The 85% Rule for Learning and Self-Taught AI: Insights into Optimal Learning and Brain Function

The 85% Rule for Learning and Self-Taught AI: Insights into Optimal Learning and Brain Function

Learning is an essential process for both humans and machines. Interestingly, it has been found that optimal learning occurs when we succeed around 85% of the time. This rule, known as the 85% rule, suggests that the ideal training accuracy is about 85%. These findings align with the 80% success rate observed in successful classrooms, as discovered by Barak Rosenshine. Despite coming from different theoretical backgrounds, these studies converge on the idea that adjusting the amount of support based on success rate is crucial.

Optimal learning theories also emphasize the importance of finding the right level of difficulty. Lev Vygotsky's zone of proximal development suggests that tasks slightly beyond our current capabilities, but manageable with assistance, maximize learning. This concept implies that learning is most effective when we are challenged but not overwhelmed. Similarly, Anders Ericsson's model of deliberate practice highlights the necessity of pushing beyond automaticity to reach our full potential. By continuously striving for improvement, we can overcome plateaus and enhance our skills.

In the realm of AI, researchers are exploring self-supervised learning algorithms that require little or no human-labeled data. Unlike traditional supervised learning, where neural networks rely on labeled datasets, self-supervised learning models mimic the way animals, including humans, learn from their environment. By exploring and interacting with the world, living beings develop a rich and robust understanding of their surroundings.

These self-supervised learning algorithms have achieved remarkable success in modeling human language and image recognition. Notably, they have shown a closer correspondence to brain function than their supervised-learning counterparts. Computational models of the mammalian visual and auditory systems built using self-supervised learning have demonstrated a strong alignment with how the brain processes information.

The brain-inspired models of artificial neural networks emerged around the same time as the revolutionary neural network named AlexNet, which significantly improved image classification. Both computational models of the primate visual system and self-supervised learning algorithms have been inspired by this breakthrough. The self-supervised algorithms create gaps in the data and prompt the neural network to fill in the missing information. Through this process, the algorithm learns to reconstruct masked images accurately.

Researchers have found that a significant portion of brain function can be attributed to self-supervised learning. The brain constantly tries to predict what will happen next, which aligns with the concept of self-supervised learning where neural networks fill in missing information. Evidence suggests that language learning is heavily influenced by predicting the next elements of speech. However, to fully understand brain function, further research is needed. Feedback connections, which are abundant in the brain, are currently underrepresented in artificial neural networks.

Actionable Advice:

1. Fine-tune the level of support: Based on the 85% rule, adjust the amount of support you provide when learning new skills or tackling challenges. Strive for a success rate of around 85% to optimize your learning experience.

2. Embrace the right level of difficulty: Find tasks that are slightly beyond your current abilities but manageable with some assistance. This zone of proximal development maximizes learning and pushes you to reach your full potential.

3. Emphasize self-supervised learning: Incorporate self-guided exploration and interaction with your environment when learning. Mimic the way humans and animals learn naturally to gain a deeper and more robust understanding of the world.

In conclusion, the 85% rule for learning highlights the importance of finding the optimal success rate for effective learning. Self-supervised learning algorithms in AI offer insights into how the brain functions and learns. By incorporating the principles of optimal learning and embracing self-supervised learning, we can enhance our own learning experiences and deepen our understanding of the world around us.