientists' term for the brain's ability to learn and
adapt to change. "Our brains can grow new neurons and the synapses between them in an hour--a
remarkable biological feature that is difficult to emulate from an engineering perspective," Parker said.
Emulating such plasticity in a synthetic brain will require a major leap in technology, similar to the leap
from cathode ray tubes to transistors. "We don't know what the new technology will look like yet, but it
will be a technology that can self-assemble and reshape itself.
As we work in the lab building neurons or constructing mathematical models, we must consider the
requirement of plasticity, even if we don't yet know what it looks like."
Aside from the daunting technological challenges, a synthetic brain or brain components will also raise
ethical and environmental issues. The role of emotions in learning are just beginning to be understood,
and it appears they are incredibly important to brain function.
"Based on what I know right now, emotions would have to be included for a synthetic brain to be able to
learn," Parker said. "It's important to understand their cause and effect."
-- Diane E. Banegas, (703) 292-8070 dbanegas@nsf.gov
Investigators
Alice Parker
Chongwu Zhou
Next
 | |  | |  |