Discrimination is the king!
Wheather you like it or not, forgetting (or discriminating) is the key to learning. It’s also the key to smarter AI.
Why to discriminate?
The more you reject as rubbish (discriminate), the easier it is to remember remaining parts. This is why the human brain has a hippocampus (Wiki Link). This small part of our brains is the middleman between a short and a long term memory.
As an algorithm designer — obsessed with reverse engineering — I believe hippocampus has this smart but simple way to discriminate informations. I have described it in my earlier article Here.
How to discriminate?
When we sleep, our brain picks up elements filtered by the hippocampus and tries to connect them with already existing memories (hence our twisted dreams). I have proposed a model for such discrimination which — in my opinion should be the fundation for all future AGI systems.
Is it even useful?
The faster and more efficiently you discriminate, the less informations you have to store and process. Hence the algorithm should be easy, fast and efficient (RAM-wise). Imagine having the luxury of rejecting 90% of data before it even gets to your dataset.
Example?
Yep, here is something I did 7 years ago: AI Researcher as a proof of concept. But now I’m preparing a browser plugin that allows you to display filtered data directly on a page and save 90% of your time while reading.
Stop reading!
To propagate Digital Hippocampus idea I’m launching a browser plugin soon. If you want to save time and let algos to filter data for you — follow me on LinkedIn and you’ll get the news shortly!
Cheers!
About Me
Marcin Rybicki, former game developer, algorithm enthusiast.
I‘m working on an ambitious project called — MarieAI — an unsupervised Machine Learning algorithm that could grow into AGI and work even without computers!
