I see a great field of AI use for especially smaller models that run on SBC systems like a Raspi5 8GB with an AI kit, way off being a chat bot for boring moments.
Think of a Smart Home set up. I have a window and outside door in the kitchen. The living room also has an outside door and two more doors of which one is a door to the kitchen. On each is a sensor indicating if the door is closed or open. The heating valve on all heaters is a smart home one.
I can set up a rule that when both the kitchen window is open and the door to the living room also the heater in the living room goes down with all kitchen heaters, but my girl, the imaginary one, would start feeling cold.
So, I either set up a sophisticated set of If - Then rules in an all alone workday that will either work or fail, but has nothing in between, or use a small reasoning AI have the decision if anyone is in the living room by analyzing all others sensors, such as TV running, radio playing, moving sensor active in the last 10 minutes, lights on, reading light on, to make an educated AI decision, that will sometimes be right, sometimes be wrong, but spare hours of creating If - Then procedures that either work or don't.
AI hallucinations can be acceptable in such non security or safety tasks especially if the AI is created to be self-learning improving over time.
Such tools don't exist yet, but they are also no job for the OpenAI multi million Dollar business economic stratosphere that need standard solutions for mass deployment. Such frameworks and tools are the space of excellence for the Open Source Community.
There is a great chance, if you repost that and comparable articles to inspire, that the world will get full of tiny and little AI systems working in the background and by their hallucinations give their places a form of soul... like your old car; Drinks, Stutters and sometimes Crashes and that boy is only all mechanics and electric wires.
#cyberpunkcoltoure
