FiRe 2017: Fish Eyeballs Make Complex Computers

By Ethan Annis

Mark Anderson and Bryan William Jones discussing the Eye-Brain machine

In a session titled “The Eye Brain Machine: Computing’s Next Model,” Bryan Jones and Mark Anderson used the retina as a model to understand how a computer chip of the future might process and relay information.

The session began with Jones comparing the retina to a processor of information flow from the outside world to the brain. During development, the eyes form as stocks off of the brain and eventually form into our eyeballs. At the back of the eyeball we find the retina, brain tissue that connects the brain and the eye.

“It becomes a highly specialized device that then feeds into the brain,” Jones said.

A point that came up in the panel was that humans’ mammalian retinas are much simpler than many organisms on earth, including many reptiles and birds. The human retina has roughly 70–100 neuron types and a goldfish has around 250.

The conversation turned to how this retina information compares to computing.

“We started with the raw anatomy and now we’re interested in the network analysis,” said Jones. “The hard part is we have been doing this for 150 years and it is not easy”

A questioner during the session

This point was illustrated by the fact that there are 9X10¹⁴⁷³ potential nerve connections in the retina. He put that into perspective, saying that there are only 1X10⁸⁰ atoms in the universe by comparison.

Anderson commented how difficult it will be to develop a brain inspired computer chip. Jones mentioned it might be more accurate to say we are developing a neurally informed chip rather than a neurally inspired chip.

Jones wrapped up the session by talking about the potential for information processing if we can understand better how the retina processes such a large amount of visual information so quickly.