5

I am a software engineer with a special interest in machine/deep learning and I have a fair amount of experience in this space. In (supervised) machine learning, you are tuning a (very) large data structure called a neural network by subjecting it over and over again with pictures/sounds/inputs that contain a particular "thing" you want it to learn and letting it analyze the characteristics of each input. Every time you show it a new input that contains this "thing" it tweaks the configuration of the network a bit more, and after many (hundreds; thousands; tens of thousands) of examples, this network converges ("learns") into a final state that is capable of recognizing the "thing" on new inputs that it has not yet seen.

I'm trying to do some research in some new areas of machine learning and I'm interested in what some of the predominant modern theories are on how the brain "works" as a possible source of inspiration. It is my belief that humans have inadvertently been building computers and software in a way that mimics the human body, probably because nature perfected the recipe a long time ago, and its probably the best recipe! So by having some decent understanding of what our top dogs in neuroscience believe with respect to have the brain "works", I might be able to think outside the box here and solve some problems that I wouldn't have thought of otherwise.

By "how the brain works" I mean:

  • Theories on how the brain converts inputs (vision, sound, smells, tactile sensations/pressure/heat/pain, tastes, etc.) into a form that it can then process and make informed decisions with
  • Theories on how the brain actually makes these decisions; how we think and uses stored information to decide to do something
  • Theories on how the brain stores and fetches memories and information (how I can recall my social security number, name or birth date, or how I know what a cat looks like; as well as how I store new info like remembering a new co-worker's name)
  • Theories on why our brains are all so different and people have different likes/interests/aversions than others (why I like writing software and hiking and other people like skydiving or eating spicy food, etc.)

Again I'm interested in what the modern and "predominant" theories are (by that I mean: gaining lots of traction/legitimacy within the appropriate scientific communities) for all these things. Knowing what those are will help me sift through the ocean of blogs, white papers and articles that might very well mislead me or take me down roads on theories that were relevant 20 years ago but have long-since been abandoned or replaced, etc. Thanks in advance for any-and-all steering here!

hotmeatballsoup
  • 151
  • 2
  • 2
    I think this question is too broad to be a good fit for the Stack Exchange format. You're asking for an introduction to entire fields (several of them) of research that exist at many scales of analysis and with very different goals. Even just introducing* these areas as you ask for is effectively a textbook-scope project*. I would say though that modern AI has long ago given up on overly biology-based methodology - if your goals are in AI/ML, stay in those fields. – Bryan Krause Dec 24 '20 at 01:56
  • Thanks @BryanKrause (+1), I did find one recent theory called the Hierarchical Mechanist Mind, are you familiar with it? It seems to address most of these and seems to be gaining traction in the neuroscience community. – hotmeatballsoup Dec 24 '20 at 02:20
  • Looks like a new name for Friston's free energy minimization paradigm. Friston's theoretical papers can give a great "bird's eye" view but as you dig into them the mechanisms proposed are surprisingly superficial and don't break much ground. – Bryan Krause Dec 24 '20 at 02:28
  • Yeah I'm kinda of getting that feeling too. I'm reading through the PDF and I'm seeing a lot of fancy words but not a lot of "concrete theory" (i.e. we believe X is what is happening, and here's our evidence...). Any chance you'd mind pointing me in the somewhat-right direction here, in your opinion? If you were me, what theory(ies) would you start with? Which ones do you think actually do break ground and give insight into how these big water bags on top of our bodies work? – hotmeatballsoup Dec 24 '20 at 02:37
  • I don't think it's a thing that can be done in a paper, otherwise I would have cited it and written an answer :) You rather have to study different sensory systems and motor systems independently, consider different species that are better for different questions, combine experimental modalities, and so on, and even then you'll only scratch the surface. – Bryan Krause Dec 24 '20 at 03:32
  • Seems taking a course might help. –  Dec 24 '20 at 17:26
  • Related: https://psychology.stackexchange.com/q/19983/7604 – Chris Rogers Dec 25 '20 at 09:57

0 Answers0