I suspect this is what’s happening. They’ve gone from a prescriptive ‘pre-programmed’ autocorrect to a more AI based ‘machine learning’ one. Hopefully this means it’ll eventually improve, although I don’t know why it’s taking so long.
I feel like they gathered data, studied it, and wrote a prescriptive autocorrect that IMO was perfectly fine and was still pretty good at catching words I was most likely to use.
Then, all of a sudden, it turned into fucking scrabble and I find myself going, “WTF are you thinking autocorrect?”
Not sure when the change started but it’s officially shitty now.
Barring code maintenance, they really had no reason to roll this out until it surpassed the old system. The AI could still train on data gathered while using the old system.
It could even be that as more and more of our lives gets funneled into these machines we’re seeing less literacy and therefore more typos to make the algorithms second guess themselves. If 10 users for every 100 type fir instead of for and don’t correct it the algo starts to see that as possible correct
1000%.
I’ve noticed across platforms, posts, texts, etc.
My guess is that there’s been a slow infiltration of “AI powered” autocorrect across the industry.
Other than that, I don’t really have a good answer to the broad, sweeping degradation of autocorrect.
But you’re definitely not the only one.
I suspect this is what’s happening. They’ve gone from a prescriptive ‘pre-programmed’ autocorrect to a more AI based ‘machine learning’ one. Hopefully this means it’ll eventually improve, although I don’t know why it’s taking so long.
(I could be wrong about all of this, of course.)
Yeah, this is precisely what I’ve been thinking.
I feel like they gathered data, studied it, and wrote a prescriptive autocorrect that IMO was perfectly fine and was still pretty good at catching words I was most likely to use.
Then, all of a sudden, it turned into fucking scrabble and I find myself going, “WTF are you thinking autocorrect?”
Not sure when the change started but it’s officially shitty now.
At least a couple years ago I’d say, and it affects voice to text recognition as well.
Barring code maintenance, they really had no reason to roll this out until it surpassed the old system. The AI could still train on data gathered while using the old system.
deleted by creator
It could even be that as more and more of our lives gets funneled into these machines we’re seeing less literacy and therefore more typos to make the algorithms second guess themselves. If 10 users for every 100 type fir instead of for and don’t correct it the algo starts to see that as possible correct