Deep Learning for a Deeper Understanding of Our Earth
News:
NVIDIA: Earth Climate at Groundbreaking 2-Kilometer Resolution
Deep learning and process understanding for data-driven Earth system science
A generalizable and accessible approach to machine learning with global satellite imagery
Machine learning in Earth and environmental science requires education and research policy reforms
Towards neural Earth system modelling by integrating artificial intelligence in Earth system science
70 years of machine learning in geoscience: a review
Advancing AI for Earth Science: A Data Systems Perspective
Artificial intelligence alone won’t solve the complexity of Earth sciences
Deep learning and process understanding for data-driven Earth
Machine learning in geosciences and remote sensing: The next generation in Geosciences
Advance in AI/DL
History
Extend of AI for Earth
View on AI
We are bound to Earth by bio bodies—fragile, short-lived.
For thousand years, we have dreamed of transcending them, replacing them with a physics being—a robot, or a hyper-intelligence—a God form.
While our bio body and brain seem arrive their limits after millennia, progress is an illusion on both.
Naturally, the key to hyper-intelligence lies in another intelligence, perhaps the artificial intelligence, a form of intelligence unlike our own, with a future shape we can barely imagine to join both dreamed parts perfectly.
This vision fuels both our excitement and fear, much like the awe and anxiety when the first steam train roared into life.
Again today, we view it with high expectations, fear its harm, witness its force, then regulate it with laws. A new power brings revolution, expecting to do everything—until it doesn't, leading to disillusionment and doubt. The same pattern unfolds now with AI.
We’re at the start of a new era, like with inventions of airplanes or spaceships that came after trains, of course with laws evolving to keep pace. The focus isn’t on AI’s importance—it’s clear—but on how to use it safely, now the center of every discussion.
In tech, however the issue is critical: we’ve hit the limits of algorithms proposed 50 years ago. Once succeeded, soon exhausted.
Most applications rely on massive computation rather than truly advancing intelligence. Quantum computing likely dominates next computing force, offering a boost to today’s costly training processes. However, it’s still immature.
We await breakthroughs in algorithms still inspired by the workings of bio brains.
Breaking ground takes time, effort, intelligence, and perhaps luck—along with major investments, tests, and talent, much like what Altman seeks. Hyper-intelligence “the big word” seem essential, though many prioritize security above all.
In years coming, another "warm winter" should arrive, as some researchers predict—a welcome balance after the recent overheated hype.
Yet, AI pathway is forward.
Tan Minh VU, Nov 2024
Sources
Pattern recognition and machine learning (Bishop 2006)
Linear analysis (Bollobás 1999, Cambridge University Press)
All of statistics (Wasserman 2013, Springer)