Real problems, solved with AI — with the evidence attached.

A directory of 1,138 documented cases where artificial intelligence contributed to actually solving a problem — in medicine, biology, climate, physics, engineering, software and everyday life. Every entry carries its primary source, a verification status, and an explicit rating of how much the problem itself matters.

12 categories · spanning 2013–2026 · 125 high-importance cases · mean importance 5.7/10 · updated hourly from the source sheet

Highest-rated problem in each field

• 10/10

  Multimodal deep learning of fundus abnormalities and traditional risk factors for cardiovascular risk prediction

  Medicine & Health · cardiology · 2023 · AI-assisted progress · confidence 88%

  Cardiovascular disease, the leading global cause of death, needed better risk identification from accessible data.

• 9/10

  Evolutionary-scale prediction of atomic-level protein structure with a language model

  Biology & Life Sciences · protein science · 2023 · Problem solved · confidence 95%

  Predicting the 3D structures of the exponentially growing catalog of protein sequences was too slow with existing methods, requiring a breakthrough in folding speed.

• 8/10

  Deep Residual Learning for Image Recognition

  Software & Coding · computer vision · 2016 · Problem solved · confidence 95%

  Very deep neural networks were difficult to train, blocking accuracy gains from increased depth in visual recognition.

• 8/10

  Learning skillful medium-range global weather forecasting

  Climate & Environment · weather forecasting · 2023 · Problem solved · confidence 92%

  Producing accurate 10-day global weather forecasts, traditionally requiring heavy compute rather than learning from historical weather data.

• 8/10

  Scaling deep learning for materials discovery

  Engineering & Materials · materials discovery · 2023 · Partially solved · confidence 85%

  Discovery of inorganic crystals for applications from microchips to batteries and photovoltaics has been bottlenecked by expensive trial-and-error approaches.

• 7/10

  Machine-learning-assisted correction of correlated qubit errors in a topological code

  Physics & Astronomy · quantum computing · 2018 · AI-assisted progress · confidence 90%

  Fault-tolerant quantum computation requires efficient detection and correction of errors in encoded quantum information, and standard decoders miss correlations between bit-flip and phase-flip errors.

• 7/10

  Neural‐Symbolic Machine Learning for Retrosynthesis and Reaction Prediction

  Science & Research · computational chemistry · 2017 · Partially solved · confidence 88%

  Rule-based expert systems for reaction prediction and retrosynthesis often fail because they ignore molecular context, causing reactivity conflicts.

• 7/10

  Discovering faster matrix multiplication algorithms with reinforcement learning

  Mathematics · scientific computing · 2022 · Partially solved · confidence 85%

  Discovering more efficient matrix multiplication algorithms is hard because the space of possible algorithms is enormous and human intuition is limited.

Full importance ranking →

Most recent

• 4/10

  Machine Learning-Assisted Polymer and Polymer Composite Design for Additive Manufacturing

  Engineering & Materials · additive manufacturing · 2026 · AI-assisted progress · confidence 65%

  Predicting the final properties of 3D-printed polymer and polymer composite parts is difficult because material composition, processing conditions and microstructure vary.

• 5/10

  Deep learning–based approaches for weed detection in crops

  Biology & Life Sciences · precision agriculture · 2026 · AI-assisted progress · confidence 72%

  Detecting weeds in crops is difficult for traditional machine vision, which lacks robustness, scalability and recognition accuracy in variable field conditions.

• 6/10

  Challenges in Testing Large Language Model-Based Software: A Faceted Taxonomy

  Software & Coding · software testing · 2026 · Partially solved · confidence 70%

  LLM-based software is non-deterministic, so traditional testing cannot verify correctness within individual test cases.

• 5/10

  LLM-assisted systematic review of large language models in clinical medicine

  Medicine & Health · medical research · 2026 · AI-assisted progress · confidence 60%

  The rapid flood of clinical LLM studies (roughly 3.2 papers per day) makes manual curation and review of the evidence base infeasible.

• 6/10

  Implementation of large language models in electronic health records

  Medicine & Health · clinical documentation · 2025 · AI-assisted progress · confidence 80%

  Electronic health records have increased clinicians' documentation workload, and real-world LLM deployments in live hospital systems remain limited.

• 4/10

  Transforming Surgical Training With AI Techniques for Training, Assessment, and Evaluation: Scoping Review

  Medicine & Health · surgical training · 2025 · AI-assisted progress · confidence 70%

  Surgical training assessment and evaluation rely on traditional methods that may be improved by data-driven AI techniques.