Authors: Tyler Phillips (Research Psychologist & Research and Content Consultant) and Dr Etienne van der Walt (neurologist and CEO & founder of Neurozone®)
Artificial intelligence (AI) has made impressive strides in recent years, advancing from simple, rule-based programs to sophisticated systems capable of complex, multi-layered tasks. Thanks to improvements in neural networks, machine learning, and deep learning, narrow AI—the kind of AI designed for specific functions—has become adept at recognizing patterns in large datasets, solving nuanced problems, and even generating creative outputs. Despite these advancements, the concept of "general AI," which could match human intellectual versatility, remains elusive. There are currently many barriers to the realization of general AI, and one of them is the issue of energy efficiency.
AI’s Energy Demands vs. the Human Brain’s Efficiency
The human brain is a marvel of energy efficiency. One group of researchers illustrates this well. With roughly 100 billion neurons, it uses just 12 watts of power—less than an average light bulb. In contrast, the energy requirements for AI systems are staggering. A typical laptop processor needs about 150 watts, while the world’s fastest supercomputers use over 21 million watts for all their applications – one of which is simulating AI. According to estimates based on Switzerland’s Blue Brain Project, simulating a human brain’s thought processes would require around 2.7 billion watts. This makes our organic brains many million times more energy-efficient than current AI systems.
This vast difference in energy consumption suggests that developing general AI, which can think as flexibly and creatively as humans, would require overcoming significant energy challenges. While the idea of an "AI takeover" stirs debate, the sheer power demands of such systems make this scenario unlikely in the near future.
The Cost of Current AI Systems
Today’s AI systems already consume considerable amounts of energy. For instance, training OpenAI's ChatGPT-3 required about 1,300 megawatt-hours of electricity, equivalent to the annual consumption of 130 U.S. homes. This raises concerns about the environmental and financial costs of further AI development. There is, thankfully, ongoing research into making AI more energy-efficient, such as through specialized chips and algorithms designed for lower consumption. Though there is a trend towards greater energy efficiency, the overall energy footprint of the AI industry remains a topic of concern.
Learning from Biological Intelligence
The efficiency of the human brain could offer valuable lessons for AI development in this regard. It is after all a marvel that a complex sensing, communicating, control, and memory network like the brain has continuously scaled in size and complexity over 500 million years without losing the ability to function adaptively in dynamic environments.
The way the brain computes is quite different from current digital systems. Digital computation relies on very reliable and precise thresholds, which are energy-intensive criteria. In contrast, biological computation is more flexible. The brain’s processes are noisy and imprecise, yet they are ‘just reliable enough’ to produce ‘just accurate enough’ results with minimal energy use. Additionally, the brain can adjust its energy consumption based on the speed and reliability needed for a task, a flexibility that is only now being implemented in digital devices. AI systems could achieve greater efficiency by adopting more flexible, adaptive approaches similar to biological computation.
The Role of Embodiment in Intelligence
Another critical aspect of biological intelligence is embodiment. According to prominent neuroscientist Antonio Damasio, consciousness (the mind) arises from the interaction between the brain, body, and environment. Damasio views that intelligence is not just about processing information but also about being informed about what is going on in the physical body as it responds to and engages with the physical world. Intelligent processing happens all over the body – not just inside the skull. This may also explain the learning abilities of some brainless organisms, like plants. Having an organic body – or a body that acts organically – seems important to reaching the capabilities of biological intelligence.
However, there are still debates about whether a form of disembodied intelligence could exist. That being said, for AI to reach a state comparable to general intelligence, it may need to integrate more closely with a physical form, much like humans do. While robotics and autonomous vehicles are making steps in this direction, humans still excel in perceptual-motor skills, navigating and manipulating the physical world with ease. AI might "out-think" us in certain areas, but we continue to "out-maneuver" AI when interacting with our surroundings – at least for now. Currently, directing AI machines to handle tasks that require such spatial and motor skills is likely to be far less energy-efficient.
The Power of Collective Human Intelligence
Another advantage humans hold is the ability to work together to generate truly innovative solutions. When individuals collaborate effectively, they often achieve results that exceed the sum of their parts. The combined effort of a team of people, each bringing their unique perspectives and skills, can generate creative and intuitive solutions that even the most advanced AI, for all its prowess with calculation and data analysis, currently struggle to achieve. This collective creative intelligence, driven by the energy efficiency of multiple human brains, offers a formidable counterpoint to the capabilities of AI. The challenge, of course, is to create the right kind of team dynamic so that collective human intelligence can thrive.
Thriving in the AI Era
So, how can humans maintain their edge in a world where AI continues to grow? One way is by taking care of our holistic, embodied intelligence. This means maintaining our physical and mental well-being by adhering to our physiological rhythms – regular exercise, sufficient sleep, and balanced nutrition.
Another way is by fostering collaboration. In real-world organizations, efficient teamwork can be challenging due to many factors such as office politics, misaligned incentives, and cultural differences. However, overcoming these obstacles to create cohesive, well-synchronized teams is crucial. It may take collaborative efforts simply to identify and remove the obstacles that prevent effective collaboration in any given organization.
Practical strategies to enhance team collaboration and individual resilience are out there. Some of them can even be found in resources like the Neurozone® Personal Report and Neurozone® Team Report. Meanwhile, the future of AI will continue to prompt questions about how best to integrate these technologies into our world. Addressing these questions will require the combined efforts of well-maintained, biologically intelligent systems working together to shape the future.