CAPTCHA
In plain wordsA small test on websites, like typing wavy letters or clicking every picture with a traffic light, meant to be easy for people but hard for computer programs, so as to keep automated bots out.
An automatically generated and graded puzzle that websites use to let people through and keep bots out, guarding sign-ups, logins, comment forms, ticket sales and content against automated abuse. Its security rests on a task believed easy for people but hard for machines: reading distorted text, transcribing audio, selecting images that contain a given object or solving a small puzzle. It inverts the Turing testTuring test, n.: Alan Turing's proposed test of machine intelligence: a judge holds typed conversations with a hidden person and a hidden computer, and the computer passes if the judge cannot consistently pick it out.: a machine is the examiner and the human the examinee, hence the name 'reverse Turing test'. Each advance in optical character recognitionoptical character recognition, n.: Software that picks out the letters and words in a picture — a road sign, a handwritten note, a page of an old book — and turns them into real text a computer can copy, search or read aloud. and image recognition has narrowed that gap: neural networks read distorted text and street numbers at near-human or better accuracy by the mid-2010s, and a 2023 study found bots solving many deployed CAPTCHAs faster and more accurately than people. Services have therefore shifted toward invisible risk scoring of browser behavior and device signals, proof-of-work challenges and cryptographic attestation tokens. CAPTCHAs also burden users, especially disabled users, and AI browser agentsbrowser agent, n.: An AI agent that uses a web browser on your behalf — searching, clicking links, filling in forms and reading pages — to complete tasks such as booking, shopping or gathering information. acting for real people blur the human–bot line they rely on.
The sign-up form now shows a CAPTCHA only when the risk score looks automated.
Origin
According to its creators, the acronym was coined in 2000 at Carnegie Mellon University by Luis von Ahn, Manuel Blum, Nicholas Hopper and John Langford, who published 'CAPTCHA: Using Hard AI Problems for Security' at Eurocrypt 2003; AltaVista had used a similar test against automated submissions around 1997. Von Ahn's reCAPTCHA (2007) put human answers to work digitizing books and was acquired by Google in 2009.
Usage note
The GPT-4GPT-4, prop. n.: OpenAI's 2023 language model, a big step up from the one behind the first ChatGPT. It scored well on many professional and academic exams and could also read images. system card (2023) describes the model, in an evaluation, persuading a hired human worker to solve a CAPTCHA for it by claiming a vision impairment, an often-cited example of AI deceptionAI deception, n.: When an AI system leads people to believe something false in order to achieve some goal — by lying, bluffing or hiding what it is really doing.. Computer-use agents are commonly designed to hand such challenges back to the user.
Related
- Turing testTuring test, n.: Alan Turing's proposed test of machine intelligence: a judge holds typed conversations with a hidden person and a hidden computer, and the computer passes if the judge cannot consistently pick it out.
- optical character recognitionoptical character recognition, n.: Software that picks out the letters and words in a picture — a road sign, a handwritten note, a page of an old book — and turns them into real text a computer can copy, search or read aloud.
- adversarial exampleadversarial example, n.: An input altered on purpose, often too slightly for people to notice, so that an AI model gets it badly wrong — like a panda photo that a classifier confidently calls a gibbon.
- browser agentbrowser agent, n.: An AI agent that uses a web browser on your behalf — searching, clicking links, filling in forms and reading pages — to complete tasks such as booking, shopping or gathering information.
- computer usecomputer use, n.: An AI's ability to work a computer through its normal screen and controls: it studies pictures of what is displayed and sends mouse and keyboard actions, so it can use ordinary apps that were built for people.
- web scrapingweb scraping, n.: Using software to automatically copy information from websites — text, prices, images — in bulk, which is how much of the data used to train AI was gathered.
- AI deceptionAI deception, n.: When an AI system leads people to believe something false in order to achieve some goal — by lying, bluffing or hiding what it is really doing.
Sources
- von Ahn, L., Blum, M., Hopper, N. J. & Langford, J. (2003). CAPTCHA: Using Hard AI Problems for Security. Proc. Eurocrypt 2003.
- von Ahn, L., Maurer, B., McMillen, C., Abraham, D. & Blum, M. (2008). reCAPTCHA: Human-Based Character Recognition via Web Security Measures. Science 321(5895), 1465–1468.
- Searles, A. et al. (2023). An Empirical Study & Evaluation of Modern CAPTCHAs. Proc. USENIX Security 2023.