Don\u2019t worry, you\u2019re not going mad.<\/p>\n
If you feel the autocorrect on your iPhone has gone haywire recently \u2013 inexplicably correcting words such as \u201ccome\u201d to \u201ccoke\u201d<\/a> and \u201cwinter\u201d to \u201cw Inter\u201d<\/a> \u2013 then you are not the only one.<\/p>\n Judging by comments<\/a> online, hundreds<\/a> of internet sleuths feel the same way, with some fearing it will never be solved.<\/p>\n Apple released its latest operating system, iOS 26, in September. About a month later, conspiracy theories abound, and a video<\/a> purporting to show an iPhone keyboard changing a user\u2019s spelling of the word \u201cthumb\u201d to \u201cthjmb\u201d has racked up more than 9m views.<\/p>\n \u201cThere\u2019s a lot of different forms of autocorrect,\u201d said Jan Pedersen, a statistician who did pioneering work on autocorrect for Microsoft. \u201cIt\u2019s a little hard to know what technology people are actually employing to do their prediction, because it\u2019s all underneath the surface.\u201d<\/p>\n One of the godfathers of autocorrect has said those waiting for an answer might never know just how this new change works \u2013 especially considering who is behind it.<\/p>\n Kenneth Church, a computational linguist who helped to pioneer some of the earliest approaches to autocorrect in the 1990s, said: \u201cWhat Apple<\/a> does is always a deep, dark secret. And Apple is better at keeping secrets than most companies.\u201d<\/p>\n The internet has been rumbling about autocorrect<\/a> for the past few years, since even before iOS 26. But there is at least one concrete difference between what autocorrect is now and what it was several years ago: artificial intelligence, or what Apple termed<\/a>, in its release of iOS 17, an \u201con-device machine learning language model\u201d that would learn from its users. The problem is, this could mean a lot of different things.<\/p>\n In response to a query from the Guardian, Apple said it had updated autocorrect over the years with the latest technologies, and that autocorrect was now an on-device language model. They said that the keyboard issue in the video was not related to autocorrect.<\/p>\n Autocorrect is a development on an earlier technology: spellchecking. Spellchecking started in roughly the 1970s, and included an early command in Unix \u2013 a coding language \u2013 that would list all the misspelled words in a given file of text. This was straightforward: compare each word in a document with a dictionary, and tell a user if one does not appear.<\/p>\n \u201cOne of the first things I did at Bell Labs was acquire the rights to British dictionaries,\u201d said Church, who used these for his early work in autocorrect and for speech-synthesis programs.<\/p>\n Autocorrecting a word \u2013 that is, suggesting in real time that a user might have meant \u201ctheir\u201d as opposed to \u201cthier\u201d \u2013 is far harder. It involves maths: the computer has to decide, statistically, if by \u201cgraff\u201d you were more likely referring to a giraffe \u2013 only two letters off \u2013 or a homophone, such as \u201cgraph\u201d.<\/p>\n In advanced cases, autocorrect also has to decide if a real English word you\u2019ve used is actually appropriate for context, or if you probably meant that your teenage son was good at \u201cmath\u201d and not \u201cmeth\u201d.<\/p>\n Up until a few years ago, the state-of-the-art technologywas n-grams, a system that worked so well most people took it for granted \u2013 except when it seemed unable to recognise less-common names<\/a>, prudishly replaced expletives with unsatisfying alternatives (something which can be ducking annoying) or apocryphally changed sentences<\/a> such as \u201cdelivered a baby in a cab\u201d to \u201cdevoured a baby in a cab.\u201d<\/p>\n skip past newsletter promotion<\/a><\/p>\n A weekly dive in to how technology is shaping our lives<\/p>\n Privacy Notice: Newsletters may contain information about charities, online ads, and content funded by outside parties. If you do not have an account, we will create a guest account for you on theguardian.com<\/a> to send you this newsletter. You can complete full registration at any time. For more information about how we use your data see our Privacy Policy<\/a>. We use Google reCaptcha to protect our website and the Google Privacy Policy<\/a> and Terms of Service<\/a> apply.<\/p>\n after newsletter promotion<\/p>\n Put simply, n-grams are a very basic version of modern LLMs such as ChatGPT. They make statistical predictions on what you\u2019re likely to say based on what you\u2019ve said before and how most people complete the sentence you\u2019ve begun. Different engineering strategies affect what data an n-gram autocorrect takes in, says Church.<\/p>\n But they are state-of-the-art no longer; we\u2019re in the AI era.<\/p>\n Apple\u2019s new offering, a \u201ctransformer language model\u201d, implies a technology that is more complex than old autocorrect, says Pedersen. A transformer is one of the key advances that underpins models such as ChatGPT and Gemini \u2013 it makes these models more sophisticated in responding to human queries.<\/p>\n What this means for new autocorrect is less clear. Pedersen says that whatever Apple has implemented, it is likely to be far smaller than familiar AI models \u2013 otherwise it could not run on a phone.<\/p>\n But crucially, it is likely to be far harder to understand what is going wrong in new autocorrect than in previous models, because of the challenges of interpreting AI.<\/p>\n \u201cThere\u2019s this whole area of explainability, interpretability, where people want to understand how stuff works,\u201d said Church. \u201cWith the older methods, you can actually get an answer to what\u2019s going on. The latest, greatest stuff is kind of like magic. It works a lot better than the older stuff. But when it goes, it\u2019s really bad.\u201d<\/p>\n","protected":false},"excerpt":{"rendered":"Don\u2019t worry, you\u2019re not going mad. If you feel the autocorrect on your iPhone has gone haywire recently…\n","protected":false},"author":2,"featured_media":264613,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[41],"tags":[165,74],"class_list":{"0":"post-264612","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-mobile","8":"tag-mobile","9":"tag-technology"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/264612","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/comments?post=264612"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/posts\/264612\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media\/264613"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/media?parent=264612"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/categories?post=264612"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us\/wp-json\/wp\/v2\/tags?post=264612"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}