What do AI and pigeons have in common? A groundbreaking discovery from Ohio State University reveals that our feathered friends, often branded as ‘dim-witted,’ operate on the same principles as artificial intelligence (AI) when solving tasks. Pigeons, leveraging innate mechanisms, tackle intricate problems that typically stymie human cognition.
Pigeon Brilliance: Outperforming Humans
Historical studies confirmed that pigeons could master multifaceted categorization tasks, areas where human cognitive strategies, such as selective attention or rule use, falter. This led Brandon Turner, the study’s lead and a psychology professor at Ohio State University, to speculate about the pigeons’ “brute force” approach – a methodology bearing a striking resemblance to AI models.
Testing their theory, Turner and his fellow researcher, Edward Wasserman from the University of Iowa, set up an experiment mimicking the suspected pigeon problem-solving method using a rudimentary AI model. The outcome? Resounding success. Turner shared, “The mechanisms steering pigeon learning seem to echo the foundational principles underpinning current machine learning and AI methods.”
The duo’s insights hint at nature’s genius in crafting an extraordinarily efficient learner in pigeons. Unlike humans, these birds don’t have the capability to generalize or extrapolate, yet, their particular brand of problem-solving can surpass our own in certain areas.
Decoding the Pigeon Method
The researchers’ experiment required pigeons to classify various visual stimuli. These ranged from lines of differing widths and angles to concentric and sectioned rings. Depending on their classification, pigeons had to tap either the right or left button. Correct choices earned them a food pellet, while mistakes went unrewarded.
Pigeons’ accuracy astonishingly leaped from 55% to 95% in easier tasks through mere trial and error. Even in trickier scenarios, their success rate rose from 55% to 68%.
Turner explained that pigeons possibly utilize associative learning – forming links between two related concepts. While this method of learning is generally perceived as too rudimentary for advanced categorization, the pigeons, to the researchers’ amazement, applied it adeptly.
AI and Pigeons: An Unexpected Parity
To parallel the pigeon’s problem-solving skills, the researchers employed an AI model. It was engineered using only the basic processes attributed to pigeons: associative learning and error correction. In tasks similar to what pigeons faced, the AI model demonstrated significant improvements in predicting correct answers, mirroring the birds’ achievements.
Humans, when faced with the pigeons’ tasks, would instinctively attempt to formulate a rule or multiple rules to simplify the challenge. This urge, Turner pointed out, could hinder humans since these tasks lack any simplifying rules. He noted, “Humans often abandon such tasks out of frustration. But pigeons, without attempting to set rules, resort to a brute force approach of trial and error coupled with associative learning. In certain tasks, this strategy propels them ahead of human performance.”
The fascinating revelation is how pigeons inherently adopt a learning strategy mirroring human-made AI. Turner mused this irony: “We laud our ingenuity in creating AI, yet we scorn pigeons as unintelligent creatures. However, the learning techniques steering these AI systems strongly align with the pigeons’ approach.”