Don’t worry, you’re not going mad.
If you feel the autocorrect on your iPhone has gone haywire recently – inexplicably correcting words such as “come” to “coke” and “winter” to “w Inter” – then you are not the only one.
Judging by comments online, hundreds of internet sleuths feel the same way, with some fearing it will never be solved.
apple released its latest operating system, iOS 26, in September. About a month later, conspiracy theories abound, and a video purporting to show an iPhone keyboard changing a user’s spelling of the word “thumb” to “thjmb” has racked up more than 9m views.
“There’s a lot of different forms of autocorrect,” said Jan Pedersen, a statistician who did pioneering work on autocorrect for Microsoft.”It’s a little hard to know what technology people are actually employing to do their prediction, as it’s all underneath the surface.”
One of the godfathers of autocorrect has said those waiting for an answer might never know just how this new change works – especially considering who is behind it.Kenneth Church,a computational linguist who helped to pioneer some of the earliest approaches to autocorrect in the 1990s,said: “What Apple dose is always a deep,dark secret. And Apple is better at keeping secrets than most companies.”
The internet has been rumbling about autocorrect for the past few years, sence 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, in its release of iOS 17, an “on-device machine learning language model” that would learn from its users. The problem is, this could mean a lot of different things.
In response to a query from the Guardian, Apple said it had upda
Apple’s New Autocorrect: A Leap Beyond N-Grams into the World of AI
Table of Contents
Apple’s recent update to its autocorrect feature signals a significant shift from customary methods to more sophisticated artificial intelligence. While previous autocorrect systems relied on predicting text based on statistical probabilities – using a technique called n-grams – the new system utilizes a “transformer language model,” a technology powering advanced AI like ChatGPT and Google’s Gemini. This change promises a more bright and nuanced autocorrect experience, but also introduces new challenges in understanding why it sometimes fails.
From N-grams to Neural Networks: The evolution of Autocorrect
For years, autocorrect relied on n-grams. Essentially, thes systems predict the next word based on the preceding sequence of n* words. They analyze vast amounts of text to determine the most likely completion of a sentence.As explained by linguist Kenneth Church, n-grams make predictions based on statistical patterns in language. https://aclanthology.org/1993.ACL-main.101/
However, n-grams are now considered outdated. We’ve entered the era of AI, where more complex models are capable of understanding context and nuance in a way that n-grams simply cannot.
Apple’s “Transformer Language Model” – What Does it Mean?
Apple’s move to a “transformer language model” indicates a ample technological upgrade. Transformers are a key innovation in the field of AI, enabling models like ChatGPT and Gemini to process and generate human language with remarkable sophistication. According to computational linguist Graeme pedersen, this suggests a more complex system than previous autocorrect iterations. https://www.cs.cmu.edu/~gpederse/
Transformers excel at understanding relationships between words in a sentence, leading to more accurate and contextually relevant predictions.
Size and Limitations: AI on Your Phone
While the new autocorrect leverages powerful AI technology, it’s crucial to understand its limitations. Pedersen points out that the model running on a smartphone *must be significantly smaller than the massive AI models like ChatGPT or Gemini, which require substantial computing power.Running a full-scale AI model on a phone would be impractical due to processing and energy constraints.
The “Black Box” Problem: Understanding Autocorrect Errors
The increased complexity of the new autocorrect introduces a new challenge: interpretability. With older n-gram systems, it was relatively straightforward to understand why a particular correction was made. However, the inner workings of transformer language models are far more opaque.
Kenneth Church highlights this issue, describing the latest AI as “kind of like magic.” While it performs significantly better than older methods, diagnosing errors becomes much more difficult.this lack of “explainability” is a growing area of research in the AI field, aiming to understand how these complex models arrive at their conclusions.
This means that when Apple’s new autocorrect makes a mistake, it may be harder to pinpoint the cause and correct the underlying issue, leaving users frustrated with seemingly inexplicable errors.