Imagination v/s AI
The creativity of the brain is powered by AI.
Artificial Intelligence has always been a game changer. Alan Turing is responsible for the birth of AI. The first AI program was a Chess-playing program. This all started with a simple coding program and eventually, the first human AI, Sophia, came into existence. A recent breakthrough happened last year on November 30, 2022, in which an OpenAI named ChatGPT steered the internet.
But wait! All these advancements have a single source behind them- Human Imagination.
I’m here today to talk to you about one of the questions that has consistently fascinated me and a question which helped me write about Artificial intelligence.
That question is: What fuels our Imagination?
We are aware that our imagination is critical to our intelligence, it’s critical to our ability to plan, create and innovate. Well, understanding how our brain works and build an artificial brain or AI program will necessarily require an understanding of imagination. Right!
But to begin with and to be able to explain a little bit more, I want to start with the following observation, and it is that imagination actually requires a connection to the real world. After all, the writers get inspiration from real events many times.
But if your thoughts are completely untethered from reality, if they’re nothing more than noise, it’s just meaningless. To put it simply, true imagination rests on a foundation of realism.
Now, the question that actually needs an answer is how does your brain develop an understanding of reality that can underpin your imagination? And that’s what I’m going to tell you about today. Now you’ve already heard a lot about this brain structure: The Hippocampus. It’s a critical brain structure for our memories and interestingly, the evidence suggests that it’s critical for our imagination, and in particular we think that it’s critical for providing the anchor of our imagination to reality.
And it’s a brain structure that is hit during Alzheimer’s disease very hard and that displays to you just how important this brain region is for our cognition, given how devastating Alzheimer’s is.
Now, as also noted the hippocampus is critical for your ability to store those types of memories-namely Episodic Memories that is personal experience memories. Your memory of your first day at school, your memory of your wedding or even just your memory of what you had for breakfast yesterday. So episodic memories in the hippocampus are organized according to a very particular structure in the activity of your brain, a structure that makes reference to reality and which allows you to remember not only what happened but where and when it happened.
This structure is called a cognitive map and roughly speaking a cognitive map is a model of the world in your brain. Evidence shows that the hippocampus is critical for your ability to engage in imagination and certain data also suggests that in particular, it’s the cognitive map in your hippocampus that tethers your imagination to reality.
The question I’m going to turn to now is how does your hippocampus learn to form a cognitive map based on your experiences? Now one way to tackle this question is using a mathematical tool called Artificial Neural Networks.
Artificial neural networks are abstract mathematical simulations of brains. We treat all the neurons in the brain as a series of numbers corresponding loosely to the electrical activity of those cells. That electrical activity changes over time according to the connections between those simulated neurons.
Now interestingly artificial neural networks are also how we build modern artificial intelligence.
All the artificial intelligence systems you’ve ever interacted with in recent years, whether it’s a voice recognition system, an image generation system or a chat bot like Chat GPT are based on artificial neural networks.
Now the process of how it works?
They train a simulated agent on artificial intelligence that possesses an artificial neural network and what they train the agent to do is they train it to predict what it’s going to see next as it wanders through a simulated maze.
Hence, findings suggest that you may learn your cognitive map by trying to predict the future and this actually fits with a lot of Neuroscience data that shows that the hippocampus is constantly making predictions about the next thing that’s going to happen to you.
Fascinatingly this is also how modern AI systems like Chat GPT are trained. These systems are at their core simply trained to predict the next thing that they will encounter such as the next word in a sentence that they’re processing and from this simple form of predictive learning emerges all of the capabilities that we now see in modern AI systems, including writing an essay, explaining a joke or just engaging in conversation. All of this suggests that imagination may not be uniquely human, it may in fact be something that emerges in any agent that’s trying to predict the future.
Because doing so will endow that agent with a cognitive map and will provide the tether to reality required to imagine potential future scenarios.
Put another way in learning to predict the future you learn how to think outside the box but just a little bit so that your thoughts remain anchored to the world you live in to some degree.
To summarize- Imagination likely rests in part on cognitive maps created when you learn to predict the world around you. With AI models we can see what happens right before our eyes and we can watch them imagine and explore the geometric structures in their simulated brains with ease. As these models improve, their “imaginations” will become more sophisticated and this may end up helping us to understand more about our own imaginations.