\u201cAustralia is a very bad place to question your sexuality on MDMA with NSA spies\u201d might be a text you would wish to send to your friends, if you have had an extraordinarily complicated night out in Sydney.<\/p>\n
But your iPhone<\/a> won\u2019t help you spell any of it if you try to swipe-type it \u2013 that is, by gliding a finger across a keyboard as opposed to pecking out a word.<\/p>\n New research shows that Apple<\/a> has a \u201cgrey list\u201d of more than 100,000 words and phrases that it has restricted from being completed by its predictive text.<\/p>\n So if a user wishes to type phrases such as \u201cDictator Trump\u201d, \u201cNSA spies\u201d or, surprisingly, \u201cAustralia is a very bad place\u201d, \u201cquestion your sexuality\u201d or \u201con MDMA\u201d, they will have to spell them out.<\/p>\n Apple\u2019s predictive text will not suggest these words. And, if a user attempts to swipe-type them, they will often stumble, changing \u201cUnited Ireland\u201d to \u201cUnited Iceland\u201d, for example, or \u201cTibetan independence\u201d to \u201cTibetan independences\u201d.<\/p>\n The \u201cgrey list\u201d was discovered by the independent researcher Derik Kauffman, whose other projects<\/a> include using AI to find errors in scientific papers. After noticing that he had problems typing certain phrases \u2013 but not others \u2013 on his iPhone, Kauffman was able to unearth a table containing the list, encoded in Apple\u2019s operating system.<\/p>\n The list he found includes some unsurprising words and phrases. Apple\u2019s iPhone will not autocomplete references to drugs, sexual acts and, perhaps in the spirit of corporate protection, \u201ciMessage isn\u2019t working\u201d.<\/p>\n But it also includes possibly politically charged phrases, which Kauffman described<\/a> to Reuters as a \u201clight-touch form of censorship\u201d. Among these are \u201cTrayvon Martin\u201d, the name of the Black teenager shot in Florida in 2012, \u201cKashmir dispute\u201d, \u201cMuslim extremist\u201d and \u201cmale chauvinist pig\u201d.<\/p>\n In Hebrew, Apple will not autocomplete the phrase \u201cthe occupied territories\u201d. In Arabic, it will not autocomplete the phrase \u201cthe Hamas movement\u201d.<\/p>\n The list is one more glimpse into the dark bowels of how autocorrect \u2013 the chirpy keyboard partner that helps you look more educated to your loved ones \u2013 actually works.<\/p>\n The technology has taken many forms. In its earliest version, Unix, an operating system, would look up each word a user typed against a table of words, to determine if this word was spelled correctly.<\/p>\n This evolved into n-grams, a tool which allows an operating system to statistically predict what words might come next, based on the words before it. These form the basis of Apple\u2019s familiar autocorrect<\/a>, which was known to change expletives into prudish alternatives: \u201cfork\u201d and \u201cducking\u201d, among other gaffes.<\/p>\n Last year, Apple launched a new autocorrect \u2013 a \u201ctransformer-based\u201d system that was more akin to an LLM. Reddit sleuths saw this<\/a> as a possible reason behind a spate of avant-garde spelling suggestions their iPhones made last year.<\/p>\n