Part 10/16:

He compares retrofitting functionalities—such as adding diving capabilities to an airplane—to the difficulty of incorporating new features into AI systems after their initial design. Addressing fundamental issues like the concept of time or comprehensive goal management must be integrated from the outset, not added later, due to the complexity and interdependence of system parts.

The Critical Role of Transparency and Safety

Transparency as a Foundation for Trust

Thoren asserts that transparency is vital for ensuring safety and predictability. He argues that systems built as black boxes—like current neural networks—lack the capacity for introspection, goal verification, or explanation, making their actions opaque and trust difficult.