Neuron Lithography Technique

An Anonymous Coward writes: "EE Times has an article about a new technique to build custom-designed networks from biological neurons using chip lithography and polymers to steer the growth of the neurons . Some of the first computers were described as "electronic brains" to the unwashed masses - will researchers have to describe these as "biological computers"?"

Comments

The strength of neurons

[Glowing Fish]

While this is an interesting development, and I can't begin to guess what is the future possibilities of it, artificially causing neurons to grow rules out of one of their main strengths.

Neurons get to make their own decision on how to grow, taking into account factors such as present of growth inducing hormones, and how much a connection a neuron makes is used. But still, to a great extant, neurons get to make their own decisions about how much and in what direction they get to grow.

If you are directing neurons into what direction they are growing totally, then what you have is a really squishy computer circuit.

Re:The strength of neurons

[dragons_flight]

Neurons get to make their own decision on how to grow, taking into account factors such as present of growth inducing hormones, and how much a connection a neuron makes is used. But still, to a great extant, neurons get to make their own decisions about how much and in what direction they get to grow.

It's not that simple. Basic nueroscience (of the poke and see what happens variety) recognizes that the brain is composed of many massively interconnected functional groups. In everyone the centers for speech, math, motor skills, etc. are roughly located in the same places. (Yes nature can compensate for damage in some cases, but retraining parts to do other work is slower and often less effective than the original)

People and animals are genetically coded to design brains in certain ways. Merely having lots of nuerons doesn't guarantee intelligence or functionality. A lot of it has to do with where inputs come in and where the outputs go out and how the groups are connected along the way. Also there are different types of nuerons with different nuerotransmitters and degrees of interconnectedness.

That said, yes the nuerons to some degree govern themselves. An architechture is built up and then nuerons respond in complicated and individual ways to some, as yet poorly understood, system for learning and development. If the brain really is all there is to intelligence than memory and learning have to be a product of something the nuerons are doing. Unless there is some uber mechanism directing all the nuerons, then learning has to be a natural result of what nuerons. Crudely put this might be divided into two categories (as we understand it today):

1. Nuerons like to fire in the same patterns they've seen before.
2. Nuerons like to make new connections.

Thoughts, especially memories, aren't random, they are similar to thoughts that have occured before. Roughly speaking it appears that the brain likes doing things it has done before, and thus learning. One way this is accomplished is by strengthening connections between nerves that have fired together in the past and weakening ones that don't often fire together.

The second thing is that nerves do like to grow. Not so much that it makes the brain random or chaotic, but enough to allow new patterns to be formed and improve on existing ones (for instance shortening the number of nuerons a common path goes through).

No one really understands how it all interacts, or how the features of nuerons relate to our preferences for certain outcomes over others (e.g. what in the brains causes pleasures to be reinforced and painful experiences to be avoided). This is however a good first step at being able to study nueronal circuitry in a highly controlled way.

Besides if you really expect functional "squishy" computers than something has to provide the initial framework that genetics and evolution has arranged in the animal kingdom. Build some nueron groups in meaningful ways, provide some mechanism for learning in an input/output environment (perhaps similar to how people try to train nueral computer networks), and then remove the restrictions on growth and connectedness and let the structures optimize and develop themselves.

Re:The strength of neurons

[plastik55]

IAANS (I am a neuroscientist.)

I think the best immediate application for this kind of technology is not in constructign useful biological circuits, but in doing basic research on the properties of the nerve cells themselves. Traditionally, most of our knowledge about how nerve cells operate has come from studies of single cells in isolation, and in vitro recording. Neither of these techniques give very much information about the mechanisms which govern interaction between neurons, especially in the ways that their growth and behavior is influenced by neighboring cells. A technique which allows us to control the growth of a nerve culture would be a great tool for studying those interactions, becaue you would be able to more tightly control the interactions present (as opposed to blindly jamming electrodes into brains and trying to infer circuits and connections from correlations in firing patterns, which is more or less the standard technique.)

Koch in Biophysics of Computation showed that given what we know about neurons, they can theoretically implement equivalents of addition, multiplication, feedback, and many other computations, all within a single cell! The range of possibilities is enourmous, and requires a controlled environment to study the different tyeps of interactions. This could be a very important tool for research.

New trend?

[gad_zuki!]

With this recent salon piece on how starfish parts could be used in computing and the neural net made out of neurons I think this could be a very interesting trend.

The site is already slashdotted, man its 4am here, but if nature has already done a good job in design why not borrow it for other purposes. Its like Junkyard Wars but on a very small and living scale.

A nice bonus with using cellurar material is there probably wont be ethical complaints i.e. who identifies with neurons?

URL of Research Group

[favalora]

The EE Times article mentioned that a lot of the work is coming from Bruce Wheeler's research group. This is the home page:

http://soma.npa.uiuc.edu/labs/wheeler/home.html

And click on "featured work."

Also, if you're an electrical engineer you might be interested in "neuromorphic engineering," in which circuitry is designed with biological inspiration. A few places to check out are:

Caltech Center for Neuromorphic Systems Engineering

Telluride Workshop on Neuromorphic Engineering

And this Introduction to Computational Neuroscience

- Gregg Favalora -

Ref to actual paper?

[Alik]

Out of curiousity, any of the readers have a reference to the actual journal article? (I'm assuming they published this *somewhere* and didn't just send out a press release.) I'm aware of previous results with neurons and polylysine, so I'd like to take a look and see precisely what the innovation is here. (My guess is that it's the microprinting.)