On the way to the library, Comet stops under the new bus-arrival sign by the front doors. "Six minutes," she reads. "Is this AI too?"
She spins around and points at the hallway thermostat. "And that? Is everything with a sensor AI now?"
Wren writes both on his clipboard. "What does the evidence say? We need to know what each one does, not how shiny it looks."
Nova hovers between them and projects two columns on the wall: Software and Physical things. "Would you like a hint?" she asks. "Follow the signal from the sensor to the output."
Comet laughs. "Okay, reviewer. You follow the signal. I will hold the door."
AI can run inside software, like a book recommender in the library catalog. It can also run inside physical things, like a sign or a thermostat.
Computers perceive the world using sensors. Perception means getting meaning from sensor signals.
An AI system then produces an output, such as a prediction or a decision. That output can affect a real place, like a warm hallway.
| Made-up Harbor Point system | Software or physical? | What goes in | Output |
|---|---|---|---|
| Library book recommender | Software | Past borrowing records | A recommendation: "You might like this book next." |
| Hallway thermostat | Physical | Temperature and motion sensor signals | A prediction of busy times, then a decision to warm the hall early |
| Bus-arrival sign | Physical | Signals from a location sensor on the bus | A prediction: "Bus in 6 minutes" |
| SH-1 Ask | Software | A typed study question | Content: a written answer |
Some AI agents talk with people in natural language, the everyday language people speak and write.
Deep neural networks such as large language models are making this possible.
The packet says SH-1 is built on a large pre-trained language model (P2). That is why Ask can answer in full sentences.
Computers build representations of the world from data structures. Reasoning algorithms use those representations to work out new information.
AI can reason about very complex problems, but it does not think the way a person does.
Nova can joke and wonder in our story because she is a story character. SH-1 would be a tool people built.
| Statement | True or false? |
|---|---|
| AI can drive software and physical systems. | ? |
| Perception means getting meaning from sensor signals. | ? |
| A system that uses a sensor must think the way a person does. | ? |
| AI can reason about complex problems, but not the way a person thinks. | ? |
| Large language models help some AI talk in natural language. | ? |
Nice tracing, reviewer. Tomorrow in the Explorer Lab you sort twelve school systems into rules and AI.