Seminar 1: Larry Abbott - Mind in the Fly Brain

MIT RES.9-003 Brains, Minds and Machines Summer Course, Summer 2015
Instructor: Larry Abbott

Modeling neural processes in the mushroom bodies of the insect brain to understand how flies learn to recognize scents. Training flies with classical conditioning with positive and negative reinforcement to study olfactory memory and behavior.

License: Creative Commons BY-NC-SA

[greencoder]

*Personal Notes*
[01:07] «The mushroom body is the soul of the fly. [...] Flies are not as intelligent as you, but [...] if you have free will, they have free will.»
- Mushroom body gives rise to intelligent / individually learned behavior of flies
[03:44] Anatomy / Antennal Lobe

[05:36] Neuronal Schematic (Olactory Receptors -> Projection Neurons)
- olfactory receptor neurons (1000x, 50 types)
- projection neurons / glomeruli
-

[14:40]
«So you've gone from stereotypy at the input stage,
[over] a wild and crazy random thing in the middle,
and then back to stereotypic to get to the output.»

[15:46] The mushroom body (~2000 Kenyon cells)

[21:21] The Connectome / Connection Matrix
«Connectomes don't come labeled.»
«Why seven connections? Seven seems awfully small to us cortico-centric people.»

[26:16] Output Neurons


[28:41] Neuronal Schematic (Mushroom Body Lobes -> Output Neurons)

[30:23] Compartmentalization allows for precise modulation by dopamine neurons (PPL1/PAM)

General Fly Architechture:
- Feed Forward Loop
- «Regions that receive output from the mushroom body also provide input to the dopamine neurons»
- «When the mushroom body acts, the dopamine neurons know about it.
And when the dopamine neurons react, the mushroom body knows about it.»

The Dopamine System
- is a reporter of behavior
- is a reporter of internal state

[34:17] Classical Conditioning

[39:35] Signal Depression upon Repetition
- Mediated by Dopamine at the output of the mushroom body
- Allows for immediate detection of unexpected/new odors

Routing / Gating of Signals by Dopamine
- Fed fly will ignore odors to which it is attracted when hungry
- Hungry fly detecting fruit odors transforms CO2 avoidance signals into attraction

Summary for Machine Learning Engineers
- One hidden layer
- Layered Output
- Intelligence resides within the complicated modulation and plasticity.
- Short-Term Memory, Long-Term Memory, Gating of Learning, Gating of Gating

[derpnerpwerp]

Sucks that we cant see the laser pointer

[glovester]

Thank you for posting. Very interesting.

[badfriday]

Evolution huh ? I see some super intelligent design here !