The Path Towards Truly Autonomous Systems
To create autonomous learning software, a program must be given an internal goal. Behind the methology of a learning-based system, is the placement of a goal oriented intent (in human terms, a “need”).
What does a program need?
A program “needs” that which a programmer therein designs.
Who decides the details of that need? Those writing the code!!
Example: A self driving car needs to get its passengers to a destination safely without breaking existing rules; speed limits, street signs and traffic rules all dictate how a car makes it from point A to B.
As humans, we learn how to get our cars from A to B. However, we also use our life experience and value system in such a way as to apply logic and reasoning when faced with situations that do not fall within statistical traffic norms.
What happens when an animal runs into your lane? — Does the car avoid the animal but jeopardize the driver with a dangerous lane change? — Does the car hit the brakes to save the animal but risk a rear ender? — Would the decision be different if the hazard was a small child....? or a basketball?
As these questions are answered and algorithms are set, future systems will have those decisions set as baseline code.
Just as a calculator is expected to have a solid arithmetic foundation, future systems should similarly have a baseline algorithmic foundation upon which carefully reasoned decisions can be made.
These systematic decisions must correspond to the basic tenets of human societal demands upon which continued developments of technology must thereby conform.
As technology evolves, without a basic understanding of humanistic need and desires (I.e., preserving and improving human lives), programs may unintentionally design themselves in a way that diverges from our ideologies and best intentions.
Early baseline code is analogous to a home’s foundation. We are currently at the foundational level of autonomous systems.
Who’s setting the precedent for the baseline decisions? And finally, how can we ensure that enlightened decisions prevail?