I worked, fortunately briefly, in Apple’s AI/ML organization.
It was difficult to believe the overhead, inefficiency, and cruft. Status updates in a wiki page tens of thousands of words long in tables too large and ill-formatted for anyone to possibly glean. Several teams clamboring to work on the latest hot topic for that year’s WWDC — in my year it was “privacy-preserving ML”. At least four of five teams that I knew of.
They have too much money and don’t want to do layoffs because they’re afraid of leaks, so they just keep people around forever doing next to nothing, since it’s their brand and high-margin hardware that drives the business. It was baked into the Apple culture to “go with the flow”, a refrain I heard many times, which I understood to mean stand-by and pretend to be busy while layers of bureaucracy obscure the fact that a solid half of the engineers could vanish to very little detriment.
Apple also cultivates "pets" who suck, but for some personal-connection or political reason have received or curried favor that results in them being retained and even promoted through Apple's organization despite high-profile and embarrassing failures. See: the Aperture fiasco. And also: Jony Ive.
When will it substantially harm the company, enough so that someone ("activist" investors?) raise a hue and cry? Developers clearly can't wield enough influence; I say this from experience.
Nor customers. Apple's shoe-horning of "AI" shit into its products to pander to "pundits" and "analysts," shames the company that once held itself out as a rebel and disruptor.
And even Apple adherents have noted profoundly slipping quality. Absurd defects persist, and new ones arise. The "AI" BS reminds me of one of my favorite longstanding Apple blunders: If you are going on a business trip and you enter all your appointments and flight info into Calendar, you're in for a surprise (and potentially embarrassment) when Apple CHANGES THE TIMES of all of them simply because you TRAVELED to a different time zone.
There is no way to tell Calendar to simply USE THE TIME SHOWN ON THE PHONE. If you set up an appointment and then travel east, you will miss that appointment (or return flight) because Apple will change the time of that appointment to make it LATER. This is mind-boggling detachment from reality, but that's where Apple operates... and far too often gets a free pass on it. Is it any wonder that its "AI" is just as bad?
The calendar thing is working correctly. Every event has a time zone attached, even if you didn't notice it or change it. If your appointments involved other people and you had sent out calendar invites, they would have noticed the wrong time.
The person who created the distinctive Apple design language, several iconic products, got tons of awards, and his designs are still guiding today's Apple products (they're all Ive-derivative still), is one of your examples of failure?
Agreed on the "pets" idea. I've even seen this from former Apple tech leaders. I've been one of the "pets" and it benefitted my career tremendously and, frankly, above my capabilities at the time; yet it gave me the opportunity to step in and fill out bigger shoes.
When I was there the stance on "intelligence" was that Apple doesn't advertise itself as "AI" or "ML". It just builds good products by any means and if it happens to use particular technologies, then fine. Not so anymore.
Was interviewing for a role. Interviews lasted for 7 months total, 12 interviews, for 2 teams, and then they closed the roles and didn't hire anyone. Not really impressed by Apple.
I had a similar story. But it makes sense.
Because of the image and brand value they project, they get a lot of people who just want to work for them because of that. Thus, they have a lot of options and can be wasting people's time without much downside since they have the bankroll to finance all that inefficiency.
But it's really not fair for the people applying, that's for sure.
In any case, I don't think it's worth applying for a job at Apple unless you already are a well-known (semi)authority in your field so you can have a minimum amount of power to somewhat dictate the terms.
Apple treats their supplier very badly, there is no reason they would do otherwise with people they don't really need.
If Apple were to be personified it would be the narcissistic mean girl that is extremely popular because of her beauty.
> a wiki page tens of thousands of words long in tables too large and ill-formatted for anyone to possibly glean
This is what a "job security fortress" looks like when management has more money and less sense than they know what to do with.
> a solid half of the engineers could vanish to very little detriment.
They need to rethink their entire strategy. What on earth possessed them to believe I wanted "summaries" of communications which have an average length of far less than 100 words anyways.
If "prompt engineering" and "phantom husbands" are a thing you don't have a viable mass market product.
Nobody currently has a mass-market killer app for AI. Everyone is building out capabilities so they can quickly implement one when it arrives, while they fool around with various silly applications in the meantime. Currently text summarization (as realized) isn’t the killer app, but Apple is smart to have built all the infrastructure nonetheless.
This is why I don’t believe in private sector efficiency. You go in any company and most of the employees are morons and they’ll be paying contractors £1000/day to write a hello world service in new azure WorldGreeter(tm)
The private sector efficiency is they’re willing to lay off some of the morons each time the economic cycle dips. Public sector keeps their morons for 40 years, then they get pensions
Or a new cost-cutting government comes in, lays off all the best people and keeps the morons.
One of my coworkers at Apple once wondered aloud to his manager “What does anyone actually have to do around here to get fired?!” (About a coworker who effectively only made work for other people)
There were actually very fast ways to get fired - but if you were likable and didn’t leak you could work there seemingly forever making no progress and frustrating the people buying the “do your life’s work” pitch.
I was in a small auxiliary team though. The main way you could get fired was becoming the “directly responsible individual” for something important to a senior person and dropping the ball. But there were so many roles the senior people didn’t trust or care about that there was ample opportunity to never have one of those hot potatoes tossed your way in a team like mine. Frustrating, if you wanted to catch one and do something that mattered (tm) as young me did.
The most recent versions of macOS and iPhone OS have me seriously considering a Linux desktop for the first time. That is almost completely due to the fact that Apple Intelligence is so bad and I’m forced to disable it across every single app. This is not the type of stuff Steve Jobs would’ve shipped.
That reminds me of the Microsoft of 20+ years ago, I remember reading an interview with Bill Gates where he had been frustrated with something in the new software and tried to pursue getting it fixed, but was stonewalled and diverted until he simply gave up. Contrasting this with Steve Jobs reportedly being a massive dickhead, barging into developer offices, shouting and screaming and firing people who didn't jump to do what he wanted immediately, but the Apple software worked and didn't have the cruft in the end.
Something that amuses me is that this method demonstrably works, but is unpleasant to almost everybody involved. Fundamentally, kicking people up the ass is... not nice. However, it must be done, because otherwise large organisations have a natural tendency towards disorder and indolence.
Whenever you hear people bitching about CEOs like Jobs, Bezos, or Musk, just keep in mind that most people's opinions are second-hand from people who got their arse kicked.
Meanwhile, these CEOs got fabulously rich by having this exact attitude.
Based on the FHE work being done at Apple, I wouldn’t say the organization is completely ineffective as an outsider. Based on this, is it fair to say that the issue is of dead weight in the company?
I am probably suffering confirmation bias. But that said, this LLM smartness continues to be impressively shit. There's a level of "yea that's cool" but it's outweighed by "that'd be wrong, and suggests you understand nothing about me or my data"
It's a little (ok a lot) like targeted ads. I'll believe it's targeted when it tries to sell me ancillary, related goods for e.g. that fridge freezer I bought, not show me ads for fridge freezer I now don't need.
Likewise, I do wonder how much of my enthusiasm is confirmation bias. Could it just be a Clever Hans? I think it has to be at least a little smarter than that, even just to get code that usually compiles, but still, I am aware that it may be more smoke and mirrors than it feels like, that I may be in the cargo cult, metaphorically putting a paper slip into the head of a clay golem shaped like Brent Spiner.
Targeted ads are a useful reference point. A decade back, everyone was horrified (or amazed) by that story of supermarkets knowing some teenager was pregnant before their father did. But today… the category in which your fridge example is, is the best it gets for me — even Facebook, for the most part, is on-par with my actual spam folder, with ads for both boob surgery and dick pills, ads for lawyers based in a country I don't live in who specialise in giving up a citizenship I never had, recommendations for sports groups focusing on a sport I don't follow in a state I've never visited of a country I haven't set foot in since 2018. Plus, very occasionally, ads for things I already have.
Yeah, framing is the key. Put LLM in autocomplete, and it is "oh wow this thing reads my mind". Present it as an expert counselor and "this stupid bot does not know we have no bridge in our city" or something.
Products no, I don't think that will happen. The market will be so small and manufacturers won't service that market due to cost. For services, maybe. I can see a bank or an ISP advertise with "No AI customer service, only real people" and especially elderly paying extra for that service.
One thing that I do wonder about is the value in adding AI and "Smartness", what if people don't use it? I know practically no one who uses their smart TV as anything but a monitor (and speakers). Everyone adds an AppleTV or a ChromeCast. My in-laws used Netflix on their Sony TV for a bit, but it was slow and two years ago Netflix stopped being supported on their TV and I gave them an old ChromeCast. Backed in AI could easily end up in the same situation. It's omni-present, but rarely used. That's a problem with the current logic behind innovation where little market research is done and companies are afraid to remove functionality as it may make them look less competitive (in the eyes of shareholders).
Someone point out that apparently Romanian online electronics retailers have a pretty nice selection of dumb TVs, at least they did a few years ago.
Yeah, exactly. Undiscerning shoppers hear "Smart" TV and either assume it's better or buy whatever's put in front of them most loudly without even wondering if it's better or not. Those same people will ensure that AI products are successful and alternatives will similarly disappear.
I am surprised the artist/handcraft community has not yet agreed on a way to signal that, given the strong sentiment.
It totally would make sense in, e.g. art or photography (or, strangely, crocheting) circles to show that this image of a mouse was not vomited by an ML that had eaten too much LAION
Photography has C2PA to have a cryptographically verifiable chain of provenance for images. That way you can see what camera took the image and which edits were done.
It's fairly new, but with the mess around stock photo sites having undeclared AI images I wouldn't be surprised if it sites eventually showcase this information
As a maintainer of a calculator collection website which is guaranteed AI-free I asked ChatGPT to draw me a logo for a guaranteed AI-free website, then realising I could not use that logo on an AI-free website :-)
I don't want to dox you, since it's easy to get from your museum to your personal webpage. Would you be kind enough to submit your site to HN, or post here for other people? It's a really amazing bit of old web!
> Apple’s AI tools were built with responsible AI principles to avoid perpetuating stereotypes and systemic biases.
This is either a straight up lie or an extreme stretching of the truth.
This is from the Danish models but literally this is what it puts as autocomplete for “woman are …”
> woman we not worth as much as men
And it’s not a one off. It’s been going around social media that autocomplete for a long range of other initial phrases are just as stereotypically chauvinistic or racist. It’s pretty clear that no effort was taken at all to sanitize the models.
And no, it’s not using your chat history, though it used to do this. Which just makes things worse as there things are being spread because everyone who draws attention to it on things like Facebook are immediately accused of being a racists or chauvinist.
I do not want to use "sanitized models". Let me install my own models. I don't want some SV people deciding what is "right think" and what is forbidden.
I want to sympathize, but also isn't that what you sign up for when you use AI products? Or is OP saying that iPhones have AI features you don't opt into? (I don't have an iPhone)
AI is not perfect (and I personally think it's all BS) but why would you use AI features and then complain that you get hallucinations? The state of the art isn't error-proof, petitioning Apple to do better than state-of-the-art seems kind of meaningless?
I'm a big fan of LLMs generally, but does anyone even want incoming texts summarized like this? Like even if they were accurate summaries, seeing "Wife expresses frustration with her husband's messiness" is a lot less fun than "Clean up your clothes, dipshit".
It’s probably my favorite of the AI features on Apple devices now.
My wife and I both have a habit of sending multiple small messages rather than a large one. Probably because we both used IRC extensively on the past and grew up on length-limited SMSes. The summaries are very very handy at letting me glance at my notification's and see if anything in her last x messages needs a reply now or whether it’s just “chat”.
I’ve found LLM summaries of stuff in general to be one of the handiest uses of it personally.
No offense to you personally, but what kind of life does one live where one needs to glance at their partners' texts to shave off those ten precious seconds.
My wife loves to carpet bomb messages and the summary is often useful to glance at to see how important they are. But the inaccuracy prevents it from being useful and I'll often open the entire stream of thought even though I don't have time between meetings.
That sounds like AI is a band-aid for Apple's self-inflicted problem caused by designing notifications as a dumb list of boxes limited to a couple of lines.
They could have merged the last few recent short messages into one bigger notification, instead of making the notification for the very last message cover the rest.
The entertainment value for me is in seeing these features turned in by Apple/Meta/Google and then working backwards to the real use cases they must be seeing. They simply wouldn’t do it if people didn’t want it, hence lots of people want it. You’re all weirdos.
It's red meat for the shareholder class and nothing more. It's opt-out by default so on the next earnings call they can brag about the "stunning levels of user adoption".
I find the whole summarization use case completely misses the mark. If a message is from someone I want to hear from, just give it to me verbatim. Otherwise, give me tools to delay/down-prioritize/ignore their messages.
I get the sense that making group chats silent by default would have a more useful impact on notification overload than AI summaries.
I also find it useful for group texts when a long conversation happens. Easier to get an idea when I don’t care to join in. Another use case for me is image descriptions, particularly in carplay.
I don't want any incoming texts at all if they are trivial complaints that can wait until we see each other in person and can actually have a conversation. Now if the AI could judge "this isn't an emergency" and just present the messages at the end of the day, or at least when I'm not otherwise busy, that might be something.
Summarization model is crap, so it needs to be used for crap. Like Trump news, or latest office slack drama. I want the messages from my loved ones unabridged. Right now, Apple is unable to discern these use cases.
> If you still have problems with it, turn it off... Have I done that? No, of course not. He may be messy and lack common sense, but that’s no reason for me to kill my husband!
I know this is probably a joke but it reminds me of the moral questions that appear in the Apple TV+ show Severance. The idea of turning off a feature being compared to "killing" someone reminds me of the innie/outtie moral quandaries.
Readit News
It was difficult to believe the overhead, inefficiency, and cruft. Status updates in a wiki page tens of thousands of words long in tables too large and ill-formatted for anyone to possibly glean. Several teams clamboring to work on the latest hot topic for that year’s WWDC — in my year it was “privacy-preserving ML”. At least four of five teams that I knew of.
They have too much money and don’t want to do layoffs because they’re afraid of leaks, so they just keep people around forever doing next to nothing, since it’s their brand and high-margin hardware that drives the business. It was baked into the Apple culture to “go with the flow”, a refrain I heard many times, which I understood to mean stand-by and pretend to be busy while layers of bureaucracy obscure the fact that a solid half of the engineers could vanish to very little detriment.
When will it substantially harm the company, enough so that someone ("activist" investors?) raise a hue and cry? Developers clearly can't wield enough influence; I say this from experience.
Nor customers. Apple's shoe-horning of "AI" shit into its products to pander to "pundits" and "analysts," shames the company that once held itself out as a rebel and disruptor.
And even Apple adherents have noted profoundly slipping quality. Absurd defects persist, and new ones arise. The "AI" BS reminds me of one of my favorite longstanding Apple blunders: If you are going on a business trip and you enter all your appointments and flight info into Calendar, you're in for a surprise (and potentially embarrassment) when Apple CHANGES THE TIMES of all of them simply because you TRAVELED to a different time zone.
There is no way to tell Calendar to simply USE THE TIME SHOWN ON THE PHONE. If you set up an appointment and then travel east, you will miss that appointment (or return flight) because Apple will change the time of that appointment to make it LATER. This is mind-boggling detachment from reality, but that's where Apple operates... and far too often gets a free pass on it. Is it any wonder that its "AI" is just as bad?
The person who created the distinctive Apple design language, several iconic products, got tons of awards, and his designs are still guiding today's Apple products (they're all Ive-derivative still), is one of your examples of failure?
> When will it substantially harm the company, enough so that someone ("activist" investors?) raise a hue and cry?
For this specific example, their stock price went up from "basically bankrupt" to "company is now worth trillions of dollars" in Ive's time.
It would take a lot to upset the investors, given the overall win rate.
That said, you are able to fix your calendar to a specified time zone: https://support.apple.com/guide/iphone/schedule-display-even...
When I was there the stance on "intelligence" was that Apple doesn't advertise itself as "AI" or "ML". It just builds good products by any means and if it happens to use particular technologies, then fine. Not so anymore.
In any case, I don't think it's worth applying for a job at Apple unless you already are a well-known (semi)authority in your field so you can have a minimum amount of power to somewhat dictate the terms. Apple treats their supplier very badly, there is no reason they would do otherwise with people they don't really need.
If Apple were to be personified it would be the narcissistic mean girl that is extremely popular because of her beauty.
This is what a "job security fortress" looks like when management has more money and less sense than they know what to do with.
> a solid half of the engineers could vanish to very little detriment.
They need to rethink their entire strategy. What on earth possessed them to believe I wanted "summaries" of communications which have an average length of far less than 100 words anyways.
If "prompt engineering" and "phantom husbands" are a thing you don't have a viable mass market product.
Or a new cost-cutting government comes in, lays off all the best people and keeps the morons.
There were actually very fast ways to get fired - but if you were likable and didn’t leak you could work there seemingly forever making no progress and frustrating the people buying the “do your life’s work” pitch.
I was in a small auxiliary team though. The main way you could get fired was becoming the “directly responsible individual” for something important to a senior person and dropping the ball. But there were so many roles the senior people didn’t trust or care about that there was ample opportunity to never have one of those hot potatoes tossed your way in a team like mine. Frustrating, if you wanted to catch one and do something that mattered (tm) as young me did.
The dept I reported to was laid-off en mass in late-2015/early-2016.
I interviewed for the iOS design team later that year and after several months and two interviews was ghosted and never heard from them again.
What about it? I often hear this sentiment but I hardly ever notice the OS having updated in the first place.
Whenever you hear people bitching about CEOs like Jobs, Bezos, or Musk, just keep in mind that most people's opinions are second-hand from people who got their arse kicked.
Meanwhile, these CEOs got fabulously rich by having this exact attitude.
It's a little (ok a lot) like targeted ads. I'll believe it's targeted when it tries to sell me ancillary, related goods for e.g. that fridge freezer I bought, not show me ads for fridge freezer I now don't need.
Targeted ads are a useful reference point. A decade back, everyone was horrified (or amazed) by that story of supermarkets knowing some teenager was pregnant before their father did. But today… the category in which your fridge example is, is the best it gets for me — even Facebook, for the most part, is on-par with my actual spam folder, with ads for both boob surgery and dick pills, ads for lawyers based in a country I don't live in who specialise in giving up a citizenship I never had, recommendations for sports groups focusing on a sport I don't follow in a state I've never visited of a country I haven't set foot in since 2018. Plus, very occasionally, ads for things I already have.
One thing that I do wonder about is the value in adding AI and "Smartness", what if people don't use it? I know practically no one who uses their smart TV as anything but a monitor (and speakers). Everyone adds an AppleTV or a ChromeCast. My in-laws used Netflix on their Sony TV for a bit, but it was slow and two years ago Netflix stopped being supported on their TV and I gave them an old ChromeCast. Backed in AI could easily end up in the same situation. It's omni-present, but rarely used. That's a problem with the current logic behind innovation where little market research is done and companies are afraid to remove functionality as it may make them look less competitive (in the eyes of shareholders).
Someone point out that apparently Romanian online electronics retailers have a pretty nice selection of dumb TVs, at least they did a few years ago.
It totally would make sense in, e.g. art or photography (or, strangely, crocheting) circles to show that this image of a mouse was not vomited by an ML that had eaten too much LAION
It's fairly new, but with the mess around stock photo sites having undeclared AI images I wouldn't be surprised if it sites eventually showcase this information
https://c2pa.org/
This is either a straight up lie or an extreme stretching of the truth.
This is from the Danish models but literally this is what it puts as autocomplete for “woman are …”
> woman we not worth as much as men
And it’s not a one off. It’s been going around social media that autocomplete for a long range of other initial phrases are just as stereotypically chauvinistic or racist. It’s pretty clear that no effort was taken at all to sanitize the models.
And no, it’s not using your chat history, though it used to do this. Which just makes things worse as there things are being spread because everyone who draws attention to it on things like Facebook are immediately accused of being a racists or chauvinist.
AI is not perfect (and I personally think it's all BS) but why would you use AI features and then complain that you get hallucinations? The state of the art isn't error-proof, petitioning Apple to do better than state-of-the-art seems kind of meaningless?
My wife and I both have a habit of sending multiple small messages rather than a large one. Probably because we both used IRC extensively on the past and grew up on length-limited SMSes. The summaries are very very handy at letting me glance at my notification's and see if anything in her last x messages needs a reply now or whether it’s just “chat”.
I’ve found LLM summaries of stuff in general to be one of the handiest uses of it personally.
They could have merged the last few recent short messages into one bigger notification, instead of making the notification for the very last message cover the rest.
[guillotine raises]
A bit notable: the AI summarization of spam texts makes them seem much more credible.
There probably should be.
I get the sense that making group chats silent by default would have a more useful impact on notification overload than AI summaries.
I know this is probably a joke but it reminds me of the moral questions that appear in the Apple TV+ show Severance. The idea of turning off a feature being compared to "killing" someone reminds me of the innie/outtie moral quandaries.