A musing on how intelligence comes to be.
The bedrock of intelligence is abstractions – the thing we do when we throw away a lot of information and just emphasise on a subset of it (e.g. calling that thing an apple instead of describing all its atoms and their x, y, z positions).
But where does the drive to form abstractions comes from? What if it rose from our desire to communicate with others? Since communication bandwidth is always limited, we are driven to find most efficient way of getting an idea across which leads to abstractions. Imagine a world where energy and time is unlimited, we might be communicating all x,y,z positions of things instead of putting labels on them.
We form abstractions when we notice some aspects of situations that we have encountered so far that we expect to find again in future situations. So the label “apple” becomes an efficient placeholder for all apple-like objects that are sweet and can be eaten.
It’s interesting to note that languages navigate a tradeoff between efficiency and fidelity. We want to communicate what we mean but we cannot spend enormous amount of energy on it. But we can’t spend too less energy too since communication channels are noisy and the recipient may not understand what we mean.
But since infinitely many abstractions for the same information is possible, how do we select which one to use? For example, why did the Arabic numbering system win over other alternative abstractions of numbers? Well, our abstractions are grounded by usefulness. In math, base-10 number system beat base-9 number system because we have 10 fingers and hence easier to count in base-10. (Of course, historical accidents may lead to initially inefficient systems such as Gregorian calendar to get adapted).
Why do we need to communicate useful abstractions? Because we are driven to co-operate by virtue of our shared gene pool. We know sharing useful abstractions are going to help the copies of our genes spread.
In summary, intelligence may have bootstrapped via the drive to help copies of ourselves survive better by sharing useful tools/abstractions.
PS; I find it interesting that in almost all AGI architectures, there’s no notion of replication, genes, survival, efficiency, inter-agent communication, energy, etc. – which might ultimately prove to be a useful framework in which intelligence arises naturally.
Join 150k+ followers
Follow @paraschopra