Examining Emergent Abilities in Large Language Models: Why Tacit Knowledge is More Important Than Deliberate Practice

Examining Emergent Abilities in Large Language Models: Why Tacit Knowledge is More Important Than Deliberate Practice

The concept of emergence, popularized by Nobel laureate Philip Anderson in his essay "More is Different," suggests that quantitative changes in a system can lead to new behaviors. This idea has been observed in various disciplines such as physics, biology, economics, and computer science. In the context of large language models, emergent abilities refer to those that are not present in smaller models but become apparent as the model scales up.

The study of emergent abilities in large language models is of great scientific interest and serves as a motivation for future research. When scaling up language models, the behavior of the model either predictably grows with scale or experiences a surge from random performance to above random at a specific scale threshold. These emergent abilities can provide valuable insights into the capabilities and limitations of language models.

On the other hand, the concept of tacit knowledge emphasizes the importance of knowledge that cannot be fully captured through words alone. Tacit knowledge is acquired through emulation, action, and apprenticeship. It involves learning by imitating and internalizing the principles behind certain actions. This type of knowledge is difficult to articulate or explain, as it relies on complex judgment and instantaneous solution selection that balances multiple considerations.

Many researchers have found that attempting to encode all the nuances and branches of tacit knowledge into an expert system is extremely challenging. Simply providing a list of procedures to follow denies individuals the opportunity to develop expertise and hinders their ability to engage in creative problem-solving. It is crucial to recognize the value of tacit knowledge and the limitations of explicit explanations in certain domains.

Deliberate practice, as defined in fields with well-established pedagogy such as music, math, and chess, can only exist in domains where tacit knowledge has already been developed over time. In other words, deliberate practice is not the primary method for acquiring tacit knowledge. Instead, learning tacit knowledge involves finding a master in the field, working under their guidance for a substantial period, and absorbing knowledge through emulation, feedback, and osmosis.

The field of Naturalistic Decision Making (NDM) focuses on making the process of acquiring tacit knowledge more effective. NDM methods aim to enhance the transfer of expertise through practical, real-world experiences rather than relying solely on deliberate practice. Understanding and exploring tacit knowledge are essential for developing expertise and fostering creative problem-solving skills.

In conclusion, examining emergent abilities in large language models and recognizing the importance of tacit knowledge go hand in hand. Both concepts shed light on the potential of scaling up language models and the limitations of explicit explanations in certain domains. To leverage these insights, here are three actionable pieces of advice:

1. Embrace the concept of emergence: When working with large language models, be open to the idea that scaling up can lead to new behaviors and abilities. Explore the patterns and behaviors that emerge as the model grows in size, as they may provide valuable insights for further research and development.

2. Prioritize tacit knowledge acquisition: Instead of relying solely on deliberate practice, seek opportunities to learn from masters in the field and engage in real-world experiences. Emulate their actions, seek feedback, and immerse yourself in the principles and judgments that underlie their expertise. Recognize the value of tacit knowledge and its role in fostering creative problem-solving.

3. Stay informed about NDM methods: Keep an eye out for advancements in the field of Naturalistic Decision Making, as they often focus on enhancing the acquisition and transfer of tacit knowledge. Stay updated on research and practices that emphasize the importance of practical experiences and learning through emulation, feedback, and osmosis.

By combining an understanding of emergent abilities in large language models with a focus on acquiring tacit knowledge, we can unlock new possibilities and enhance our expertise in various domains. Embracing emergence and recognizing the value of tacit knowledge are essential steps towards advancing our understanding of complex systems and fostering innovation and creativity.