Machine Learning Street Talk (MLST)
#038 - Professor Kenneth Stanley - Why Greatness Cannot Be Planned
Episode guests
Podcast summary created with Snipd AI
Quick takeaways
- Embracing divergence over convergence fosters genuine progress in AI research.
- Balancing objectives with novelty search sparks innovation and creativity in machine learning.
- Curiosity-driven AI systems enhance adaptability and engage in creative behaviors.
- Encouraging risk-taking in research challenges the dominance of incremental innovations in academia.
- Evolutionary algorithms offer new perspectives on creativity and problem-solving in AI experiments.
- Collaborative processes leveraging diverse contributions lead to innovative outcomes in novelty search.
Deep dives
The Importance of Open-Endedness in AI Research
Open-endedness in AI research, particularly in areas like neuroevolution and AI-generating algorithms, is highlighted as a promising paradigm to achieve artificial general intelligence. Professor Kenneth Stanley emphasizes the significance of exploring divergence over convergence, populations over single objectives, and the preservation of diversity for continuous innovation and problem-solving.
Challenges with Objective Optimization
The podcast delves into the limitations of objective-driven optimization in machine learning and AI research. Objectives are criticized for hindering creativity, innovation, and the exploration of novel solutions. The discussion emphasizes the need to balance objective optimization with novelty search to foster genuine progress and breakthroughs.
The Role of Curiosity and Interestingness in AI Systems
Curiosity and interestingness are portrayed as essential components in designing AI systems that can engage in intelligent and creative behaviors. By allowing for exploration based on gradients of novelty rather than fixed objectives, AI systems can enhance their adaptive capabilities and engage in genuine discovery processes.
Risk-Free Research and Innovation in Machine Learning
A humorous yet thought-provoking critique is presented regarding the state of machine learning research, where risk-free approaches and incremental innovations dominate academic pursuits. The podcast challenges the trend of rewarding safe, high-prestige research over daring and potentially transformative exploration, highlighting the need for a balance between stability and innovative risk-taking.
Achieving Innovation Through Evolutionary Algorithms
Through examples from nature and AI experiments like Pickbreeder, the podcast emphasizes the power of evolutionary algorithms in fostering creativity, problem-solving, and innovation. The concept of breeding pictures to evolve diverse and novel visual concepts illustrates the potential of evolutionary processes in generating novel solutions and expanding problem-solving capabilities.
Encouraging Diverse Collaborations and Contributions in Novelty Search
Encouraging diverse collaborations and contributions is essential for novelty search processes, where individuals build upon each other's work. This collaborative approach allows for the creation of innovative ideas that stem from diverse perspectives and experiences, leading to a rich and varied array of outcomes.
Balancing Objectives with Open-Ended Exploration
Balancing objectives with open-ended exploration is crucial in research and innovation processes. While objectives can drive focused outcomes, open-ended exploration fosters creativity and innovation by allowing for unexpected discoveries and novel solutions. Finding a balance between structured goals and unbounded exploration enables a more holistic approach to problem-solving.
Importance of Subjectivity and Interestingness in Decision-Making
Acknowledging the role of subjectivity and interestingness in decision-making processes is vital for fostering creativity and novelty. Embracing diverse perspectives and individual intuitions can lead to unique insights and breakthroughs that may not be captured by conventional metrics or objective criteria. By valuing subjectivity and interestingness, we can nurture a culture of curiosity, exploration, and innovative thinking.
Elevating Human Intuition and Creativity in Research and Funding Initiatives
Recognizing and elevating human intuition and creativity in research and funding initiatives can drive transformative outcomes and groundbreaking discoveries. By trusting experts' intuitions and fostering a culture that values creativity, we can unlock new possibilities and uncharted territories in science, technology, and beyond. Embracing creativity as a driving force behind research endeavors can lead to remarkable advancements and paradigm-shifting innovations.
Exploration and Autonomy in Innovation
Exploration and autonomy are crucial in innovation processes, as discussed in the podcast episode. The key point underscored is the importance of allowing individuals the freedom to explore ideas without being constrained by consensus-driven decisions. Autonomy fosters creativity and divergent thinking, enabling individuals to follow their interests without external interference, leading to unique discoveries and solutions.
The Significance of Divergent Search Spaces
Divergent search spaces are highlighted for their role in fostering innovation. By creating an environment that encourages exploration and divergent thinking, individuals can uncover unexpected solutions and ideas. The discussion emphasizes the value of avoiding committee-driven approaches and granting individuals the freedom to pursue their interests to uncover novel and valuable outcomes.
Challenges in Defining Interestingness
The podcast delves into the nuanced concept of 'interestingness' in the context of exploratory research. It explores the challenges in defining what makes a discovery truly interesting and the balance between novelty and relevance. The conversation touches on the subjective nature of interestingness and the need for evolving metrics to capture the essence of exploration.
Balancing Objectives and Exploration in Innovation
The podcast debates the interplay between objectives and exploration in the innovation process. While objectives are essential for optimization and refinement, exploration plays a critical role in discovering new knowledge and breakthroughs. The conversation calls for a balanced approach that integrates the benefits of both objective-driven optimization and exploratory search to drive meaningful progress and innovation.
Professor Kenneth Stanley is currently a research science manager at OpenAI in San Fransisco. We've Been dreaming about getting Kenneth on the show since the very begininning of Machine Learning Street Talk. Some of you might recall that our first ever show was on the enhanced POET paper, of course Kenneth had his hands all over it. He's been cited over 16000 times, his most popular paper with over 3K citations was the NEAT algorithm. His interests are neuroevolution, open-endedness, NNs, artificial life, and AI. He invented the concept of novelty search with no clearly defined objective. His key idea is that there is a tyranny of objectives prevailing in every aspect of our lives, society and indeed our algorithms. Crucially, these objectives produce convergent behaviour and thinking and distract us from discovering stepping stones which will lead to greatness. He thinks that this monotonic objective obsession, this idea that we need to continue to improve benchmarks every year is dangerous. He wrote about this in detail in his recent book "greatness can not be planned" which will be the main topic of discussion in the show. We also cover his ideas on open endedness in machine learning.
00:00:00 Intro to Kenneth
00:01:16 Show structure disclaimer
00:04:16 Passionate discussion
00:06:26 WHy greatness cant be planned and the tyranny of objectives
00:14:40 Chinese Finger Trap
00:16:28 Perverse Incentives and feedback loops
00:18:17 Deception
00:23:29 Maze example
00:24:44 How can we define curiosity or interestingness
00:26:59 Open endedness
00:33:01 ICML 2019 and Yannic, POET, first MSLST
00:36:17 evolutionary algorithms++
00:43:18 POET, the first MLST
00:45:39 A lesson to GOFAI people
00:48:46 Machine Learning -- the great stagnation
00:54:34 Actual scientific successes are usually luck, and against the odds -- Biontech
00:56:21 Picbreeder and NEAT
01:10:47 How Tim applies these ideas to his life and why he runs MLST
01:14:58 Keith Skit about UCF
01:15:13 Main show kick off
01:18:02 Why does Kenneth value serindipitous exploration so much
01:24:10 Scientific support for Keneths ideas in normal life
01:27:12 We should drop objectives to achieve them. An oxymoron?
01:33:13 Isnt this just resource allocation between exploration and exploitation?
01:39:06 Are objectives merely a matter of degree?
01:42:38 How do we allocate funds for treasure hunting in society
01:47:34 A keen nose for what is interesting, and voting can be dangerous
01:53:00 Committees are the antithesis of innovation
01:56:21 Does Kenneth apply these ideas to his real life?
01:59:48 Divergence vs interestingness vs novelty vs complexity
02:08:13 Picbreeder
02:12:39 Isnt everything novel in some sense?
02:16:35 Imagine if there was no selection pressure?
02:18:31 Is innovation == environment exploitation?
02:20:37 Is it possible to take shortcuts if you already knew what the innovations were?
02:21:11 Go Explore -- does the algorithm encode the stepping stones?
02:24:41 What does it mean for things to be interestingly different?
02:26:11 behavioral characterization / diversity measure to your broad interests
02:30:54 Shaping objectives
02:32:49 Why do all ambitious objectives have deception? Picbreeder analogy
02:35:59 Exploration vs Exploitation, Science vs Engineering
02:43:18 Schools of thought in ML and could search lead to AGI
02:45:49 Official ending
Remember Everything You Learn from Podcasts
