AI Velocity: How Artificial Intelligence Accelerates Everything
Where AI dwells, innovation follows.
As you know, it’s been a few months since the last post on WattMind. I want to thank everyone who has been reading and reached out with comments and suggestions. It is greatly appreciated!
Going forward, I will be broadening the subject of WattMind beyond the intersection of AI and energy. Why? Simply put, I don’t want to limit my writing to energy or AI only. There’s so much going on in the world in all domains - technology, culture, politics, and so much more. I enjoy thinking and writing about all of these things!
So going forward - after this article - you can expect a variety of topics in this publication, which will of course continue to include Artificial Intelligence. Because AI is not going away, it is in fact accelerating everything.
That is what I want to talk about here - a general framework for how AI advances technological progress in any domain it touches.
AI Velocity: The Rate of Rapid Machine Learning
If you’re working in a domain with thousands of possible inputs that could impact the capability of the final product, it can take years to find the best combination of inputs.
Enter AI. Regardless of application, one of the major benefits of AI with robust compute is the ability to test and iterate much faster than any human can. I call this idea AI Velocity.
AI Velocity refers to the unprecedented pace of experimentation, testing, and iteration possible through artificial intelligence. This comes to life when AI is applied to a range of problem-solving scenarios, vastly outstripping the speed and accuracy of human efforts.
Take, for instance, the development of new pharmaceuticals. Traditionally, this process involves countless hours of trial-and-error tests, with researchers manually examining individual combinations of chemical compounds for potential efficacy. However, with AI Velocity, hundreds of thousands of combinations can be evaluated almost instantaneously, making the discovery of viable drug combinations significantly faster and more efficient.
The wind energy sector is also substantially advanced by AI Velocity. Conventionally, wind turbine design is a tedious, long-drawn process involving extensive simulations for each design change.
But with AI Velocity, machine learning algorithms swiftly evaluate numerous design alterations, accurately foreseeing their impact on turbine performance. This results in much more efficient wind turbines getting developed much faster than traditional methods, thereby expediting the shift towards renewable energy sources.
By introducing the concept of AI Velocity into these diverse fields, we're amplifying our ability to test, iterate, and ultimately reach our goals faster than ever before.
What Drives AI Velocity?
AI Velocity is driven by a combination of machine learning techniques that exponentially speed up testing and refining processes. Here's how each technique plays its part:
Supervised Learning
This is the training phase where the AI learns from a dataset. It's as if the AI is being tutored, using labeled data to learn a function that can be applied to new data. This speeds up AI's ability to understand and predict outcomes based on input variables.
Reinforcement Learning
In this process, the AI learns by interacting with its environment and receiving rewards or penalties based on its actions. This allows the AI to learn optimal behaviors and solutions, thus quickly improving its performance.
Optimization Algorithms
Once the AI has learned, optimization algorithms help fine-tune its performance. They guide the AI in finding the most effective variables or model parameters, improving the quality of its output. This contributes to the speed of iterating and refining solutions.
Deep Learning
Involving deep neural networks, this technique allows AI to learn from vast amounts of data in a way that mimics the human brain. This accelerates the AI's ability to analyze complex, unstructured data and derive meaningful insights quickly.
Generative Adversarial Networks (GANs)
These are like the brainstorming and refining teams of the AI world. The Generator comes up with new ideas (or data), and the Discriminator judges if these ideas are convincing or not. They constantly challenge each other, leading to better and more realistic solutions in less time, thereby adding to the AI's speed of innovation.
AI Velocity in the Energy Industry
This framework is the key driver of innovation in the energy sector. Whether it’s for solar cell optimization, EV material inputs, or fusion-reactor designs, the speed of testing and potential breakthrough is accelerating.
Solar Cell Development
Researchers at MIT and Google Brain have developed a system called a differentiable solar cell simulator that makes it possible to evaluate multiple proposed designs at once and provide information about which changes will provide the desired improvements. This could greatly increase the rate for the discovery of new, improved configurations. The system is available as an open-source code that anyone can use and contribute to. It can simulate one-dimensional solar cells with multiple layers and different properties.
EV Development
MATLAB, Simulink, and Simscape enable engineers to front-load the development of EVs through the systematic use of data and models.
Engineers are able to design complex EV architectures and optimize systems, model batteries and develop battery management systems (BMS), model fuel cell systems (FCS) and develop fuel cell control systems (FCCS), model traction motors and develop Motor Control Units (MCU), deploy, integrate, and test control algorithms, use data-driven workflows and artificial intelligence (AI) in EV development. More info on this process here.
This holistic approach allows for the virtual creation and full testing of new electric vehicle designs. Doing this on a computer reduces costs to manufacture and test by an order of magnitude.
Beyond the Present: AI Velocity’s Potential
In some ways, it’s hard to imagine where AI velocity will take us. One breakthrough could lead down a previously unknown path, unlocking some impossible technology (maybe room-temperature superconductors!).
But we can speculate - based on some early signs of progress.
Nuclear Fusion
DeepMind collaborated with the Swiss Plasma Center to train a deep reinforcement learning algorithm to control the plasma inside a nuclear fusion reactor.
Plasma is a superheated state of matter that requires extremely high temperatures and magnetic fields to sustain nuclear fusion, which produces clean and unlimited energy.
DeepMind used two neural networks, a critic and an actor, to learn how to adjust the magnets in a virtual and a real reactor. The AI achieved zero resistance and partial levitation of the plasma.
This process can help physicists explore different plasma conditions and reactor configurations, and potentially accelerate the development of commercial nuclear fusion.
Turbulence Tracking in Fusion Reactors
The AI impacts on fusion energy go beyond plasma control.
Researchers have demonstrated the use of computer vision models to identify and track turbulent structures, called blobs, that appear in plasma created in controlled-nuclear-fusion experiments. They created a synthetic video dataset of plasma turbulence to train four computer vision models, which can draw boundaries around blobs in the same ways that humans would.
The trained models are then tested using real video data from experiments. It was found that blobs can be identified with high accuracy - along with an estimate their size and speed. This could provide more detailed information about the plasma boundary, helping to solve one of the many outstanding obstacles to creating a viable fusion reactor.
Battery Development
Researchers at MIT have developed a new kind of battery made entirely from abundant and inexpensive materials that could provide low-cost backup storage for renewable energy sources such as wind and solar.
Aluminum-sulfur battery architecture: The new battery uses aluminum and sulfur as its two electrode materials, with a molten salt electrolyte in between. The battery is safe, cheap, and resistant to dendritic shorting, a common problem in battery reliability.
Fast-charging and high-capacity performance: The battery cells can endure hundreds of cycles at exceptionally high charging rates, with a projected cost per cell of about one-sixth that of comparable lithium-ion cells. The battery can also maintain its operating temperature without external heat sources.
Potential applications and spinoff company: The new battery technology is ideal for installations of about the size needed to power a single home or small to medium business. It could also be used for electric vehicle charging stations. The technology is the basis for a new spinoff company called Avanti.
Looking Into the Future
The future of clean energy is bright - thanks to the power of AI Velocity. What comes next, no one knows for sure. But given where the current resources are flowing, I would expect fully functional and cheap perovskite solar cells this decade, along with battery breakthroughs that enable the elusive solar energy + storage paradigm we are entering.
Beyond that, a nuclear fusion breakthrough could come in the mid-2030s as more funding is thrown at this technology. Coupled with AI Velocity - the sky (or the sun) is the limit.
Chris Prato, creator of WattMind, combines 16 years of energy industry experience with a passion for clean energy and AI. With degrees in Mathematics, Philosophy, and an MBA, Chris offers unique insights into the world of AI-driven energy tech. Follow Chris on LinkedIn or Threads for more energy hot takes.