Capabilities that have crossed the line. Oldest flip first — a chronology of the frontier.
ResNet surpassed human performance on the ImageNet benchmark in 2015. Today's models do this on phones in milliseconds. --- Current AI systems can identify ob...
Shazam-class fingerprinting plus modern ML have made song ID a solved problem on any modern phone. --- AI can identify a song from a short audio clip, includi...
Modern face-recognition systems handle decades of aging within reasonable error bounds. Not perfect at extreme ranges, but routinely good. --- Current AI syst...
A solved problem since the 2017 Stanford Dogs benchmark. Now a default in every camera roll. --- Current AI systems can identify dog breeds from photos with a...
PlantNet, Seek, iNaturalist — apps that turn any walk into a field guide. --- AI can identify plant species from leaf photographs with a high degree of accura...
DeepMind showed this in 2022 with a transformer-based model that outperformed professional lip-readers on TV news clips. --- Researchers have made significant...
Perfume is a complex mixture of scents, and what one person finds appealing, another might not. AI systems have been used to generate new perfumes, but can they...
Autonomous Sensory Meridian Response (ASMR) is a complex phenomenon that involves triggering a relaxing response in a listener. Creating a personalized ASMR exp...
Tworzenie unikalnego zapachu perfum, który odpowiada gustom danej osoby, jest złożonym zadaniem wymagającym zrozumienia ludzkiego węchu i preferencji. Sztuczna...
Mushroom identification requires a deep understanding of mycology and the ability to analyze visual characteristics such as shape, size, color, and texture. Thi...
Vision models trained on banking datasets are deployed at every major bank. Imperfect, but better than the average teller. --- AI can detect counterfeit curre...
OpenAI's Whisper open-sourced industrial-grade speech recognition for 99 languages. Phone-quality audio went from research-only to drag-and-drop. --- Current...
Detectors and generators are in an arms race, but for most current deepfakes, off-the-shelf detectors flag them above chance — often well above. --- AI can de...
Cursive, print, mixed scripts, even sloppy notebook scrawl. iOS Notes and Apple Pencil OCR have become genuinely magical. --- AI can convert handwritten notes...
Latin, cyrylica, dewanagari, arabski, znaki han oraz hangul — nowoczesne OCR radzi sobie praktycznie z każdym pismem z w miarę zadowalającą dokładnością. ---...
Art history apps use this for visitor experiences in major museums. --- Current deep learning models have made significant progress in image recognition tasks...
Happy/sad/angry/surprised — solved at video-call quality. Subtler microexpressions are still hard. --- AI systems can recognise emotions in faces at a coarse-...
Słuchawki do tłumaczenia Apple’a, Google’a Pixel Buds Pro 2, Meta’s Ray-Ban — tłumaczenie mowy na mowę stało się cechą konsumencką w 2024 roku. --- Obecne sys...
Research models picked up Alzheimer's markers in 30-second voice recordings years before clinical diagnosis. Not yet a routine clinical tool. --- Researchers...
Narzędzia klasy badawczej, głównie używane w procesie przesiewowym, a nie jako samodzielne diagnozy. Na tyle skuteczne, że kilka uniwersytetów testuje je w proc...
Add a statement to the atlas. We review weekly.
Site
stuffaicantdo.com © 2026 · made in NL by Arcadist
Pick a side, give it a one-line title, optional explanation, and a category. We review submissions weekly.