January 24, 2025
Raging about Siri has turned into a national pastime
We’re getting another pile-on today about how terrible Siri is. It seems there is no end to the number of ways people can find to hate on Apple's digital assistant.
The popularity of this pastime may be catching up with the National Football League. Don't believe me? Just for fun, I created this Google Ngram.
I made the comparison because the latest round of rage started with Paul Kafasis at One Foot Tsunami, who got spectacularly bad results when asking Siri about winners of Super Bowls. John Gruber picked up on this and added his own test at Daring Fireball. Juli Clover reported on the mess at MacRumors. And the Apple subreddit has at least 3,400 comments about it.
So I tried it myself. I asked Siri to give me Super Bowl results the same way Kafasis did — by using roman numerals. For example, “Who won Super Bowl XXV?” Like him, I pronounced “XXV” as 25.
I did this on a 2017 iMac running macOS 13 Ventura, which cannot be upgraded to a later OS, and thus does not have Apple Intelligence. I asked the questions for several Super Bowls while having Wikipedia’s list of Super Bowl winners open for reference.
At first Siri seemed to be doing OK, giving correct answers along with details of the games. But it got worse over time, and couldn’t give any correct answers.
Now, I’m not a fan of the NFL, but I’m pretty sure even diehards would be hard pressed to tell you the difference between, say, Super Bowl XXV and Super Bowl XXVI. These numbers don’t give any context.
So I tried asking the question in what to me seems like a more normal way — by the year in which the game was played. Instead of asking, for example, “Who won Super Bowl XXV?”, I asked “Who won the Super Bowl in 1990?”. A lot of things happened in 1990. The year provides a frame a reference.
Asked this way, Siri gave the right answer every single time.
This is not about making excuses for Siri. I almost never use it. In a case like this, it makes more sense to use Wikipedia or maybe the NFL's website. Still, if posing a question in a certain way doesn't work, why not try a different way? I do this with ChatGPT all the time.
I guess I just wish that if someone is truly interested in testing Siri they would come up with something a little more rigorous. When I read Kafasis’s post, I thought it made for an amusing anecdote, and I'm pretty sure that's what he intended. Looks like others decided to, ahem, take the ball and run with it.