Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

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

arstechnica.com

Post reply on HN
Print view View original

81–90 of 217 posts

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

#81

Post by throwaway202020 »

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 strong points in imaging (thanks to their 1000+ people team) and wrist rejection, virtually everything else they do that requires AI sucks and lags behind competition, including, Siri, maps, autocorrect, spell check, iCloud, search, calendar, spam detection, recommendations etc. For most of these things, most people don't even count them as real competition. Google on the other hand is able to achieve very competitive performance through software and AI without such large team on phone camera and frankly quite pathetic hardware.

These kind of reality distortion pieces aren't going to help them. They have $100+B, in cash, they can easily start reputable open research lab that can rival FAIR, OpenAI or DeepMind. Even a smaller companies like Intel and Adobe is starting to realize that this is necessary so they can tap into expertise on demand. At minimum that will be totally worth for a talent pipeline that can be motivated to do "rotation" or "sabbaticals" into product groups from an open lab.

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

#82

Post by zwegner »

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 recently discovered a neat trick for fixing the duck problem, that is, adding swear words to the iOS dictionary: add a contact with the chosen words in the name fields. E.g. I have a contact named "shit shitty fuck fuckin fucking oh shit"

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

#83

Post by htatche »

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.

I feel your pain.

On a day to day basis I do use 4 different languages, sometimes even mixing them in a single conversation (i.e the other person is bi/trilingual too). For me the autocorrection is only a distant dream and will probably remain so for a long time.

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

#84

Post by lacker »

"Google is an amazing company, and there's some really great technologists working there," he said. "But fundamentally, their business model is different and they're not known for shipping consumer experiences that are used by hundreds of millions of people."

What does he mean by this? Google search and Android are both used by more than a billion people. YouTube over 2 billion. These are all bigger than any Apple product. If anything, Google is known for shipping consumer experiences that reach a large number of people. Apple by contrast is known for its high-quality, high-price consumer experiences that reach fewer people.

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

#86

Post by anorakoverflow »

nfg wrote: post #46
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…

The one the drives me mad the most is when I type a word which it incorrectly autocorrects to a proper noun (which gets capitalised), I then use backspace to delete that word and type the - completely different - correct one, but it insists on upper-casing the first letter to match the original casing of the incorrect autocorrect. It really should keep track of the fact that it introduces the case-change and invalida…

I know exactly what you mean! This is even more annoying in German, where every noun is capitalised.

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

#87

Post by the_gipsy »

llampx wrote: post #48
moandcompany wrote: post #23

Earlier quoted context omitted.

When it comes to machine learning, the subtlety here is that there are at least two sides or facets to machine learning: (1) training and (2) inference. It's fair to say that there are multiple areas for AI leadership. It is generally believed that: Model Creation (1) Those with the access to the best (which is not necessarily the most, but often believed to be) data have a strong starting point for training models;…

Re: Prediction at the edge, I would think that it's better if there aren't going to be any updates to the model. Or if internet access is limited. Correct me if I'm wrong, but most of the ML inference actually takes place on the cloud nowadays, not on-device.

IIRC, for example phone keyboards use federated learning, where the model is further trained locally. You don't want to send every word the user types, for privacy reasons and others. Some kind of "diff" of the local model can then be sent to the cloud at some time to add up to the base model.

There is still an attack vector, you can infer a bit from the "diff", but you probably can't tell exactly what the user wrote.

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

#88

Post by bitxbit »

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…

That’s an interesting take. Does Apple really lag? I think quote-unquote AI in SV has pretty much stalled in the past 3-4 years. There’s been a lot of volume in ML but nothing groundbreaking.

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

#89

Post by pcr910303 »

I like this article. I like Apple’s approach to ML because it blends in. When applied, the feature should not expose that it’s based on ML — that’s a failure. And so Siri, Alexa, Google Assistant are failures. But Face ID, plam rejection are successes.

If you have to explain the customers that it’s ML based, that’s the same as asking for the customers to understand it’s unreliability. And unreliable features are worse than no features, and that’s why nobody uses Siri, Alexa, Google Assistant except for a few reliably-working requests.

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

#90

Post by throwaway202020 »

bitxbit wrote: post #88
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…

That’s an interesting take. Does Apple really lag? I think quote-unquote AI in SV has pretty much stalled in the past 3-4 years. There’s been a lot of volume in ML but nothing groundbreaking.

That's not true at all. AI is everywhere. When Gmail completes your sentence, auto-playlist gets created, camera detects a pet, you take night shot of a city, grammar suggestions are shown, wrong keystrokes auto-corrects, voice mail is transcribed, phone call is responded by quick text, your sleep is analysed, heart rate is monitored on runs, news articles gets recommended, spam email gets filtered, device idle detected... I can go on.

Most of these problems cannot be simply solved by just throwing in best engineers. No amount of classical algorithms you learned as CS major is going to help you implement the best solutions for these problems. The state of art solutions to these problems requires intense narrowly focused researchers who have studied these problems for many years, knows which 10% of the papers even worth looking at and pro/cons of different techniques. Something as benign as running neural net on phone hardware is intensely researched subject and your implementation can be literally 10X to 100X better at speed and power consumption if you have kept up with the field.

Post reply on HN