Listen to the Podcast:
New materials for carbon capture and wind turbines could minimize pollution. A small but growing number of companies want to use AI to help them grow faster.
Companies have promoted numerous ways that artificial intelligence can make our lives easier since ChatGPT went viral last fall. They have promised superhuman virtual assistants, tutors, lawyers and even doctors.
What about a superhuman chemical engineer?
Orbital Materials, a London-based startup, wants to achieve exactly that. The company is trying to use generative AI, the technology behind tools like ChatGPT, to speed up the development of clean energy products. Essentially, the goal is to create computer models that are powerful and sharp enough to discover the optimal formulations for things like sustainable jet fuel or batteries free of rare-earth materials.
Jonathan Godwin, co-founder of Orbital Materials, wants a system as simple and efficient as the one software engineers already use to develop designs for things like furniture and airplane wings.
“Historically, that has been too difficult for molecular science,” he commented.
ChatGPT works because it’s good at predicting text: here’s the next suitable word or sentence. For the same concept to work in chemistry, an AI system must predict how a new molecule will act in the real world, not just in a laboratory.
Researchers and companies have used artificial intelligence to search for better, greener materials. Symyx Technologies, a materials discovery company founded in the 1990s, closed its doors after a sale. More recently, petrochemical alternatives and cell programming have gained ground.
However, the technology for many items needed to decarbonize the world is not yet available.
Sophisticated new materials can take decades to make their way from discovery to sale. That timescale is too slow for companies and nations looking to cut emissions quickly to reach net-zero ambitions.
According to Aaike van Vugt, co-founder of the materials science company VSParticle, “That should happen in the next ten years, or sooner.”
AI researchers think they can help. Godwin spent three years at DeepMind, Google’s artificial intelligence unit, researching advanced materials discoveries before founding Orbital Materials. That lab published AlphaFold, a model for predicting protein shapes that could help discover new drugs and vaccines. This, combined with the rapid adoption of tools like ChatGPT, convinced him that AI would soon be able to dominate the physical world.
“What I thought would take ten years happened in 18 months,” he said. “Things are getting better and better.”
Godwin compares his orbital materials method to AI imagers like Dall-E and Stable Diffusion. These models are built from billions of photos online, and when users type in a text message, a photorealistic creation appears. Orbital Materials intends to train models on the molecular structure of materials using massive amounts of data. Enter a desired characteristic and material—an alloy that can withstand extremely high temperatures—and the model returns a proposed molecular formula.
According to Rafael Gomez-Bombarelli, an assistant professor at MIT who advised Orbital Materials, this approach is effective in theory, since you can imagine new molecules and test how they would work. (He said that he is not an investor).
Using AI to develop climate-friendly materials
Many IT investors are looking for startups that can benefit from producing greener materials. This is especially true in Europe, where regulators require companies to reduce carbon emissions or risk heavy fines. In the future, markets for sophisticated materials in areas such as renewable energy, transportation, and agriculture are expected to increase by tens of billions of dollars.
Some researchers, such as those at the University of Toronto, have set up “autonomous labs” that combine artificial intelligence systems with robots to search for new materials at unprecedented rates. A Dutch company called VSParticle makes equipment to make parts for renewable gas and hydrogen sensors.
According to co-founder van Vugt, consider it comparable to a DNA sequencer in a genomics lab. He believes his technology can help reduce the 20-year time horizon of advanced materials to a year and eventually “a couple of months.” His business is currently seeking financing to grow.
Orbital Materials, which has secured $4.8 million in previously undisclosed seed funding, intends to start by focusing its AI on carbon capture. The company is developing an algorithmic model that can more efficiently separate CO2 and other harmful chemicals from other pollutants than current methods.
(According to Godwin, the startup, which employs numerous AI experts, intends to publish peer-reviewed results on the technology soon.) Carbon capture isn’t working at scale yet, but thanks to a spate of government incentives, particularly in the United States, interest in implementing the technology is growing rapidly.
Godwin stated that Orbital Materials would eventually aim to expand into areas such as fuel and batteries. He envisions a business model similar to synthetic biology and drug development companies: build the brainpower, then license the innovative software or materials to manufacturers. “It will take us a bit of time to get to market,” Godwin admitted. “But once you’re there, everything happens very quickly.”
However, perfecting the AI is only half the battle. Manufacturing sophisticated materials in the fuel and battery manufacturing fields requires the collaboration of large, established companies and tangled supply chains. According to MIT’s Gomez-Bombarelli, this can be even more expensive than inventing new drugs. “The economics and de-risking make it much more difficult,” he explained.
According to Heather Redman, managing partner at Flying Fish Partners, which backed Orbital Materials, most tech investors looking for the bright penny of generative AI have been unable to see beyond chatbots. She acknowledges the dangers associated with energy businesses, but feels the potential for $1 trillion of products like batteries and carbon capture is worth the investment risk.
“We love the big hills as long as there is a big, gigantic market and opportunity at the top,” he explained.
Gomez-Bombarelli understands how big these hills can be. In 2015, he helped found Calculario, a startup similar to Orbital Materials that used AI and quantum chemistry to accelerate the development of new materials. He didn’t get enough traction and was forced to focus on the OLED market.
“Maybe we didn’t make our case,” he said. “Or maybe the market wasn’t ready.”
It is debatable if it currently is. However, there are some positive indicators. Computer science has certainly advanced. Newcomers may also have an easier time selling AI because potential customers can see the benefits more clearly. Gomez-Bombarelli’s pitch is simple: “Look at ChatGPT.” We can apply the same logic to chemistry.”
Also Read: Google’s Next-Gen Gemini AI
Subscribe to our latest newsletter
To read our exclusive content, sign up now. $5/month, $50/year
Categories: Technology
Source: vtt.edu.vn