When Biologists Play Atari

Fascinating research study which attempted to apply the same neuroscience techniques and tests that we use currently to study the brain, and used them to try and study a classic in processor history the 6502, used notably in the Atari video game systems as well as the Apple I and Commodore 64. They chose the Atari processor due to the fact that is functions are incredibly well understood and basic enough that we have a very thorough and deep understanding of exactly what is happening in those processors to make them work. Additionally, since so many of the researchers grew up playing Atari, there was some great sentimental value to the work as well.

"To an extent, the fact that their simulator worked is a validation of the approach. But, at the same time, the chip is incredibly simple: there is only one type of transistor, as opposed to the countless number of specialized cells in the brain. And the algorithms used to analyze the connections only got the team so far; lots of human intervention was required as well. "Even with the whole-brain connectome," Jonas and Kording conclude, "extracting hierarchical organization and understanding the nature of the underlying computation is incredibly difficult."
They then used the simulator to try out a variety of approaches that have been used in neurobiology. The first is termed a lesion analysis, where they disable individual transistors and see what happens. While this was great for understanding which transistors were essential for which game, it actually didn't tell them much at all about how the processor operated. And, in fact, the results were largely artifacts. Even though they could identify transistors that were essential for one game or another, 'a given transistor is obviously not specialized for Donkey Kongor Space Invaders.'"

They found using these techniques lead to some very interesting results and absolutely none of data is relevant in actually getting close to understanding the functional processes occurring. They use this to make the argument that even though we are attempting to document when each neuron fires and which other neurons it communicates with, we simply do not have a thorough enough understanding of how the brain works to even begin to make use of that data and get closer to actually understanding how the brain functions.

While it's certainly a valid point that these techniques were developed to study the biological processes in the body and that a silicon processor simply does not work the same way, I find this research incredibly compelling. The steps we are doing to document how the neurons interrelate in the brain are absolutely important, but I think we are fooling ourselves if we think that this will do much more than raise more questions.