Modern AI is built on neuroscience from the ’50s and ’60s. The potential of AI with the latest breakthroughs is unimaginable.

Modern AI is built on neuroscience from the '50s and '60s. The potential of AI with the latest breakthroughs is unimaginable.

Neural Networks and the Brain: Bridging the Gap between A.I. and Neuroscience

Neural Networks

In the world of artificial intelligence (A.I.), the term “neural networks” is often thrown around without fully appreciating its origins and potential. It’s fascinating to see how the field of neuroscience has inspired the development of deep-learning models, yet many A.I. professionals remain unaware of these breakthroughs. However, if we truly want A.I. systems that can push the boundaries of science, it is crucial to bridge the gap between neuroscience and A.I.

A.I. versus the Human Brain

The capabilities of the human brain are nothing short of extraordinary. Back in the 1930s, Donald Hebb and others laid the foundations for understanding how neurons learn, which served as inspiration for the first deep-learning models. In later years, David Hubel and Torsten Wiesel’s Nobel Prize-winning discoveries on the brain’s perceptual system greatly influenced the development of convolutional neural networks, a vital component of modern A.I. deep learning.

While neuroscience has experienced explosive growth in the last few decades, many recent breakthroughs have yet to manifest in A.I. systems. Most A.I. professionals today are unaware of these advances and fail to recognize the potential impact of recent neuroscience breakthroughs on A.I. This lack of awareness must change if we want A.I. systems capable of pushing the boundaries of scientific knowledge.

For example, researchers have identified a common brain circuit that can serve as a template for A.I. But the disparities between the brain’s efficiency and A.I. systems’ energy consumption are staggering. While the human brain consumes a mere 20 watts of power (less than half the energy of a light bulb), large language models like ChatGPT consume electricity equivalent to hundreds of thousands or even millions of people. Moreover, the computational resources needed to train such A.I. systems have been doubling every 3.4 months since 2012, rendering the situation unsustainable in terms of energy consumption and costs.

Not only is the brain incredibly energy-efficient, but it is also truly intelligent. In contrast to A.I. systems, the brain can understand the structure of its environment, make complex predictions, and carry out intelligent actions. Humans learn continuously and incrementally, unlike A.I. models that require retraining to correct mistakes. This highlights the limitations of current A.I. systems and presents an opportunity for improvement through the integration of neuroscience principles.

Bridging the Gap between Neuroscience and A.I.

Despite the urgent need for collaboration between neuroscientists and A.I. practitioners, cultural differences hinder effective communication between the two fields. Neuroscientific experiments require meticulous recordings, measurements, and analysis, often resulting in research papers that appear as incomprehensible jargon to A.I. professionals and computer scientists.

To bridge this gap, neuroscientists must step back and explain their concepts from a big-picture standpoint, helping A.I. professionals understand the relevance and potential applications of their findings. Additionally, we need more researchers with hybrid A.I.-neuroscience roles to facilitate collaboration and translate neuroscientific insights into brain-inspired A.I. techniques.

Exciting breakthroughs have already demonstrated the benefits of applying brain-based principles to large language models. Through interdisciplinary collaboration, A.I. researchers can gain a better understanding of how to translate neuroscientific findings into practical applications. This integration can significantly increase efficiency and sustainability while reducing the need for extensive training data.

Various organizations, including government agencies, academic researchers, and technology companies like Intel and Google DeepMind, are making progress in applying brain-based principles to A.I., for instance, utilizes Numenta’s technology to explore brain-inspired solutions. These efforts are crucial in expanding A.I. while minimizing its environmental impact as deep learning systems continue to grow in size.

Building a Sustainable and Efficient Future

Throughout history, society’s greatest breakthroughs have emerged from interdisciplinary collaboration and multiple contributions. A.I. and neuroscience must follow this path if we are to create a future where A.I. systems genuinely interact with scientists, enhance human capabilities, and improve lives in all aspects.

Whether we like it or not, A.I. is already an integral part of our lives. It is our responsibility to make it sustainable and efficient by bridging the neuroscience-A.I. gap. This requires incorporating cross-disciplinary research, commercialization, education, policies, and practices that prioritize the improvement of the human condition while protecting the environment.

In conclusion, the integration of neuroscience discoveries with A.I. has the potential to revolutionize the field and create smarter, more efficient systems. By acknowledging and leveraging the remarkable capabilities of the human brain, we can push the boundaries of A.I. while ensuring a sustainable and collaborative future. It’s time to unleash the true potential of A.I. through the marriage of neuroscience and technology.

Subutai Ahmad is the CEO of Numenta.

The opinions expressed in commentary pieces are solely the views of their authors and do not necessarily reflect the opinions and beliefs of ANBLE.

More must-read commentary published by ANBLE: