SDS 611: Open-Ended A.I.: Practical Applications for Humans and Machines — with Kenneth Stanley

Objective Challenges

Dr. Kenneth Stanley challenges the conventional reliance on objectives in machine learning, suggesting that they can often mislead progress. He introduces the concept of "Novelty Search," which prioritizes discovering interesting outcomes over strictly following predefined goals 1. This approach questions the effectiveness of traditional reward functions, as Jon Krohn reflects on the limitations of optimizing for a single objective like happiness or contentment 2. Stanley explains,

Objectives are an absolute embarrassment, like, they should not be losing to an algorithm that doesn't know anything about what it's trying to do.

--- Kenneth Stanley

This perspective encourages a broader exploration beyond rigid objectives, potentially leading to more innovative discoveries.

   

Learning Experience

The journey of Jon Krohn and Kenneth Stanley highlights the value of embracing unexpected career paths and experiences. Krohn shares how pivotal encounters led him to make significant career shifts, emphasizing the importance of following one's interests rather than predefined goals 3. Stanley supports this by suggesting that AI insights can lead to self-discovery, proposing that understanding AI can help us better understand ourselves 4. He notes,

There's a much bigger implication...when has ever there been an algorithmic insight that leads to social critique or to understanding yourself better?

--- Kenneth Stanley

This approach advocates for a more flexible and exploratory mindset in both personal and professional development.

   

Education Reform

The current education system, according to Kenneth Stanley, often stifles creativity by focusing too heavily on objectives. He argues that true exploration and innovation are hindered by the regimented nature of traditional education, which only begins to allow for genuine exploration at the PhD level 5. Stanley envisions a system where exploration is encouraged from an early stage, akin to his concept of algorithms that prioritize novelty over objectives 6. He imagines,

Instead of trying to get around the track, which is the normal objective function, what if I just think about it, I'm just going to try to do something new that's interesting.

--- Kenneth Stanley

This perspective suggests a need for educational reform that fosters curiosity and innovation from the ground up.