Consciousness , brains and machines

Weekend Blog: #10 : This weekend I delve into something fundamental, a question which has been asked to mystics, but has now become a quintessential question for science ,as we delve more and more in the human brain and see its relevance for AI

Over the centuries, we have been engulfed in trying to control the facets of the physical world we live in. Our scientific pursuits to look at the planet(outside-in and inside-out), the forces of nature , the way things work and the way we make them work , be it a bulb, mobile phones, autonomous cars or rockets as the new space age dawns upon us.

Few people and attempts have tried to look at the most beautiful machine in the universe, one that sits within us aka the human brain. Its complex but its the one that makes us the most dominant species on this planet.

This 1.1 kg mass is unique. It has unique physical structure. It has billions of neurons (the basic building blocks) , ones that are attributed to provide us reasoning. Every neuron is also connected to around 10,000 more using structures like the synapses , making it the most complex jamboree of networks. Neuroscientists have also shown that synapses are the seat of our memory. This complex structure has inspired the growth of artificial neural networks (ANNs) and deep learning, manifestations of which exist today as chatgpt.

While the brain is the hardware, the mind is the software of this machine, one that makes us do what we do. All our beliefs, disbeliefs, fear, trials and tribulations are models that run in the software. The mind is the software that models the world around us and imprints it within the brain for us to live, look at the world and constitute beliefs.

The above is analogous to the deep reinforcement learning(a branch of deep learning which learns via rewards), agent , which models its world to solve a problem, because it sees a reward in the future. The steps that it takes are called Sequential Decision Making, is guided by the ultimate reward in the future. This may include playing a game and maximizing the score you get, ability for us to walk sp that we can put our best next foot forward, other tasks etc. While doing so, it sometimes leaves rewards for the lull of the maximum final reward.

Human mind is the same , The ultimate goal could be attaining nirvana, more money, good life etc. The are the goals we have where we start making mental models of the world around us. We keep updating them as we grow, we keep pushing some to the background. The mind thus is the software that runs on the hardware of the brain. This software comes to life due to myriads of interconnections between neurons, creating experiences via electrical impulses that travel across the body, also creating memory on the way to take actions that affect the future.

We are constantly predicting the future without even realizing it

1. When we take the next step forward, we are predicting a safe landing of the foot to balance the body
2. Which side should I turn the door knob when I reach my home ?
3. What should I do tomorrow? What are some of my pending tasks?
4. How should I connect with colleagues tomorrow? Would they mind if I go to office wearing this ?
5. What would happen if I do 'X' and what may not happen ?

And just as an artificial neural network starts with a seed of random weights in the beginning of a forward propagation to run a uniform approximation mechanism to predict, the initial seed or weights( not random here) for a human brain is the intuitive physics and biology/chemistry fed into us as we are born

1. Consider a new born baby and simulate its fall while keeping them in your hands( This is a thought experiment. Please don't try this. This typically happens when we have a new born in our arms), you would notice that the child flays its hands in air , as if its trying to reach out for something. It seems the concept of Gravity is embedded within babies as a primal intuitive physics.
2. Babies also know they need milk to satiate their desire
3. They smile when they see some faces and are indifferent to others

These mental models are ingrained in babies and these mental models keep getting updated as one grows, or simply when more data is received.

Where does consciousness comes in , in all of this ? I know physicists may define it differently , a computer scientists' view point is that consciousness, is that model of the world which is internal. In this model, the mind has placed us in it as a simulation agent. When the external model changes, this model tries to internalize and puts itself back again in that situation to make a decision

A computer scientists' distinction from a physicists is the way we consider these models of the software which is the mind. There is an external model and an internal model. An external model is the model of things we make to realize how the world works . How do you cycle? What happens when temperature rises. The consciousness is used to position and simulate ourselves within that model.

At Maker's Lab, and generally my research has been virtually along these lines. I took the route of a language and how it could explain features to machines , but it seems I was seeing the world narrowly. Its the multi modal approach needed to understand the phenomena and catapult it within AI systems.

As a thought experiment , and my weekend project, I am trying to model a game world, a world that can be divided into bocks and levels. Blocks have an agent resident in it. The block is self sufficient and each one has been provided with some charactersitics that the agent interacts with. Our methodology is to see how agent evolves as it ages. The age is defined by a color code. As the agent interacts with the things/situations provided in each block, we observe how it evolves. We divide this into a test group and a control group. We are trying to see how humans and agents can work together to create a harmonious outcome

This is not a simple problem to solve. This would need me not just ANNs, but also symbolic reasoning, Bayesian approaches for probability and probabilistic programming as well . Doomsday theorists can call this the beginning of the end of the world, I would say, its just the beginning and there wont be an end.

The part of the brain we are trying to model is the neocortical region which is the seat of intelligence. Flight or fright mechanism which resides in Amygdala is not even remotely considered here. We do hope we should be able to produce some results.

This quest of mine also needs partners and like minded individuals. so do reach out to me if you are interested to work in this topic.

For my previous blog on brains and beyond neural networks, do read https://www.dhirubhai.net/pulse/beyond-neural-networks-nikhil-malhotra/