biological networks

Fun Fact 110

Bhutan is well-known for its proactive conservation initiatives. More than 40% of the country is made up of national parks, reserves and protected areas and another 9% of the country has been dedicated to creating a network of biological corridors to link all its protected areas. Around 56% of its population is involved in conservation, forestry or agriculture.

Neural Networks

My wish is to show people the way to control fear. You control fear by learning the technique called self-observation. Once we begin to observe our thoughts we realize that most of what we think during our day is nonsense. Useless mental chatter. That is not surprising. You see the brain does not work in a linear manner like a computer. 

The brain operates on a biological neural network. The brain works incredibly fast and it spews out a flood of “thoughts” based on external stimuli. Some of them are quite useless and the brain filters these out. Many of these thoughts are fear inducing because the brain is trying to keep us alive. We sort through these rapid fire thoughts and select the ones which ring true. 

Often, however, we attach importance to the wrong thoughts. Mental discipline will allow us to better reject fear inducing thoughts which are not grounded in reality. We become clearer. Calmer. More rational.

Our neural net allows us to be creative by generating ideas which contain a degree of randomness to them. It is similar to the way random mutation adds variety to the evolution of organisms.

3

In Big Step for Artificial Intelligence, Machine Learns To Master Video Games

by Michael Keller

Competitive gamers beware: There’s a new top dog in the classic arcade category. This champion has cracked 49 vintage Atari 2600 titles, from Breakout to Star Gunner and Space Invaders, outperforming professional game testers by more than 1,000 percent in some cases. Success didn’t come easy. Improvements happened one attempt at a time through an intense period of training, which included playing and then studying each frame of every game millions of times without a break. 

It’s understandable if this newly minted master’s name isn’t familiar to the millions who play Call of Duty or GTA every day, because its creators have been keeping it under wraps while they improved its algorithms. The name is deep-Q network, but you can call it DQN.

Its developers at a Google enterprise called Deepmind say the system is capable of quickly learning how to excel at games even though it starts with minimal background information. DQN represents a significant advance in artificial intelligence, combining machine learning and the principles of neuroscience to make a computer program learn like animals do.   

“This work is the first time anyone has built a single general learning system that can learn directly from experience to master a wide range of challenging tasks–in this case a set of Atari games–and perform at or better than human level on those games,” says Demis Hassabis, an AI researcher and neuroscientist. His team’s work was published today in the journal Nature.

Keep reading

Researchers create organic nanowire synaptic transistors that emulate the working principles of biological synapses

A team of researchers with the Pohang University of Science and Technology in Korea has created organic nanowire synaptic transistors that emulate the working principles of biological synapses. As they describe in their paper published in the journal Science Advances, the artificial synapses they have created use much smaller amounts of power than other devices developed thus far and rival that of their biological counterparts.          

Scientists are taking multiple paths towards building next generation computers—some are fixated on finding a material to replace silicon, others are working towards building a quantum machine, while still others are busy trying to build something much more like the human mind. A hybrid system of sorts that has organic artificial parts meant to mimic those found in the brain. In this new effort, the team in Korea has reached a new milestone in creating an artificial synapse—one that has very nearly the same power requirements as those inside our skulls.

Up till now, artificial synapses have consumed far more power than human synapses, which researchers have calculated is on the order of 10 femtojoules each time a single one fires. The new synapse created by the team requires just 1.23 femtojoules per event—far lower than anything achieved thus far, and on par with their natural rival. Though it might seem the artificial creations are using less power, they do not perform the same functions just yet, so natural biology is still ahead. Plus there is the issue of transferring information from one neuron to another. The “wires” used by the human body are still much thinner than the metal kind still being used by scientists—still, researchers are gaining.

As part of this latest effort, the team placed 144 of their artificial synapses on a 4 inch wafer and connected them together in a two dimensional mesh with wires that were just 200 to 300 nanometers on average. The idea was to test the possibility of causing the synapses to fire (open or close) based on information coming from a wire, or being sent from other artificial neurons. Each synapse mimicked the natural kind in shape as well—they were long and thin and were made of two types of organic material that allowed for holding or releasing ions.

The new artificial synapses are one more step on the road towards a computer that works in ways very similar to the human brain, and most believe if we ever get there, the machines we create will be far more powerful than anything nature has ever produced.

Image: Schematic of biological neuronal network and an ONW ST that emulates a biological synapse.

Credit: Science Advances (2016)

How Termites Build Complex Homes Without a Master Plan

by Charles Q. Choi, Inside Science

Termites are tiny insects, but they are capable of moving tons of soil to build giant nests. Now scientists are discovering simple rules these insect architects might follow that could help explain how they build complex homes without a master plan.

Such research could lead to robot swarms that can organize to assemble intricate structures. These findings could also help decipher the rules governing complex systems ranging from blood vessels to neural networks.

Keep reading

Insect Nervous System Copied To Boost Computing Power

by Charles Q. Choi

Brains are the most powerful computers known. Now microchips built to mimic insects’ nervous systems have been shown to successfully tackle technical computing problems like object recognition and data mining, researchers say.

Attempts to recreate how the brain works are nothing new. Computing principles underlying how the organ operates have inspired computer programs known as neural networks, which have been used for decades to analyze data. The artificial neurons that make up these programs imitate the brain’s neurons, with each one capable of sending, receiving and processing information.

However, real biological neural networks rely on electrical impulses known as spikes. Simulating networks of spiking neurons with software is computationally intensive, setting limits on how long these simulations can run and how large they can get.

Keep reading