All activity, all intelligence, rests in finding new levels of effective context sensitive heuristics that work in practice at near real time.
No one can work anything from first principles in real time.
All any of us can do is use useful approximations.
Making systems recursively flexible to allow new sets of approximations at any level is about the best definition of intelligence available.
No person has awareness of reality.
We live in reality.
Our awareness resides in a subconsciously created model of reality.
Understanding that, is step one to understanding what we are, and how to get to AI.
Look at how our sensory systems actually work.
Heuristic hacks at every level condensing information and influencing probabilities to action.
Reflexes – hacks that act even before the brain is notified.
The brain is slow.
The only way consciousness can work is by residing in a slightly predictive model of reality.
Look at how our neurons are structured for sequence and prediction.
Expected vs actual – update to actual.
Start to notice in life where the model and the actual don’t match. In a sense that is what meditation and mindfulness training is all about.
Just look at the numbers of being human.
We tend to consider people as units.
Yet every person consists of about 10,000 times as many cells as there are people on the planet. To see all of those cells, at 3 per second, would take a million years.
Inside every cell is about 5 times as many molecules as there are cells in our bodies.
If we could have somehow taken a snapshot of all those molecules, and blown them up to a size we could see, and if we had been looking at that one image at 100 molecules every second, since the universe began some 14 billion years ago, we would be about 1% of the way through looking at that one image.
To slow down atomic action to the point that we could magnify and watch it, just to the point we could see an average water molecule moving near an enzyme in our body (not the fast ones, just the average) it would take 30,000 years to watch 1 second’s worth of action in real time.
We are complex.
Understanding anything requires simplifying heuristics – hacks that work in practice, mostly, in context.
That is as close as any of us can ever expect to get to understanding.
Absolute knowledge is total myth – Santa Claus stuff.
Absolute security is the same.
All any of us has is useful probabilities, useful hacks that seem to work in practice in the particular contexts of our lives.
Any claims to anything more than that are pure “snake oil”.
Those are my useful hack!
AI, any intelligence, is going to be subject to very similar constraints.
Some classes of problem lend themselves to rapid and useful approximation, some don’t.
And there is no possible way, (in either logic or reality) of being sure which is what.
Welcome to life.
Anything less than total acceptance of that is delusional.