With many parts of the globe facing heatwaves, wildfires and floods, former Google CEO Eric Schmidt said the use of artificial intelligence (AI) could help predict extreme weather. I believe it can help. Semiconductor giant Nvidia’s experiment to build an AI-powered “digital twin” of the planet could be the first of many steps in this direction.
the concept of digital twin Represent scenarios in which actual, real-world products, processes, or systems serve as digital counterparts for simulation, testing, monitoring, and integration. While the concept itself is not new, the growing power of generative AI is fueling excitement about future possibilities.
A digital twin approach to solving climate prediction
What Nvidia is trying to do here is create a digital twin called Earth-2. It uses FourCastNet AI model weather forecasts that use massive amounts of data from the earth system to forecast the next two weeks in a faster and more accurate way. Schmidt believes such a system could potentially generate thousands of predictions instead of just 50 today.
These accurate disaster risk predictions could provide vulnerable people with valuable time to prepare and evacuate, Schmidt said, while climate modeling is just an early use case for AI. There is potential, and with sensible regulation and the right support for innovative uses, the challenges of science could be met. the most pressing issues.
AI can change the way science works
“Building a future where AI-powered tools will save us from mindless and time-consuming labor, lead us to creative inventions and discoveries, and facilitate breakthroughs that would otherwise take decades. You can,” he said in an article he wrote. Published in MIT Technology Review.
Schmidt, co-founder of Schmidt Futures, which bets on brilliant minds to make the world a better place, said the world limits AI with large-scale language models, but there are many different things that could have a greater impact. He states that there are many model architectures. Scientists have used AI models to identify antibiotics to fight pathogens. Google Deep Mind This model controls the plasma in fusion reactions and brings clean energy options to the world.
AI is not just about large language models
“Over the past decade, most scientific progress has come through small-scale ‘classical’ models focused on specific questions. These models are already making great strides. Recently, large-scale deep learning models that have begun to incorporate cross-domain knowledge and generative AI have expanded the possibilities,” he says.
Schmidt says that while the core of the scientific process is the same as what is taught in elementary school, AI starts with understanding the background and then learns about each component of scientific research, including identifying hypotheses, conducting experiments, analyzing data, and drawing conclusions. I think it has the potential to revolutionize the way we do things.
The scientific process itself is simplified
Scientific literature review is already simplified by AI tools like PaperQA and Elicit. The next step is to generate hypotheses over a wider range, narrow them down more quickly, and generate stronger options. He draws attention to the example of scientists at Caltech using his AI fluid simulation model to design better catheters that ward off infections.
In the experimental process, Schmidt said, AI could do these things faster, cheaper, and on a larger scale, while simultaneously AI-powered machines with hundreds of micropipettes running day and night, and human suggested that the process could produce samples at unmatched speeds. Instead of a scientist running a handful of experiments, with AI tools he can run 1,000 experiments, he added.
When it comes to analyzing and drawing conclusions, self-driving labs can go beyond automation and use large language models to interpret results or recommend alternative experiments. In fact, AI lab assistants can even order consumables to replace previously used ones and set up and run the next set of recommended experiments overnight.
While AI tools can lower the barriers to entry for new scientists and open opportunities for researchers who may have been left out of the process, he said AI can help recognize areas where human touch is still important. warned that the best way to make more effective use of .
