Interesting "Minority Report" result: Apple Intelligence reports the tennis winner before he wins the tournament. AI is proving to better at weather forecasting vs. current SOTA systems.
Not really, Apple Weather is so unreliable, it is a joke.
I ride a bike to work and my desk doesn't have a direct view out of the window - I have daylight, but an oblique angle out the window, where it is hard to see if it is raining. Usually it will be raining outside and Apple Weather says that there is 0% chance of rain and that it isn't currently raining... I've pretty much given up on it and generally get up and walk to the window, to see whether I need to put on my rain gear...
WÖW weather is showing light rain and high winds (60km/h) outside, Apple weather says 0% chance of rain and winds up to 21km/h... It is currently raining and the wind is trying to uproot nearby trees...
We won't get Apple Intelligence here until the spring, at the earliest. At the moment, I don't really have FOMO, given how badly it is failing. But that needs to be put into context, a lot of the other AI services I've used aren't any better.
I was watching a YouTube video with my wife at the weekend and it was auto-translating into German (she doesn't speak any English). It was a recipe and it was using dates (the fruit), the translation started going on about "pureeing the appointments"...
I keep trying different AI tools, but always come away disappointed, because the error rate is just too high to use them for anything serious at the moment. It is the same as the translation tools, I worked in a translation office for a while and they did everything by hand, because things like Google Translate were simply useless at doing a good job at translation, let alone a professional job.
Before I worked at the translation office, I had to translate an instruction manual to German in a hurry, I put it into Google Translate, hoping to save time...
"Do not open the case, high voltage inside" -> Gehäuse öffnen, Starkstrom drinnen (open the case, high voltage inside)
"Do not open the case, no user serviceable parts inside" -> Gehäuse öffnen, nichts drin (open the case, nothing inside)
GT now get this right, because I actually too the time to put the correction into it. But it still regularly makes simple mistakes. Knowing English and German, I can spot the mistakes easily, but if it does that badly with other languages, I wouldn't trust it. Just imagine translating a zoo sign with "do not enter the enclosure" drops the "not" in the translation...
Edit: I'm not against AI and I have been using it in certain situations for several years, such as phot editing software for removing artifacts or objects in the background. There it often does a great job. But the "general" intelligence models are, at the moment, generally providing poor information much of the time. They certainly aren't ready for prime time, yet they are being marketed as the Second Coming...