---
With that said - I love this question!
---
The premise, as interpreted:
**"A self-teaching application which can adjust behavior and performance with minimal input"**
---
Proposal A: [NMAP/PRISM/ETC for AI/ML systems] ((I'd be sure Palantir has this covered...))
This one will be long winded and off-the-cuff based on some work we did with tracking software in the past (people places, things, things-that-kill-people-places-and-things, etc)
I would like an AI awareness layer whereby an AI/ML system can locate adjacent 'other' systems and either silently learn - or actively map/learn etc...
We have had netmon tools for decades. I'd prefer a single firewall device;
I plug it in at home - and its my VPN/PROXY that ALL my familial devices run through. ALL.
So its just a simple VPN etc - all my mobile devices in the family of devices only route through my home device - which in-turn also keeps a small constellation of other nodes several hops out - such that BGP is sorta accomplished - cant get to primary node, go here instead.
I've designed SDNs on mass-fiber btwn HK, JP, Indonesia, Malaysia, etc...
Its doable (this is when Palo Alto Networks was just coming online with their product line...)
Doing this in a manner which utilized AI/ML/etc to more intelligently setup these systems...
Yes I could lead a team that does that....
Lets think this thing out into the zeitgeist and if its lame, then let it die - if it is not lame - then I will work with anyone who may want to explore it?
