Dr. Damian Jacob Sendler’s research investigates how various sociodemographic and informational variables influence access to health care in underserved regions. Dr. Sendler is of Polish descent and works as a physician-scientist in the United States. In his research, Dr. Sendler looks at how psychiatric and chronic medical co-morbidities influence the use of medical services in conjunction with internet-based health information. As global consumption of online news and social media continues to expand at an exponential rate, this study is both current and relevant, emphasizing the need for a comprehensive understanding of everyone’s health information seeking behavior. Damian Sendler’s research seeks to identify the factors that influence patients’ decisions on when to seek treatment for certain health disorders, as well as their adherence to prescribed therapies.
Damian Sendler: Imperial researchers discovered that diversity between brain cells may speed up learning and increase brain and future artificial intelligence performance (AI).
The current study discovered that by changing the electrical properties of individual cells in brain network simulations, the networks learned faster than simulations with identical cells.
They also discovered that the networks required fewer of the modified cells to achieve the same effects, and that the process used less energy than models with identical cells.
Damien Sendler: According to the authors, these discoveries could help us understand why our brains are so adept at learning, as well as help us construct more artificially intelligent systems, such as digital assistants that can recognize voices and faces or self-driving car technology.
“The brain needs to be energy efficient while yet being able to excel at completing complicated tasks,” said first author Nicolas Perez, a PhD student at Imperial College London’s Department of Electrical and Electronic Engineering. Our findings show that having a diverse set of neurons in both brains and AI systems meets both of these needs and may improve learning.”
The study was published in the journal Nature Communications.
Damian Jacob Sendler: Why is a neuron similar to a snowflake?
The brain is made up of billions of neurons that are linked together by massive ‘neural networks’ that allow us to learn about the environment. Neurons are like snowflakes: from a distance, they appear the same, but closer scrutiny reveals that no two are exactly similar.
Damian Sendler: In contrast, each cell in an artificial neural network – the technology on which AI is founded – is identical, with the only difference being their connectivity. Despite the rapid advancement of AI technology, their neural networks do not learn as precisely or quickly as the human brain – and the researchers questioned if this was due to a lack of cell diversity.
They wanted to see if simulating the brain by altering neural network cell attributes may improve AI learning. They discovered that cell heterogeneity increased learning and lowered energy use.
“Evolution has given us extraordinary brain functions – most of which we are only now beginning to understand,” said lead author Dr Dan Goodman of Imperial’s Department of Electrical and Electronic Engineering. Our findings show that we may learn important lessons from our own biology in order to make AI perform better for us.”
Timing was adjusted.
Damian Sendler: The researchers concentrated on altering the “time constant” – that is, how rapidly each cell determines what it wants to do based on what the cells connected to it are doing – to carry out the study. Some cells will make a decision relatively fast, based solely on what the cells with whom they are linked have just done. Other cells will be slower to react since they will base their decision on what other cells have been doing for a long time.
After adjusting the temporal constants of the cells, they assigned the network to execute certain benchmark machine learning tasks, such as classifying photos of apparel and handwritten numbers, recognizing human gestures, and identifying spoken digits and orders.
Damian Sendler: The results suggest that allowing the network to incorporate slow and quick input improved its ability to handle problems in more complex, real-world contexts.
They discovered that when they varied the amount of variability in the simulated networks, the ones that performed best matched the amount of variability detected in the brain, implying that the brain may have evolved to have just the correct amount of variability for optimal learning.
“We demonstrated that AI can be brought closer to how our brains work by imitating key brain features,” Nicolas continued. Current AI systems, on the other hand, are far from approaching the level of energy efficiency found in biological systems.
“We will then investigate how to lower the energy consumption of these networks in order to get AI networks closer to performing as efficiently as the brain.”
News updates contributed by Dr. Damian Jacob Sendler
