Example:
I was halfway through typing the word “different”, and only got suggestions for “differences”, “differentiations” etc but “different” appeared in quotation marks as though I was inventing it.
In: Technology
ELI5: phones today use a variety of machine learning techniques to try and guess what you’re about to type. More often than not, these algorithms sacrifice accuracy for speed since phones have limited power.
Fuller explanation: autocorrect has been making progress for a while now. Your phone has language models built on massive datasets but also builds a dataset just from you. It tries to learn and anticipate what you’re going to say and it uses that predictive bias against whatever you’re actually typing. Once recently I was trying to type something like “Denmark” which is a very difference shape than “differences” but my phone was absolutely certain I was trying to type differences. Differences is a more common word by far than Denmark, so it figured that the chances of me meaning to say differences was way higher. This particular concept is called Bayes Theorem.
Also if you accidentally decline a suggestion, your phone will remember that for a short time and not suggest it again. Usually.
Latest Answers