Advancements in technology are normal nowadays, and we get something new and exciting every passing day, like the evolution of AI-enabled drones. We feel the same when we learn that NeuroAI researchers have designed a new AI model, which they claim to be more efficient and responsive as the efficiency of the human brain inspires it.
Artificial intelligence has positively changed a lot, and it is now better and more advanced than ever before. It reads, talks, designs, is creative, a problem solver, and even recommends business decisions; overall, it has a lot of data stored in it. However, despite all these advancements, AI is still under pressure due to certain limitations and critical shortcomings.
Kyle Daruwalla, the NeuroAI Scholar at CSHL, which is Cold Spring Harbor Laboratory, says that although ChatGPT and all other AI technologies are impressive in the way they interact with the physical world, sadly, they are still limited. The actions they can perform, like solving math problems or writing essays, are good, but to make them well-written, these AIs need billions and billions of training examples.
Daruwalla wants to fill this gap of efficiency in AI, searching for unconventional ways of designing and making new AI models that are capable of overcoming these computational obstacles. And it looks like he is successful in his efforts.
Brain-Inspired AI By Kyle Daruwalla
This new machine-learning model supports an unproven theory that establishes a relationship between working memory and academic performance and learning. The basic reason behind it is the management of data movement. Modern computing consumes most of its energy by transferring data. Artificial neural networks consist of billions of connections, which causes data to travel long distances.
So, for the solution to this problem, Daruwalla wants inspiration from something that exists and is computationally efficient and energy-saving: the human brain.
Daruwalla’s new AI algorithms are designed to mirror the human brain’s movement and processing of new information. According to this design, the AI “neurons” should individually receive feedback and adjust instantly instead of waiting for an entire circuit update. As a result of this short-distance data travel, the processing of data will be done in real-time. It’s similar to our brain, where connections are constantly changing and adjusting; instead of first, you pause, then adapt and resume it.
Daruwalla’s design will work as a pioneer for further advancement in AI technology, where AI learns like humans. This results not only in the efficiency and accessibility of AI but also represents a full-circle moment for neuroAI. Remember that Neuroscience has been providing AI with valuable data for many years, even before the ChatGPT AI uttered its first digital syllable, which now has become much more advanced with the memory function. And AI may likely return this favor soon.