A New Type of Brain-Inspired Ultrafast AI Developed

The human brain dynamics, notwithstanding the slow speed, inspire scientists to build ultrafast AI algorithms that outperform existing ones today.

The human brain is a fascinating organic machine.

The intricate network of neurons that makes us who we are has long been the subject of intense research by scientists looking for bio-inspired solutions.

Aside from the brain-computer chip projects, like Elon Musk’s Neuralink, the brain is inspiring scientists in many fields, perhaps the most so in artificial intelligence.

After all, the concept of AI itself and computers come from the desire to create brain-like machines.

The Slow Learning Brain Gives Rise to Ultrafast AI

Neuroscientists are always improving their algorithm designs as they learn new things about how the brain works.

In a new study, a research team demonstrated how the dynamics of the human brain could lead to a new type of ultrafast AI algorithm.

New insights into the workings of the human brain allowed researchers at Bar-Ilan University in Israel to create this machine learning algorithm. It even outperforms existing state-of-the-art learning algorithms in learning rates.

The brain dynamics are very slow. Compared to computers, the human brain processes at a much slower rate. But it is incredibly efficient.

There are way fewer neurons in the brain than there are bits in an average PC. And even the first-ever computers beat the brain in the computational speed.

While the human brain presents different inputs in a timed order, a computer is based on synchronous inputs. Meaning all inputs are presented simultaneously.

Processing an event with multiple objects. A synchronous input where all objects are presented simultaneously to a computer (left), versus an asynchronous input where objects are presented with temporal order to the brain (right). | Image courtesy of Prof. Ido Kanter

“When looking ahead, one immediately observes a frame with multiple objects. For instance, while driving one observes cars, pedestrian crossings, and road signs, and can easily identify their temporal ordering and relative positions. Biological hardware (learning rules) is designed to deal with asynchronous inputs and refine their relative information,” said Prof. Ido Kanter, lead author of the study.

Thanks to advance experiments on neuronal cultures and large-scale simulations, the team claim to be rebuilding the bridge between neuroscience and advanced artificial intelligence algorithms that “has been left virtually useless for almost 70 years.”

They call for this bridge to be reinitiated and think of this ultrafast AI architecture to be a step in that direction.