Thinkorithm

Four Applications of AI, Explained

Reading handwriting, supporting medical imaging, predicting proteins and forecasting weather — what the systems do and where people remain in the loop.

Published · 2 min read

Behind the headlines, much useful AI is quiet and specialised. The four examples below are documented in official or peer-reviewed sources, and each shows a different way that learned models and human experts work together.

1. Reading handwriting

One of the earliest large-scale uses of neural networks was reading handwritten characters. A 1998 paper by Yann LeCun and colleagues describes a system based on convolutional networks that was deployed commercially and read several million bank cheques per day. The task was narrow, the data plentiful and errors could be caught by further checks — a pattern that still predicts where AI works well.

2. Supporting medical imaging

The US Food and Drug Administration maintains a public list of AI-enabled medical devices authorised for marketing in the United States; it now runs to more than a thousand entries. Many are in radiology: software that helps detect findings on scans, or that flags time-sensitive images so that a clinician reviews them sooner. Such tools are cleared for specific intended uses, and the clinical decision remains with medical professionals.

3. Predicting protein structures

A protein’s function depends on its three-dimensional shape, which has traditionally been determined through slow and expensive experiments. AlphaFold predicts that shape from the protein’s amino-acid sequence. The 2024 Nobel Prize in Chemistry recognised this work; according to the Nobel committee, its developers have predicted the structures of almost all of the roughly 200 million known proteins. Predictions are a starting point for researchers, who still confirm important structures experimentally.

4. Forecasting the weather

Weather forecasting has long relied on physics-based simulations running on supercomputers. Since 2025, the European Centre for Medium-Range Weather Forecasts has also run a machine-learning model, AIFS, operationally. ECMWF reports that it is faster and uses far less energy per forecast, and that in its 2025 evaluation it outperformed the physics-based model for most surface and upper-air variables in the medium range — while the physics-based model remains essential for high-resolution detail.

Sources

  1. LeCun, Y., Bottou, L., Bengio, Y., Haffner, P. Gradient-Based Learning Applied to Document Recognition. Proceedings of the IEEE, 86(11), 1998.
  2. U.S. Food and Drug Administration. Artificial Intelligence-Enabled Medical Devices.
  3. Nobel Prize Outreach. The Nobel Prize in Chemistry 2024 — Popular information.
  4. ECMWF. ECMWF’s AI forecasts become operational, 2025; 2025 forecast evaluation met with positive feedback, 2026.