The building owner is using a replication of the copyrighted work. The owner should absolutely compensate the original creator.

But the printing company that the building owner hires to print the poster doesn’t owe the original creator anything. Even though it is directly replicating copyrighted work, and certainly benefiting from doing so. If the printer were selling the copyrighted works directly then that would be a different matter and they would have to compensate the original copyright owner. So clearly context matters.

An AI doesn’t even make a replication of the original work as part of its training process.

If the AI then goes on to create a replication, or a new work that is similar enough to the original that copyright applied, and intended to use the work in a context where copyright would apply, then absolutely. That would constitute a breach of copyright.

It is the work itself that is copyrighted, not the knowledge/ability to create the work. It’s the knowledge of how to create the work which is encoded in the neural network.

Lots of people benefits from freely distributed content. Simply benefiting from it is not enough to justify requiring a license fee.

Hypothetically speaking, let’s say a few years down the line we have robot servants. I have a robotic care giver that assists me with mobility. Much as I may have a human care giver today.

If I go to the movies with my robot care giver, they will take up a seat so I would expect to pay for a ticket, just as I would for a human care giver. Do I then need to pay an extra licensing fee for the robots AI brain to actually watch the movie?

What if it’s a free screening? Should I still have to pay for the robot brain to “use” the movie?

Is the robot “using” the movie in some unique and distinct way compared to how I would be “using” the movie?