How can we truly state that the functionality and capabilities of A.I. are synonymous with that of Human or Biological intelligences, when we aren’t even entirely sure how the brain really works?
Those were my original thoughts on A.I. and Machine Learning…
That was until I spent a few weeks collaborating with grad students at Georgia Tech’s A.I. and Neuroplasticity project in Atlanta, Georgia.
“What an incredible power: The ability to… grow up” – Rose Quartz (Steven Universe)
Biological vs Artificial Intelligences
Living beings are complex. And while a complete theory on how the brain functions overall does not yet exist, we do have some fundamental ideas. Neurons within our nervous systems send signals and messages throughout the entire body. For most living species these neurons accumulate near the head or eyes resulting from billions of years of evolution.
You know, the basic stuff you learn in high school biology.
Scientists discovered that those pathways and connections made by neurons can be represented by matrices or tensors. Basically… the hippies were right. Our thoughts and consciousness quite literally exist in higher dimensions.
The goal of most researchers in machine learning or “Artificial Intelligence”, like myself, is to replicate these fundamental functions of the brain. Allowing computer algorithms to adapt and perform functions based on the knowledge it acquires or is given by the programmer/trainer.
Similar to the pathways and connections between biological neurons, A.I. uses “artificial neural networks” that carry information or predictions of the most desired outcome specified by the coder. This process involves that same hippy, multidimensional philosophy, married with statistics and higher order calculus.
Until recently, in my own (and arguably pretty popular) opinion, the apparatus of machine learning in and of itself is linear and static. I mean sure, we are only in the lower end of the technological development curve in terms of Artificial Intelligence. Tech is about to go CRAZY. And FAST. But A.I. still relies heavily on statistics and human data. Which is great for basic tasks like facial recognition and doing some lazy college student’s algebra homework.
But…
The word “Intelligence” implies “creativity”.
This is where modern forms of A.I. are lacking… for now.
What is Neuroplasticity?
We’ve all dealt with a tough break up… Or maybe toxic parents. Moving to a new city, learning the language, how to get around, culture, customs etc., Why blind people acquire keen hearing abilities. Walking again after an intensive leg or knee surgery.
These are everyday examples of Neuroplasticity.
The Brain’s unique ability to reshape its neurological structure based on new knowledge or lived experiences.
And its HUGE flex.
A direct link to humanities extreme levels of creativity.
Beethoven composed Fur Elise for the love of his life after she fell in love and married another man. The same reason a traumatic experience can alter a person for the rest of their life is the same reason we humans are able to produce these profound, supernatural strokes of genius, creativity and incredible works of art. Seemingly out of nothing.
Modern A.I. is not equipped with the ability to alter the structures of their own learning algorithms. Relying solely on new data and statistical weights on that data in order to produce different outputs. Less equipped with the ability to apply this knowledge across multiple domains seamlessly.
Like an 80 year old Republican senator, the way of thinking for most artificial intelligences today is static and rigid.
They do not self-evolve.
Progress of Neuroplasticity in A.I.
Conclusions we found at Georgia Institute of Technology:
In the brain, neuronal connections are not random; connections depend on distance.
Geometric graphs can represent more than just physical space; e.g., similarly in characteristics.
Adding Hebbian plasticity to the k-cap process leads to convergence on directed random graphs.
With plasticity, this process exhibits powerful computational properties (Assembly Calculus)
This work corresponds to a low-plasticity setting.
Investigating the effect of degree heterogeneity, existing deterministic convergence on two alternating sets with rates that are unclear.
In layman’s terms:
In computational neuroscience, “network theory” is the idea that the firings of neurons in a “biological brain” are not random. That they follow structured yet complex patterns, mirroring assembly calculus; a branch of mathematics dealing with calculating the intricate firings of neurons in the brain.
In several labs, we experimented with using this branch of mathematics to allow a small set of artificial neurons to emulate neuroplastic firing (the k-cap process) we find in the neuronal connections of biological neurons.
With Hebbian plasticity (multiple artificial neurons firing at once), we found that the network begins to converge into “random” configurations found in biological neuronal connections. Hinting to structured yet adaptable learning behavior.
Citations:
Steven Universe. “Greg the Babysitter.” Season 3, episode 14. Aired September 15, 2016, on Cartoon Network.
Akgün, Ergün & Demir, Metin. (2018). Modeling Course Achievements of Elementary Education Teacher Candidates with Artificial Neural Networks. International Journal of Assessment Tools in Education. 5. 10.21449/ijate.444073.
Reid, Mirabel, and Santosh S. Vempala. 2022. “The $k$-Cap Process on Geometric Random Graphs.” arXiv preprint. https://arxiv.org/abs/2203.12680.
Bullmore, Ed, and Olaf Sporns. 2009. “Complex Brain Networks: Graph Theoretical Analysis of Structural and Functional Systems.” Nature Reviews Neuroscience 10 (3): 186–98. https://doi.org/10.1038/nrn2575.
Papadimitriou, Christos H., Santosh Vempala, David Mitropolsky, Anna Bulatov, and Marina Vardi. 2020. “Brain Computation by Assemblies of Neurons.” Proceedings of the National Academy of Sciences 117 (25): 14464–72. https://doi.org/10.1073/pnas.2001893117.
I am always open to collaboration or getting involved in projects. Submit your name and request below and I’ll be sure to reach you as soon as possible. Or even if you just wanted to check in and say “what’s up?”
