Episode 96: Kenneth Stanley on the Pursuit of What’s Interesting
The Theory of Anything
Deep Dive
What is the main thesis of Kenneth Stanley's book, 'Why Greatness Cannot Be Planned: The Myth of the Objective'?
The book argues that setting specific objectives can be counterproductive in complex systems, including machine learning, business, and personal life. Instead, pursuing what is interesting and novel can lead to better innovation and discovery.
Why does Kenneth Stanley believe that setting objectives can be counterproductive in complex systems?
Objectives in complex systems are often deceptive, leading to dead ends and blinding individuals or systems to other potential stepping stones that could have led to better outcomes. This is because complex spaces are inherently deceptive, making it difficult to predict the correct path towards a distant goal.
How does novelty search work in machine learning algorithms?
Novelty search algorithms measure and reward novelty instead of a specific objective. They compare current behaviors to past behaviors to find new and interesting paths, rather than optimizing towards a predefined goal. This can lead to unexpected and useful discoveries, even in simple domains.
What is the difference between pursuing an objective and following interestingness or novelty?
Pursuing an objective involves comparing current progress to a predefined goal, while following interestingness or novelty involves comparing current actions to past experiences. Interestingness and novelty allow for more exploration and can lead to innovative solutions that objectives might miss.
Can you provide a practical example of the objective paradox in action?
In the Pick Breeder experiment, people found interesting and novel images when they were not explicitly looking for them. This shows that sometimes the best discoveries are made when we are not focused on a specific goal, highlighting the limitations of objective-driven approaches.
What are quality diversity algorithms, and how do they relate to novelty search?
Quality diversity algorithms are successors to novelty search that aim to balance both the quality and diversity of solutions. They focus on finding a wide range of high-quality solutions rather than a single optimal one, making them more practical for real-world applications while still capturing the essence of open-ended exploration.
How does Kenneth Stanley’s view on evolution differ from traditional Darwinian perspectives?
Traditional Darwinian evolution focuses on optimization and survival, leading to convergence. Stanley’s view emphasizes divergence and open-endedness, which are crucial for the long-term creation of new species and complex life. He argues that the divergent nature of evolution is what makes it truly interesting and innovative.
What practical implications does Stanley's research have for funding and structuring research institutions?
Stanley suggests that research institutions, both in government and industry, should balance objective and non-objective pursuits. Government-funded research labs could take more risks by following interestingness and novelty, while corporate labs need to justify non-objective research as a shield against future disruption.
Why do humans have a psychological need for objectives?
Humans crave certainty and a sense of security, which objectives provide. Objectives create a veneer of control and predictability, making them psychologically satisfying. However, this can limit creativity and open-ended exploration, which are necessary for significant innovation.
How literally does Kenneth Stanley mean the idea that creativity is a kind of search?
Stanley means it quite literally. He views creativity as searching through a vast space of possibilities, looking for interesting and novel ideas or solutions. This framework helps in visualizing and understanding the process of innovation and discovery.
Shownotes Transcript