Neurons in a Dish Learn to Play Pong

Neurons in a Dish Learn to Play Pong

Hundreds of thousands of neurons have been trained to play a version Pong of the classic computer game Pong .

In doing this, the cells join an increasing pantheon Pong gamers, including pigs who can control joysticks using their noses and monkeys wired with their minds to control the game. (Google’s DeepMind artificial-intelligence (AI) algorithms mastered Pong many years ago and have moved on to more-sophisticated computer games such as StarCraft II. )

Gamer cells do not respond to visual cues from a screen, but to electrical signals from the dish’s electrodes. These electrodes stimulate the cells and record changes to neuronal activity. The stimulation signals and cellular responses were then converted by researchers into a visual representation of the . game. The results are reported in Neuron .

Intelligence in a dish

The work is a proof-of-principle that neurons in a dish are capable of learning and displaying basic signs of intelligence, according to Brett Kagan, chief scientist at Cortical Labs, Melbourne, Australia. He says that neurons are often viewed primarily in terms of their implications on animal or human biology. “They’re not thought about as an information processor, but a neuron is this amazing system that can process information in real time with very low power consumption.”

Although the company calls their system DishBrain the neurons are far from a brain and show no signs or consciousness, Kagan said. Kagan defines intelligence as “the ability to collect information and apply it in an adaptive manner in a given environment.” This is a hotly debated topic.

Cortical Labs is a continuation of work by Steve Potter (now at the Georgia Institute of Technology, Atlanta) and his colleagues. In 2008, the team reported that neurons cultured from rats can exhibit learning and goal-directed behaviour.

The work of Cortical Labs brings more sophisticated technology and analytical tools to bear. Potter says His original dishes contained dozens of electrodes, but each DishBrain contains thousands. Potter’s group focused on rodent cells; however, the DishBrain team tested neurons derived directly from human cells.

The researchers used their system in order to teach neurons to respond to an electric signal that is a replacement for the ball in Pong .. To intercept a ball, the player moves a vertical paddle across the screen. The paddle was controlled by neurons in the experiment.

The authors represented the ball’s route by stimulating neurons along its path relative to the paddle. The paddle was moved up or down by neurons in another part of the network.

To teach neurons how to hit the ball, Kagan and his team used the theory that neurons are prone to repeat activity in a predictable environment. The neurons that responded in a manner that corresponded to hitting the ball were stimulated at the same frequency and in the same location. If they missed the ball the electrodes stimulated the network in random locations and at different frequencies. The neurons learned to hit the ball to get the patterned response over time.

Not just

This work is an important step in developing assays that could, for example, be used to test the potential effects of a new drug on neuronal functions, says Takuya, a neuroscientist at the RIKEN Centre for Brain Science, Saitama, Japan. He said that it was not clear if the neurons behaved as they did to create a predictable environment or in response to other aspects of the signals they received. He says, “I believe the next step is to explain in detail what kind of stimuli can make that difference.”

Cortical Labs aims to eventually use neurons in order to develop “biological processor units” for computing. Potter says that DishBrain’s techniques are quantitative enough to allow for comparisons between learning patterns in different animals or between brain cells from different regions.

In the meantime, he said that the decision to make DishBrain activity the game Pong was a great move. “People are very interested in AI and want to play Pong Potter. “That was a brillian

Read More