Intelligence and Inherent Limitations
It is well-known and established that current strides in artificial intelligence can be summarized as narrow or domain-specific.
This means that we train our algorithms to be tuned to solve and/or represent a specific problem. That could be credit card fraud detection, facial recognition, brain lesion segmentation and countless other tasks thousands of practitioners are working on.
Essentially, these are standalone algorithms that can really only understand what they have seen, and they fail to generalize on adjacent and far-away problems.
And as we move toward algorithms that get better and better at a group of related tasks to semi-related tasks to unrelated tasks, we may need to start thinking about how we encode things that drive the algorithm to solve tasks.
I am talking about more nebulous things like — reducing the effect of bias, optimizing fair loss functions, and ultimately the utility of the algorithm’s use for a specific task.
But — there is one interesting aspect that is sort of talked about these days and has come up in a book I recently read — A Stange Bird by Jeff VanderMeer. And it is this idea of having something inherent within a body that makes it feel purpose. Or feel that your time is limited. That you need to finish something in order to reach some desired outcome.
Again, nebulous, I know.
But as humans, we have values that guide our thinking and actions, we have teachers that tell us what is out there, we have a physical body that is bound by genetics, we have goals, we suffer from anxiety, etc.
These are all limitations, sometimes hard-wired, sometimes self-inflicted. But they work in a way to push us in a direction that requires us to make another decision and another and another. Down a path where constant predictions are made.
This may be one of the crucial ingredients to making an intelligent life that we can relate to, one that has inherent limitations set by us or set by nature.
Who knows.
Thanks for reading.
