it's strange how we seem to know everything and nothing about our brain we know it's billions of neurons connected together forming trillions of electrical circuits endlessly firing in patterns that somehow Define us but how exactly we don't know people have tried to explain the brain by breaking it into different regions performing different functions memory here planning here perception here but recent Neuroscience has revealed that brain function don't exist in separate regions but as an interplay of many regions and this is too complex to reverse engineer because the brain was not engineered the way a human
would have made it the brain evolved to create what we call intelligence the modern boom and artificial intelligence has forced us to confront the question what makes human intelligence so special a recent book A Brief History of intelligence takes an evolutionary view based on the most modern research it follows the evolution of intelligence through five key leaps in learning ability so let's go back in time to the beginning before brains there was a simpler brain fabric called a nerve net which we see in the first multicellular animals it allowed them to both sense and react
to their world enabling simple Behavior such as grasping at the presence of food this set sets up our first breakthrough which is the ability to navigate towards food there's a significant Milestone with the emergence of our ancestors that resembled a tiny worm with a simple brain consisting of only 302 neurons and they had a primitive smell defined by small clusters of neurons within the brain each able to detect different chemicals in the environment and their brains applied a simple rule for finding food if the con concentration of a good thing increases keep going forward if
it decreases then change direction however these animals couldn't learn new Behavior within their lifetimes their brain circuits were genetically encoded and fixed in life this is a limitation that life forms would overcome with a new learning mechanism based on experience this leads to our next breakthrough learning from trial and error we see this evolutionary leap during the Cambrian explosion around 5 140 million years ago this was a period of time marked by an explosion of new life forms including another distant ancestor resembling A Primitive fish they evolve new neural circuits capable of adjusting Behavior based
on outcomes but in order to learn from experience you must have both a learnable sense and learnable behavior so let's start with sense these animals developed a new brain layer that acts as an interface with the world known as the cortex it receives input from all sensory organs Eyes Ears smell touch and processes these inputs into patterns of neuron activations within its deeper layers this enables the cortex to associate patterns with specific things in the world such as new sources of food or danger unlike what we saw before where one neuron corresponds respond to one
specific thing this evolutionary leap allows a pattern of neuron activations to represent different things in the world this is what allows fish to recognize new objects so effectively including complex patterns they would not have evolved to recognize such as human faces to demonstrate this power researchers didn't experiment where they taught fish to spit water at pictures of specific human faces to get rewards and they were quite good at this even if the faces were rotated so now let's turn to how Behavior can be learned alongside [Music] sense to control Behavior the neurons which represent Perceptions
in the cortex connect to newly evolved dopamine neurons now here's the key whenever dopamine signals increase it makes the most recent Behavior that the animal did more likely to occur and when dopamine decreases it inhibits the most recent Behavior a well-known experiment was led by Schultz he showed the monkeys different cues such as pictures of shapes 5 seconds before presenting the reward of sugar water first they observe that the monkeyy dopamine neurons fired in response to newly presented unexpected sugar water but not when the shape was presented prior after a few repeated trials they noticed
the dopamine neurons began firing immediately after the shape was presented and not when the sugar water was given so the dopamine response shifted to the earliest known expectation of the reward this made it clear that dopamine was not a signal for pleasure but an increase in the amount of expected future pleasure or reward but as we all know unlearning is just as important as learning which dopamine also facilitates for example they did more trials where they showed the shape but never presented the reward in these instances the dopamine neurons showed a decrease in activity exactly
when the reward previously occurred this reduction in the dopamine firing rate can be interpreted as a form of negative feedback what we'd feel as dissatisfaction or disappointment and this ability to perform arbitrary sequence of actions based on rewards allowed animals to better learn from their environment how to exploit rewards and avoid pain you can think of this process as intuition building however this form of learning was limited to the experience of actual rewards and pain this leads us to our next breakthrough simulation now let's fast forward 4 100 million years to a world ruled by
dinosaurs alongside them was another ancestor think of it as a primitive rodent while much smaller and weaker these animals evolved something much more powerful than simple trial and error learning they grew as simple imagination something dinosaurs didn't have this required a new brain region called the neocortex the neocortex is amazing in that it can imagine possible f Futures like mental movies which activate the brain in the same way as if we actually experienced it and critically those dopamine circuits can be triggered from these simulations in the same way as direct experience leveraging existing intuition circuits
but applied to imagined actions and this allows animals to learn from mental trial and error for example example it was known for a long time that when rats were put in mazes they would often pause when they had a choice to make and look back and forth pondering which way to go it was only recently that David reddish LED an experiment where they recorded neuron activity in rats while searching through mazes they found that when a rat stopped at a decision point and turned its head back and forth its brain was actually simulating going down
each pathway the exact same neur patterns were activated in the rat's mind when it walked down each path and when it imagined walking down it and what you can see is a sequence running from the rat to the goal this ability to run mental simulations from your current internal State also improves complex motor skills compare this to animals which don't have a neocortex such as lizards since they don't see their next step before they take it however these early mammals the simulation ability was still limited compared to our own these mental movies were always from
their firstperson perspective and based on their direct experience which leads to our next breakthrough now we can fast forward to around 15 million years ago and our ancestors had evolved into the first primates While most mammals live in solitary or small family- based groups primates such as chimpanzees live in tribes with Dynamic social hierarchies using empty kerosene cans to make his display more impressive Mike Rose from very low status to the position of top ranking or alpha male successful chimps must have the ability to know what others are thinking to do that they must be
able to break out of their firstperson perspective and think from a third person perspective what we call theory of Mind such as thinking about what others are thinking or thinking about your own thinking Olie is headed for a termite mound in her mouth she carries two grass tools she has picked them even before sighting the mound a demonstration of true forethought to do this primates use a newly evolved brain region which grew out of the neocortex called the granular prefrontal cortex and it allows the brain to simulate things more generally for example the primatologist Emil
menel observed a female chimp named Belle who came up with strategies to try and trick another dominant male chimp from stealing food at First Belle simply sat on top of a food source when she found it until Rock realized this and pushed her to get at the food in response she came up with a new strategy where she would avoid new food locations at first waiting for rock to look away and then immediately run towards it and once Rock figured that out Bell began trying to lead Rock in the wrong direction walking away from food
and eventually Rock realized he should search for food in the opposite direction that Bell was leading him and this process of ever escalating deceptions revealed that Bell and rock were able to understand what was going on in each other's minds and there have been many science experiments done to prove just this that primates can in fact understand what others hold in their mind and how that might change the key point for this breakthrough though is that primates can hold accumulated knowledge that never originated in their mind or their own personal experience the simplest example is
watching someone get hurt and then not getting hurt yourself learning from their experience not yours this form of imitation learning extends to learning more complex behaviors for example we see chimps learn to use tools by first observing and then mimicking their parents these patterns may be passed from generation to Generation by observational learning the chimpanzee has his own primitive culture and this sets up our final breakthrough language in humans we see a unique ability to communicate our imaginations with symbols we create and alongside the development of language was the development of writing and we can
chain these symbols together in sequences of any length to represent thoughts of any kind or of any level of complexity and when reading or listening to language we can see someone else's thoughts and simulate those thoughts as if we had them too and this Unleashed humans from the Practical limitation of imitation learning unlike chimps which could only learn by observing others actual actions we can learn from other people's imagined actions and this allows knowledge to accumulate without action at all just by observing symbols writing is important because it allows those thoughts to accumulate outside living
minds and build up across Generations resulting in a technological explosion unlike anything else in nature language allowed thought to break free from the human mind completely so we've seen intelligence arise out of the evolution of progressively more complex sources of learning starting with your own actions then your own imagined actions and then other people's actions and finally learning from other people's imagined actions recently the field of AI has exploded due to large language models like Chachi BT which have learned to model human language by reading all the text in the world these systems take the
exact opposite approach as that taken by Evolution human intelligence ended with language while large language models begin with language they take the final output of human thought as their input this is the big debate in the field of AI can artificial intelligence shortcircuit the steps that human intelligence had to take and go straight to language or will it need to build up from more foundational experience and if so what form does that experience need to take to put it simply does your brain need your body thanks to Jane Street for sponsoring this artof the problem
video Jan street is a quantitative trading firm with offices in New York London Hong Kong Amsterdam and Singapore they use things like machine learning distributed systems programmable hardware and statistics to trade on markets around the world you might like to know that Jan street is looking for intelligent people from all backgrounds who love solving complex problems to join their team and what's nice is you don't need a background in finance to apply they look for Curious people who like to solve problems after all the stock market is a form of distributed intelligence which no human
or computer can predict longterm perhaps the next great trading idea will come from you if you want to learn more about Jane Street's work and explore their opportunities check out their website at janest street.com
