AI Improvements to Material Science Are Worth Getting Excited About
Over the last 6 to 9 months, AI has taken the world by storm. Normal people have started using AI to create art and write content. AI has become immensely popular, but people have forgotten that AI isn’t just for generating words and pixels on a screen. AI helps us solve important real-world problems we face in energy storage, computer chip manufacturing, and consumer goods.
Electric car batteries are hugely expensive, take a long time to recharge, provide shorter ranges than gas, and depend on foreign-sourced rare earth materials.
Computer chip manufacturing has gotten a lot harder. Fundamental limits of physics and the way our current chip manufacturing processes work with them no longer allow us to regularly shrink down chip sizes and increase transistor counts as we have for the last several decades. And given our complex computer chip manufacturing processes, we heavily rely on foreign manufacturing plants and foreign-sourced rare earth materials to build them.
Consumer goods such as non-stick cookware and stretch-resistant and resilient clothing produce toxic byproducts and negative hormonal effects respectively that consumers actively want to avoid.
All these problems can be boiled down to material science issues. By finding better material components, better complementary materials, or improving current materials manufacturing process, we can solve them, but the current materials science research process has its issues.
Material science research is currently a giant bottleneck for innovation
Finding materials with the right mix of special properties to use in important consumer and industrial applications using the current materials science research process takes a lot of time.
Materials science involves the study of complex systems, often at the atomic or molecular level. This can make it difficult to fully understand and predict the properties and behavior of materials.
Materials science often involves conducting experiments to test and verify theories and predictions. These experiments can be time-consuming, particularly if they involve synthesizing and characterizing new materials.
Materials science experiments often generate large amounts of data that need to be analyzed and interpreted. This can be a slow and labor-intensive process, particularly if the data is complex or difficult to understand.
AI models for material science dramatically speed up the materials research process
State of the art AI models are fantastically good at identifying and recognizing patterns in huge datasets of raw data. That makes AI models a great fit for modeling the structure and behavior of alternative materials and detecting the important differences so that we can identify and implement better alternatives.
Replacing current materials science research processes with new processes aided by the use of AI reduces the length of new materials search from weeks, months, or even years to mere seconds.
Innovation in energy storage has been elusive. For the last 100 years, researchers have performed a long series of experiments and while they have made substantial progress, we have never gained a fundamental understanding of how energy storage and battery technology works as the details of these micro-level interactions remained hard to measure.
Energy storage technologies, such as batteries, are still relatively expensive compared to other forms of energy generation and storage, such as fossil fuels.
Batteries degrade over time, which reduces their efficiency and lifespan. Shorter lifespans prevent many battery technologies from being used in grid-scale energy storage systems expected to operate over long periods of time.
Some energy storage technologies aren't scalable, which makes it difficult to integrate them into electric grids or meet the needs of large industrial facilities.
Many battery technologies, such as lithium-ion, pose safety risks. If they are not designed and operated properly, batteries can catch fire and potentially explode when they're damaged or charged and discharged too quickly.
Developing a better understanding of the structure and behavior of battery ingredient materials is key to producing higher capacity batteries, faster charging cycles, and ultimately switching to cheaper and more sustainable materials.
AI helps us model out and understand the micro-level behavior of battery technology materials and reduce the number of defects in the battery manufacturing process. By reducing defects, we can create longer lasting batteries. By modeling alternative battery materials, we can move away from dangerous technologies like lithium-ion to safer technologies such as solid-state electrolytes.
Computer chip manufacturing
We've been able to reduce the size of computer transistors as measured by die sizes every 18 to 24 months for decades, but reducing computer chip die sizes below 5nm has proven particularly challenging because of quantum mechanics.
Smaller computer chips are more efficient and use less power. However, as chips continue to shrink in size, it becomes increasingly difficult to manufacture them with the same level of precision and reliability.
As transistors get smaller, they start to exhibit quantum mechanical effects, which makes it difficult to design transistors that are both small and reliable.
As chip technology advances, it becomes increasingly difficult to produce chips that are fast, efficient, and have low power consumption. This requires constantly improving manufacturing processes and developing new materials and technologies.
Modern computer chips are highly complex with billions of transistors and other components packed into a small area. This complexity makes it difficult to design and manufacture chips that are reliable and free from defects.
AI helps us build smaller transistors by modeling alternative structures and materials that can reliably communicate information at smaller die sizes where quantum mechanics become important. AI helps find alternative materials that are less dependent on foreign-sourced rare earth materials. AI also helps us design computer chips beyond the complexity of traditional computer chip design processes that dramatically improve performance and energy efficiency.
Cooking is an essential avenue to health and yet popular non-stick cooking materials seem to make a bad tradeoff between health and utility.
Non-stick coatings, such as those made with perfluoroalkyl substances (PFAS), have been linked to cancer and reproductive problems. This has led to concerns about the safety of non-stick cookware, particularly when used at high temperatures or when the cookware is scratched or damaged.
Producing non-stick coating releases toxic chemicals into the air and water. Disposing of non-stick cookware poses environmental risks as well.
Non-stick coatings wear off and damage over time, which reduces their effectiveness and lifespan.
Non-stick cookware isn't suitable for high-heat searing or grilling, which damages the non-stick coating and produces harmful fumes.
Non-stick cookware requires special care in order to maintain its non-stick properties, such as using gentle cleaning methods, avoiding metal utensils, and avoiding high heat.
AI can help us model and test alternative non-stick cooking materials resilient to modern cooking techniques, hold up to cooking and baking temperature changes and utensils, and are safe throughout their lifetime of use. AI modeling for non-stick cooking materials will help us make cooking safe while still keeping it easy.
Stretch-Resistant and Resilient Clothing
Most clothing used today is made using some blend of synthetic fabrics. Polyester, nylon, acrylic, and microfiber are made from plastic, which is created from oil.
Evidence suggests that a number of synthetic fabrics such as polyester interfere with the production of human hormones.
One study found that men who wore polyester-cotton blend underwear had lower levels of free testosterone compared to those who wore cotton underwear. Researchers suggested that this may be due to the estrogen-like effects of certain chemicals used in these synthetic fabrics' production.
Synthetic fabrics, such as polyester and nylon, can release harmful chemicals when they break down or are washed. These chemicals can have negative impacts on human health, particularly if they are inhaled or absorbed through the skin.
Plastic-based clothing, particularly synthetic fabrics, can shed tiny plastic fibers during washing and wear. These microfibers can end up in the environment, where they can be ingested by wildlife and potentially enter the food chain.
AI can help us model and find non-estrogenic, non-hormone disrupting, and non-toxic clothing materials. AI can also help us find better processes to produce safe clothing that offer the utility of today's synthetic materials without the negative health effects.
The Path Forward
While AI currently provides consumers with the ability to generate art and content, breakthroughs it can help us make in material science are far more compelling and have the potential to greatly benefit society as a whole.
Developing new materials, better understanding them, and making better predictions of their behavior leads to significant cost savings, improved consumer safety, and dramatically speeds up scientific discovery. These advancements are far more important and impactful than the fun that consumers currently get from using AI to generate art and content.
Science's next generation of innovation will come from applying methods AI researchers used to train and build today's art and text generation models to generate lists of new alternative materials with dynamic properties we can use to improve energy storage, computing chip manufacturing, and consumer goods.