Why Apple believes it’s an AI leader–and why it says critics have it all wrong

Post by sgt » Sat, Aug 08, 2020, 3:46 PM UTC

tedeh wrote: post #61

jointpdf wrote: post #34

Setting aside all the other work Apple is doing in this area, can someone explain why iMessage autocorrect is unacceptably bad (to put it diplomatically) in the era of weapons-grade language models? Its behavior is totally bizarre. It’s like an underpowered chess engine that makes flagrant blunders: capitalizing random words in the middle of a sentence (rock -> Rock), the context sensitivity of an actual rock, forcin…

I'm bilingual and write approximately the same amount of messages in both languages using iOS. Autocorrect always suggests words from BOTH languages mid sentence. Like what the hell, they should be able to infer which language I'm using just from the first word.

Post by odyssey7 » Sat, Aug 08, 2020, 3:59 PM UTC

linguae wrote: post #115

throwaway202020 wrote: post #81

You can't be AI leader when every AI leader is staying away from you by 10ft pole. When Ian Goodfellow joined Apple, there was literally rain of criticism on him from ending his career as researcher to bowing down to money. I don't know of any researcher who does want to have continued research career willing to join Apple. They simply don't allow that kind of freedom or publishing results. While Apple has some stron…

It's a shame that Apple doesn't have an industrial research lab; it certainly has the funding to create a lab that would rival Microsoft Research or perhaps even legendary labs like Xerox PARC and Bell Labs. Apple used to have a lab: during the "interregnum" years of 1985 to 1997, Apple had a fantastic research group called the Advanced Technology Group, led by the late Larry Tesler. In some ways this group was a sor…

Post by dylz » Sat, Aug 08, 2020, 4:02 PM UTC

shrimpx wrote: post #108

throwaway202020 wrote: post #81

You can't be AI leader when every AI leader is staying away from you by 10ft pole. When Ian Goodfellow joined Apple, there was literally rain of criticism on him from ending his career as researcher to bowing down to money. I don't know of any researcher who does want to have continued research career willing to join Apple. They simply don't allow that kind of freedom or publishing results. While Apple has some stron…

Apple has fallen way behind in software/apps, and maybe they’re ok with it. The hardware and OS are good, and some staple apps for creators are good. But at everything else they seem really lethargic or absent. For example, as a web developer Apple has had zero impact on my tooling and knowledge. But their laptop is nice and I have one, which may be good enough for them.

Post by DerSaidin » Sat, Aug 08, 2020, 4:07 PM UTC

skybrian wrote: post #18

Despegar wrote: post #3

This piece is interesting because Apple was saying this all along but no one really believed them because it sounded like excuse making. But here JG is basically saying the same thing: >Yes, I understand this perception of bigger models in data centers somehow are more accurate, but it's actually wrong. It's actually technically wrong. It's better to run the model close to the data, rather than moving the data around…

It's not my area, but I read Google's AI blog and they seem to be doing machine learning research that ends up in its phones, such as various camera improvements and offline Google Translate. It seems like it's the sort of thing you do when you can, but often it comes as a second phase after getting it to work in the data center. I didn't see anything in this article that was obviously unique to Apple.

Post by parekhnish » Sat, Aug 08, 2020, 4:21 PM UTC

jointpdf wrote: post #34

Setting aside all the other work Apple is doing in this area, can someone explain why iMessage autocorrect is unacceptably bad (to put it diplomatically) in the era of weapons-grade language models? Its behavior is totally bizarre. It’s like an underpowered chess engine that makes flagrant blunders: capitalizing random words in the middle of a sentence (rock -> Rock), the context sensitivity of an actual rock, forcin…