Can a pigeon compete with artificial intelligence in terms of intelligence? On a very basic level, yes.
In a new study, University of Iowa psychologists looked at how the pigeon’s brain works and how the bird’s “raw power” of learning bears similarities to artificial intelligence.
The researchers gave the deaf complex categorization tests where high-level thinking, such as using logic or reasoning, wouldn’t help. Instead, after extensive trial and error, the pigeons were finally able to remember enough scenarios in the test to achieve an accuracy of almost 70%.
Researchers equate the repetitive trial-and-error approach of pigeons to artificial intelligence. Computers use the same basic methodology, the researchers say, “learn” to identify patterns and objects that are easily recognizable by humans. Certainly, given their vast memory and storage capacity – and increasingly powerful in those areas – computers far surpass anything the deaf brain can conjure up.
However, the basic process of association formation – considered a lower-level thinking technique – is the same between pigeons passing tests and the latest advances in AI.
“You hear all the time about the wonders of AI, all the amazing things it can do,” says Ed Wasserman, Stuit Professor of Experimental Psychology in the Iowa Department of Psychological and Brain Sciences and corresponding author of the study. “He can punch people in the pants while playing chess or other video games. He can beat us at all sorts of things. How is that? is it smart No, it uses the same or equivalent system as the dove here. »
The researchers tried to disentangle two types of learning: The first, declarative learning, relies on the exercise of reason based on a set of rules or strategies – a so-called higher level of learning, which is mainly attributed to humans. The other, associative learning, focuses on recognizing and making connections between objects or patterns, such as “sky blue” and “wet with water”.
Many animal species use associative learning, but only a few—including dolphins and chimpanzees—are considered declarative learners.
But AI is all the rage as computers, robots, surveillance systems and so many other technologies seem to “think” like humans. But is that really the case, or is AI just the product of clever human input? Or, as the study authors put it, have we compromised the power of associative learning in human and animal cognition?
Wasserman’s team developed what he calls a “freakishly difficult” test to find out.
Each test pigeon was shown a stimulus and had to decide to which category this stimulus belonged by pressing a button on the right or left. Categories included line widths, line angles, concentric rings, and subdivided rings. A correct answer resulted in a tasty pellet; A wrong answer has no effect. What made the test so challenging, Wasserman says, was its arbitrariness: No rules or logic would help decipher the task.
“These charms are something special. They are different and never repeat themselves,” says Wasserman, who has studied pigeon intelligence for five decades. “You have to remember the individual stimuli or the regions from which the stimuli come in order to solve the task. »
Each of the four test pigeons responded correctly about half the time at the beginning. But after hundreds of tests, the foursome finally increased their score to an average of 68% correct.
“Pigeons are like AI masters,” says Wasserman. “They use a biological algorithm that nature gave them, while the computer uses an artificial algorithm that humans gave them. »
The common denominator is that both AI and deaf use associative learning, and yet it was this basic thinking that allowed deaf to score successfully. If people took the same test, Wasserman says, they would get poor results and likely drop out.
“The goal was to see how well a simple associative mechanism could solve a task that would bother us because people rely so much on rules or strategy,” adds Wasserman. “In that case, these rules would hinder learning. The pigeon never goes through this process. It doesn’t have that high-level thought process. But that doesn’t stop them from learning. In a way, it even makes it easier. »
Wasserman sees a paradox in the way associative learning is viewed.
“People are impressed that AI is doing amazing things with a learning algorithm similar to the pigeon,” he says, “but when people talk about associative learning in humans and animals, it’s seen as rigid and undemanding. »
The study, Solving the Paradox of Associative Learning by Categorical Learning in Pigeons, was published online Feb. 7 in the journal Current Biology.
Study co-authors include Drew Kain, who graduated in Neuroscience from Iowa in 2022 and has a PhD in Neuroscience from Iowa; and Ellen O’Donoghue, who received her PhD in psychology from Iowa last year and is now a postdoctoral fellow at Cardiff University.
The National Institutes of Health funded the research.