The human brain is the most advanced supercomputer in the world. It can simultaneously learn and adapt without the use of massive data centers like today’s AI systems. This month, a team of engineers from the Texas A&M Department of Computer and Electrical Engineering and other leading institutions introduced a brain-inspired AI system that could unlock better performance and open the door for smarter protocols. Here’s what you need to know.
The Limits of Traditional AI and the Turing Architecture
Artificial intelligence is one of the fastest-growing and most exciting tech industries in the world. In a few short years, AI algorithms went from being only accessible to high-level Google and Microsoft engineers to becoming a common tool for the average person. Along this journey, AI systems have continually increased in performance and capabilities. Alongside this increase in performance, an increase in hardware capabilities has also occurred.
Notably, AI systems rely on the Turing architecture. This style of computing is named after its inventor, Alan Turing. He devised that computers could provide better performance if certain parts of their operations were separated.
This structure means that computers have their processing and memory physically separated within the chip design. This separation has dictated the hardware design of machines ever since. His vision for computers remains the most prominently used architecture, despite it being decades since his discovery.
The Energy Crisis: Why Modern AI Needs a New Blueprint
The problem with this layout is that AI systems require massive amounts of data. In some instances, entire data centers need to be built from the ground up, utilizing the most advanced hardware as part of the system requirements. Sadly, this structure is not sustainable.
Large data centers are massive power drains that can consume as much energy as entire cities. In many instances, these systems rely on renewable energy, which has helped reduce environmental impact. However, the growing demand for AI systems means that there needs to be an upgrade in efficiency in order to make these protocols sustainable in the future.
Why Today’s AI Models Cannot Adapt in Real Time
Additionally, this structure means that the scalability of these protocols is based on their hardware rather than their software constraints. Even the most advanced system still must wait for the data to be physically sent between the hardware components on a device.
Introducing a Super-Turing AI: Inspired by the Human Brain
Another drawback of Turing-based AI systems is that they rely on preset data points. Today’s models don’t offer any way to modify their performance during operation. These systems rely on the data provided to complete backpropagation.
This approach differs from your brain, which utilizes a combination of knowledge and environmental conditions to make accurate assessments with a fraction of the energy. Notably, it’s biologically impossible to replicate back propagation because it would require the person to have already experienced the situation prior. Recognizing these facts, engineers have sought to develop an AI system that mimics human learning.
How Synaptic Resistors Learn Like Our Brains
The study “HfZrO-based synaptic resistor circuit for a Super-Turing intelligent system“1 introduces a neuromorphic AI system that can mimic how the brain learns to provide real-time adaptability based on conditions. The study was put forth by a team of Researchers from UCLA, Texas A&M, and other research facilities and was published in Science Advances. To accomplish this task, the team created a “synstor” (synaptic resistor) circuit that can learn like your brain.
Source – Texas A&M Department of Computer and Electrical Engineering
Brain-Like Efficiency: AI That Consumes Microwatts, Not Megawatts
The engineers behind this study recognize that the human brain is still far ahead of AI in many aspects, such as energy consumption and flexibility. The brain’s neural processes allow it to conduct complex bodily functions while simultaneously adapting to changing conditions around you.
In terms of energy efficiency, the human brain only requires about 20 watts to operate. In comparison, today’s advanced AI system can consume up to 1 billion watts during operations. This power consumption comes from the processing pathways and components, alongside the AI systems architecture.
Understanding Synaptic Plasticity in Artificial Intelligence
The engineers began by studying synaptic plasticity in brains. Your brain has millions of neurons that power your memory and learning. These areas of communication where nerves meet are called synapses. Synaptic plasticity refers to the brain’s ability to strengthen connections that are used more often. It also refers to the brain’s ability to create new pathways when it’s deemed more efficient.
The Brain’s Secret: Switching Between Turing and Super-Turing Modes
Keenly, brains can operate in both Turing mode and a super-Turing mode. While in a Turing operational state, your brain will utilize pre learned tasks and actions. When unexpected scenarios arise or the conditions change in a way that is not predictable, your brain will automatically switch to a super-turing stance. This approach allows your brain to always adapt to your surroundings.
How Super-Turing AI Learns and Adapts Instantly
Scientists find this super turning state to be at the core of the brain’s unique capabilities. They utilized this structure to reimagine AI architectures, envisioning a more human-like operational flow where learning and inference occur simultaneously.
This strategy reduces energy consumption considerably. The elimination of the requirement to move massive amounts of data across hardware components reduces consumption and improves performance. It also supports real-time adaptability while shrinking power consumption to new lows.
Inside the Synstor: The Ferroelectric Circuit Mimicking Human Learning
At the core of this study is the creation of an inclusive sensor circuit capable of simultaneously processing, learning, and operating smoothly. The team fabricated a circuit of Hf0.5Zr0.5Ο2-based synstors as part of their study.
This custom-built circuit featured synaptic resistors integrating ferroelectric HfZrO materials. Notably, ferroelectric materials, like Hf oxides, provide engineers with the ability to create scalable, Si-compatible, nonvolatile analog memory and neuromorphic devices.
Drone Simulation: Human vs AI vs Synstor
The team set up a drone flight simulation and had a traditional AI, a human, and a super-turing AI system navigate the simulation to test their theory. Before any turbulence or obstacles, the synstor circuit, human operators, and ANN successfully learned to drive the drone toward its target position without any problems or delays.
However, when trees or strong winds were added to the equation, the performance of all parties decreased. Specifically, the human and synstor circuit AI system completed the task. The traditional AI system was unable to adjust to the new environment and didn’t complete the test as the added obstacles reduced its performance.
AI Adaptability in Action: The Drone Obstacle Course Results
The study showed that the engineers’ newly developed synaptic resistor circuits support efficient and adaptive Super-Turing AI systems that can navigate uncertain and dynamic real-world environments. They examined why only the human and synstor-powered drone could complete the course. Their success was due to their ability to adapt on the fly.
Why Synstor Circuits Are a Leap Forward in Sustainable AI
There are several benefits that a synstor circuit AI system could bring to the market. For one, they will open the door for more advanced and capable AI options. These protocols will be capable of concurrent real-time inference and learning, providing a stepping stone into next-generation AI protocols.
Eco-friendly: The engineer’s AI system architecture utilizes a fraction of the electricity required to run traditional AI protocols. This upgraded power usage allows Synstor AI systems to claim the title as the most eco-friendly AI architecture for data. As demand for these services increases, the need for sustainable options will also climb.
Efficiency: The engineers successfully created a system that mimics the human brain. Their approach reduces energy requirements across the board for AI systems and allows AI systems to maximize their power consumption when executing or training new models.
Adaptability: One of the main benefits of this AI structure is its adaptability. AI systems need to become faster and more adaptable to handle the growing number of tasks handed to them. From piloting your EV home to deciding what songs to play for you, future AI systems will take into account their environment as part of their processing procedures.
Future Applications: Brain-Like AI in Cars, Wearables, and Military Tech
The brain-inspired AI system has the potential to revolutionize the industry. It offers engineers the prospect of sustainable AI development and could support the creation of tomorrow’s advanced autonomous systems. This technology will also find its way into robotics, wearables, and the medical sector.
Advanced Computing: You can expect to see advanced AI systems utilizing this tech shortly. The reduced power consumption will reduce operating costs for these systems, making them an attractive option to the status quo. Additionally, as the synstor chip design improves, it will drive further adoption.
Military: The military is one of the biggest advocates of AI. AI systems can assist in tracking, targeting, planning, and logistics, enhancing the operational capabilities of any army utilizing them. In the future, you will see AI systems powering advanced loitering munitions and drones and helping planners maintain situational awareness..
Autonomous Vehicles: One of the core use case scenarios for advanced AI is piloting autonomous vehicles. The dream of fully autonomous cars has been slow to take hold, but recent advancements have made the vision closer than ever. This latest AI breakthrough will help improve performance and open the door for more advanced autonomous vehicles capable of flight and more.
This technology will one day help to make communities safer and your everyday tasks easier. However, there are still some details and hurdles that the tech must overcome before it goes mainstream. Despite the holdups, potential commercial adoption is estimated within 5–10 years, with nearer-term applications in specialized low-power AI tasks.
Meet the Research Team Behind This AI Breakthrough
Engineers from the Texas A&M Department of Computer and Electrical Engineering and the University of California-Los Angeles collaborated to bring this study to light. Specifically, Jungmin Lee, Rahul Shenoy, and Atharva Deo led the research.
They had help from a team of engineers, including Sun Yi, Dawei Gao, David Qiao, Mingjie Xu, Shiva Asapu, Zixuan Rong, Dhurva Nathan, Yong Hei, Dharma Paladugu, Jian-Guo Zheng, J. Joshua Yang, R. Stanley Williams, Qing Wu, and Yong Chen. Notably, the study received funding from the Air Force Office of Scientific Research.
The Business Angle: How NVIDIA and Others Are Advancing Neuromorphic AI
The AI race is on, and competing firms are spending billions to try and dominate the sector. The relatively disruptive nature of AI means that there are new opportunities for firms like OpenAI to come from nowhere and secure a top spot in the market. While it’s very difficult to determine who the next new player in the sector will be, it’s easy to see who has dominated the market for the last few years.
NVIDIA (NVDA -7.81%) entered the market in 1993 and is based in California. The now world-famous 3D graphics card maker pioneered ASIC and other game-changing designs over the years. These upgrades helped the company rebrand from a gaming graphics card firm into one of the most recognizable names in the AI sector.
NVIDIA Corporation (NVDA -7.81%)
NVIDIA’s rise didn’t go unnoticed. Its stock has seen considerable growth, and in June 2024, the firm hit a $3T market cap. Today, NVIDIA is a leader in neuromorphic and edge AI hardware. As such, it remains heavily involved in designing AI inference efficiency and brain-inspired architectures. Those seeking a reliable and proven AI stock should take the time to do more research into NVIDIA.
For a closer look at NVIDIA, make sure to look over our spotlight of this industry giant.
Latest on NVIDIA Corporation
The Future of AI: Natural Intelligence, Artificially Engineered
The engineers behind this study have unlocked a new level in AI capability. Their new chip design will allow AI to develop faster and smarter algorithms. These systems will one day power nearly every component in your life. For now, this research demonstrated how scientists can find inspiration from nature and the human body to drive innovation to new heights..
Learn about other cool AI projects now.
Studies Referenced:
1. Lee, J., Shenoy, R., Deo, A., Yi, S., Gao, D., Qiao, D., Xu, M., Asapu, S., Rong, Z., Nathan, D., Hei, Y., Paladugu, D., Zheng, J.-G., Yang, J. J., Williams, R. S., Wu, Q., & Chen, Y. (2025). HfZrO-based synaptic resistor circuit for a Super-Turing intelligent system. Science Advances, 11(9), eadr2082. https://doi.org/10.1126/sciadv.adr2082