Source (Consciousness -> Meaning -> Symbols) -> Communication ->(Symbols -> Meaning -> Consciousness) Receiver
When a person communicates with another the process may be described as above:The receiver interprets the symbols she received, extracting a meaning that she internalizes with her consciousness. The source was conscious and had the intention of communicating and a meaning of what she wanted to communicate; she codified the meaning in symbols that have been communicated.
In this act of communication both source and receiving parties need to have some shared model to base the encoding and interpretation of the meaning; a shared context and culture (a common knowledge).
When a machine with an AI system generates a message (like in a chatbot) it generates it by algorithmically assemble symbols based on a statistical model. There is no consciousness, no meaning to form the intention of the communication.
The receiver would be wrong to assume that the machine intented to meant the information carried in the message. It’s a purely mechanical product.
It’s the receiving party that can attribute a meaning to the sybols, not the source.
Meaning is only in the eyes of the beholder.
Think of an abstract painter. She is concious and wants to commnicate something to the observers; she codifies the message in the painting but then eventually the observer looks at the painting and is unable to attribute a meaning; the observer just sees symbols and cannot extract a meaning.
This is symmetric to what happens with AI: the machine just mechanically assembles symbols and then the receiver attributes a meaning.
The machine does not know what the message means, it does not know it is producing a message; it does not even know that its output is a message and that someone will receive it.
With AI it’s like looking at rocks, stalagmites or clouds and seeing an animal.
The clouds don’t know it’s an animal; they don’t even know it’s a shape. They don’t know.
It’s just us, seeing the clouds who see an animal in them.