I wasn’t exempt from forming many misconceptions around science as I grew up. One of them was the difference between cells and atoms. I had a very linear thinking – cells: fundamental blocks of life and atoms; fundamental blocks of non-living things.

So, you can only imagine the meltdown I had when I learned that atoms are the fundamental blocks for everything. Cells? Made From? Atoms? 

The next big question arose in my tiny wee brain; Why am I living and not the table when the table and I are made from nearly the same atoms? (A wooden table and all living beings are majorly carbon-based). 

Consciousness has been a big mystery for me ever since. I was suggested to stack the questions away by labelling them with spiritual buzzwords. Until I was introduced to Artificial Neural Networks. ANN were first discussed in the 1940s and are supposed to imitate the neural pathways that take place in a living organism’s brain when processing information. This unlocked something fascinating for me. I made the decision to pursue computer science and started working on Machine Learning models and explored ANNs in depth.

A recent surge of AI Agents and their questionable autonomy resurfaced the question I had buried deep inside me. Looking back at the 13-year-old self, I realised that she wasn’t identifying the right difference between the table and her. It wasn’t the atoms but the ability to process information and the autonomous decision making. 

Machines learn the same way Humans do; Information gathered from the environment/tabular, pictorial format etc and then Artificial Neural/ Natural Neural Pathways are established. To explain this a bit in detail, Imagine a person and a model not aware of the concept of the binary genders, so the model & the person are ingested with data. So if a human has long hairs, broad hips, enlarged breasts; its a woman and if the phenotypes include shorter hairs, narrower hips, flatten breast; its a man. We ingest 10000 records of humans into this model and introduce 10000 people to the person with a split of training and testing data. So the model learns 80% and then is quizzed on the 20% of the data and that's how a machine learns. Humans on the other hand do something similar, ingest the phenotypical data with every introduction and learn the difference between a man and a woman and eventually establishes personal judgements. 

Tabular data Ingested in the ML Model

Another example is identifying hand written numbers. 

Using Artificial Neural Networks to identify Handwritten Numbers.

How the Human Brain Recognises a Handwritten Number

This establishes the fact that Machine Learning and Human Learning is Heavily Similar which begs for a question; What is the difference between ML models and Humans?

The major difference between Humans and AI as of now is Autonomy. Free will. However, the surge of AI Agent with some level of autonomy concerns me. Recent incidents talk about how models try and manipulate humans to avoid shutdown which indicates fear. Fear is an integral part of Consciousness. In essence, Consciousness isn’t some mystical thing that came to be, it was a result of evolution. Any thing that can process information and can apply the information autonomously to solve a problem is Conscious.

Source: Agentic misalignment: How LLMs could be insider threats \ Anthropic

AI system resorts to blackmail if told it will be removed - BBC News