![]() Such a strong surge of neural firing is exactly what synapses need in order to change their strength-what neuroscientists call “long-term potentiation.” When a pupil pays conscious attention to, say, a foreign-language word that the teacher has just introduced, she allows that word to deeply propagate into her cortical circuits, all the way into the prefrontal cortex. ![]() ![]() With conscious attention, the discharges of the sensory and conceptual neurons that code for an object are massively amplified and prolonged, and their messages propagate into the prefrontal cortex, where whole populations of neurons ignite and fire for a long time, well beyond the original duration of the image. This is utterly different from the extraordinary amplification that occurs in our brain whenever we pay attention to an object and become aware of it. Unattended objects cause only a modest activation that induces little or no learning. Information about it is discarded early on, and it remains confined to the earliest sensory areas. If I don’t pay attention to the Frisbee, this part of the image is wiped out: processing goes on as if it did not exist. Nowadays, any sophisticated artificial intelligence system no longer connects all inputs with all outputs-it knows that learning will be faster if such a plain network, where every pixel of the input has a chance to predict any word at the output, is replaced by an organized architecture where learning is broken down into two modules: one that learns to pay attention, and another that learns to name the data filtered by the first.Īttention is essential, but it may result in a problem: if attention is misdirected, learning can get stuck. When describing the Frisbee, the network concentrates all its resources on the corresponding pixels of the image and temporarily removes all those which correspond to the person and the park-it will return to them later. Today, if artificial systems manage to successfully label a picture (“A woman throwing a Frisbee in a park”), it is because they use attention to channel the information by focusing a spotlight on each relevant part of the image. Very quickly, the idea of learning to pay attention spread like wildfire in the field of artificial intelligence. They showed that attention brought in immense benefits: their system learned better and faster because it managed to focus on the relevant words of the original sentence at each step. Their first model learned to translate sentences from one language to another. It was only in 2014 that two researchers, Canadian Yoshua Bengio and Korean Kyunghyun Cho, showed how to integrate attention into artificial neural networks. At each stage, our brain decides how much importance it should attribute to such and such input and allocates resources only to the information it considers most essential. This is why a pyramid of attention mechanisms, organized like a gigantic filter, carries out a selective triage. Initially, all these messages are processed in parallel by distinct neurons-yet it would be impossible to digest them in depth: the brain’s resources would not suffice. Our brain is constantly bombarded with stimuli: the senses of sight, hearing, smell, and touch transmit millions of bits of information per second. Why did attention mechanisms evolve in so many animal species? Because attention solves a very common problem: information saturation. These are ancient mechanisms in evolution: whenever a dog reorients its ears or a mouse freezes up upon hearing a cracking sound, they’re making use of attention circuits that are very close to ours. In cognitive science, “attention” refers to all the mechanisms by which the brain selects information, amplifies it, channels it, and deepens its processing. In the space of a few minutes, your brain went through most of the key states of attention: vigilance and alertness, selection and distraction, orientation and filtering. making you forget which check-in counter you were supposed to go to. This message, which your brain considers a priority, takes over your attention and invades your consciousness. Suddenly, you turn around: in the crowd, an unexpected friend just called your first name. Advertisements all around call out to you, but you do not even see them-instead, you head straight for the check-in counter. Your mind on alert, you look for the departures sign, without letting yourself be distracted by the flow of travelers you quickly scroll through the list to find your flight. ![]() Imagine arriving at the airport just in time to catch a plane. Everything in your behavior betrays the heightened concentration of your attention.
0 Comments
Leave a Reply. |