For some time now there seems to be a direct connection between the pigeons and the world of information technology. Yet what can two worlds apparently so distant from each other have in common? Recently a well-known youtuber he had wondered if in the data transfer a carrier pigeon or a Gigabit fiber optic connection were faster.
If that was a somewhat provocative experiment, this time a group of researchers from Ohio State University carried out a deeper investigation concluding that pigeons use the same principles underlying modern ones artificial intelligences to address various problems.
Brandon Turner, lead author of the new study and professor of psychology at Ohio State Universityworked with Edward Wasserman, a psychology professor at the University of Iowa, arriving at a series of interesting conclusions, published in iScience.
Associative learning underlies modern artificial intelligence and the thinking of… pigeons
L’associative learning is a type of learning in which an organism or system learns to connect two or more stimuli or events, so that the occurrence of one causes a certain expectation or consequential response. This type of learning is often associated with psychology: a subject learns to associate a neutral event with a stimulus that evokes a response. In the Pavlov’s classical conditioningFor example, a dog can learn to associate the ringing of a bell (neutral stimulus) with the delivery of food (unconditioned stimulus); alone, the ringing of the bell causes an anticipated salivary response. This is an example of associative learning.
How artificial intelligences make use of associative learning
In the context ofartificial intelligenceassociative learning is relevant in various applications, including:
- Machine learning: The algorithms of machine learning they often rely on associative learning to detect correlations and patterns in data. For example, thesupervised learning involves associating inputs with outputs, so that the model can then make predictions based on new inputs.
- Artificial neural network: Artificial neural networks are inspired by the functioning of biological neurons and use weighted connections to associate inputs with desired outcomes. Learning in these networks involvesweight update to improve performance.
- Pattern recognition: Associative learning is widely used in recognition of pattern. Think about application fields such as Facial recognition and recognition vocal: the system learns to associate pattern specific to identities or commands.
- Personalized recommendations: In recommender systems, AI uses associative learning to match the user’s past behaviors with future recommendations, such as suggesting products or content based on previous purchases or views of a set of web pages.
Turner and Wasserman explain that thanks to associative learning, a pigeon can find solutions to complex problems that are difficult for humans or other primates to reach. Primate thinking is typically guided byselective attention and by the explicit use of rules, characteristics that can hinder the resolution of some types of problems.
Image source: Ohio State University.
Parallels between the world of pigeons and the functioning of artificial intelligences
As part of the study, a large family of pigeons was faced with four different tasks. For the simplest tasks, it was found that the pigeons could learn the correct choices over time and increase their success rates from about 55% to 95%. In carrying out the more complex tasks, the birds did not show such a clear improvement over time, going from 55% to 68%.
However, the results served to show narrow parallels between the performances scored by the pigeons and those highlighted by the artificial intelligence models. In both cases the use of associative and learning techniques emerges error correction to guide their decisions towards success.
Turner noted the methodology that combines trial, error and associative learning based on “ brute force” helps achieve better results than humans.
Until yesterday, pigeons were considered rather dull animals. The question is: how come they suddenly seem to become smarter than humans?
The issue, as explained above, does not lie in these terms. Turner and Wasserman demonstrated that pigeons and humans use different approaches to solve specific tasks. In some specific cases, pigeons appear to be more effective at using associative learning and interfacing with errors to solve certain learning tasks. visual categorization.
Humans are known for their extraordinary general intelligence, which includes reasoning, problem-solving, conceptual learning, creativity and much more. These abilities greatly surpass those of pigeons in many contexts. L’intelligence it is a complex and multidimensional concept, and varies between species and even between individuals within the same species. Therefore pigeons have developed different skills and strategies to face the specific challenges that arise in their environment.
