Report describes Apple’s “organizational dysfunction” and “lack of ambition” in AI
- Thread starter JournalBot
- Start date
You are incorrect. The name is Advanced RISC Machine. It derived from the Acorn RISC Machine with a shit-ton of changes and new instructions introduced by Apple to meet the specific needs of the Newton requiring a pretty capable mobile processor. VLSI was also a partner to provide expertise on the layout and manufacture.
Apple continued to provide direct contributions to silicon through the PowerPC era as well with the vector unit designed almost entirely by Apple. Apple invented Firewire as an isochronous interface was needed that could guarantee bandwidth for audio and video applications, Apple also co-developed Thunderbolt with Intel, and on and on.
Some of us were here when the Ars community was saying that Apple couldn't possibly have helped develop the A4 processor that was all Samsung, and that the A5 processor couldn't have possibly had Apple designed cores, that must have been either ARM or Samsung. Even the competition bought into this when they were absolutely fucking blindsided when the A7 shipped as the first 64 bit ARM processor - being called a gimmick. And ARM could never scale to desktop, Apple will never drop Intel. And then the M1 ships and fucking runs circles around anything in the market at the same power envelope, and so on.
Yes, we know - Apple is incompetent. Apple has always been incompetent. Apple will always be incompetent. All they do is marketing. Anyone who says otherwise is lying to you.
Upvote
32
(33
/
-1)
They are market leaders or top competitors in:
- Phones
- Tablets
- Laptops
- Headphones
- Watches
- Set top boxes
- Home automation
- Car entertainment systems
- Mobile payments
- Music
- Movies and television content
- Health data
- Desktop and mobile processors independent of the above (Apple Silicon)
- Compilers independent of the above (LLVM, Clang, etc.)
- Programming languages independent of the above (Swift)
- I/O interfaces independent of the above (Thunderbolt)
- Video/image codecs independent of the above (ProRes, HEIF)
What was the last product/service that Apple canceled? Can you give even a single example of what you are describing?
Upvote
26
(26
/
0)
Apple never said NFT or blockchain either and my portfolio is just fine.
Upvote
16
(16
/
0)
This is another dumb article that lumps Apple in with the rest of "big tech." So-called AI is all OpenAI does, so it is nowhere near a competitor to Apple. Complaining that Apple is "falling behind" them is like complaining that Burger King is falling behind Baker's Square in making pies.
Google and Microsoft sell software and SaaS, which is not what Apple does either. This article fails to state exactly what Apple is supposed to be doing with AI that it's not.
Junk.
Google and Microsoft sell software and SaaS, which is not what Apple does either. This article fails to state exactly what Apple is supposed to be doing with AI that it's not.
Junk.
Upvote
8
(8
/
0)
After reading most of this story and reading the earliest comments I find myself in agreement. Apple's last major mistake was the Newton, not because it didn't work, but because you technology to make it work cost too much and it was too bulky. Now 100s of millions of iPhones later, we have a Super Newton at a successful price point. I get the feeling from this story that the AI folks at Apple are in a lot more of a hurry than Tim Cook. Sure Apple has the cash to put this "Thing" into the marketplace, but will it be another Lisa or Newton? The verdict up top seems to think so. Apple's best and brightest AI engineers will have to wait a little longer, like ten years longer as in the time between the demise of the Newton and the advent of the iPhone.
Upvote
6
(6
/
0)
Also failure modes matter: if Photos misses 3% of my pictures of a friend, it’s perhaps a little annoying but nothing like what could happen if an LLM emits a very confident falsehood or the police use some automated phrenology system to decide who to arrest, etc. Facebook and Google tend to be both far more cavalier about that kind of thing and unwilling to provide effective (or any) feedback mechanisms to correct errors.
Upvote
16
(16
/
0)
This actually very reassuring to see apple is sticking to a more measured and especially local first approach
Upvote
8
(8
/
0)
Yeah right. Thus the total lack of a patent portfolio.
You're right, Apple is not like the idiots who race to type "First!" at the start of a comment section. You point isn't much more meaningful than that.
Upvote
15
(15
/
0)
You think the Altair is a personal computer the likes of the Apple I or Apple II?
It wasn't. Oh it was cool flipping switches and watching the lights flash.
Upvote
7
(8
/
-1)
An article with a headline alleging “organisational dysfunction” within Apple should really have reported on something remotely approaching that. The entire body of the article could be summarised as “Apple conservative on LLMs; some engineers unhappy.”
Upvote
19
(19
/
0)
Upvote
6
(6
/
0)
Even if Apple were behind (and they probably are), they'll just buy a company to catch up. Hell, I could see them dropping a huge check for OpenAI if they really felt behind. They are among the richest of the rich. Apple dropping 30-40 billion on OpenAI tomorrow would still be near half of what Microsoft is spending just for Call of Duty.
Upvote
3
(4
/
-1)
If Apple's AIs lack ambition, perhaps they just need more motivation?
Or maybe they're just slacker AIs.
Upvote
-9
(0
/
-9)
Upvote
1
(3
/
-2)
I get the impression that some internal tension of people wanting to move fast and break stuff is rubbing up against Apple's general strategy of waiting until a technology is mature and clearly useful before entering the market.
Upvote
14
(14
/
0)
Upvote
9
(9
/
0)
Hey Siri, turn the light on in the st...
..actually don't bother, I'll do it myself... using this switch thingy (which is 10x faster and less infuriating)
There was a time when Siri used to turn it on and say "voila!", these days Siri doesn't talk to me, or play the music I request, or even grunt to acknowledge when I ask a question....
..actually don't bother, I'll do it myself... using this switch thingy (which is 10x faster and less infuriating)
There was a time when Siri used to turn it on and say "voila!", these days Siri doesn't talk to me, or play the music I request, or even grunt to acknowledge when I ask a question....
Upvote
-10
(0
/
-10)
Upvote
-7
(1
/
-8)
I think this is fine. I'm a huge skeptic that there are going to be many pro-social applications for LLMs. For pro-social applications, accuracy is important. MS might build some useful auto complete functions into Word and Outlook. (Like "autocomplete this cover letter, 2-3 sentences in"). There's a bit of potential.
There's massive potential for anti-social uses of the technology, like spamming, phishing, misinformation campaigns, ... because in those applications, accuracy is generally not important.
This case is pretty much the reverse of that quote for Jurassic Park that always gets pulled out, Apple is not asking whether they CAN do things with AI but whether they SHOULD. The world might be a better place if that was the rule rather than the exception.
There's massive potential for anti-social uses of the technology, like spamming, phishing, misinformation campaigns, ... because in those applications, accuracy is generally not important.
This case is pretty much the reverse of that quote for Jurassic Park that always gets pulled out, Apple is not asking whether they CAN do things with AI but whether they SHOULD. The world might be a better place if that was the rule rather than the exception.
Upvote
5
(5
/
0)
Google likely faces the bigger challenge. From business insider:
"Pichai has said that Google continues to hire top researchers in the AI field, and there's no doubt the company still retains much strong talent. Still, many of the major contributors to Google's most seminal AI papers have left to join competitors like OpenAI or start their own firms. Daniel De Freitas and Noam Shazeer, two researchers behind Google's large language model LaMDA, left the company, frustrated by its hesitation to release a chatbot like ChatGPT, The Wall Street Journal reported.
That frustration over Google's slow movement has been corroborated by other former Google researchers who spoke to Insider. One former employee said that at such an exciting time for AI, startups offer the opportunity for researchers to have more ownership over their work and possibly make a greater impact. Below are some of the most notable papers in the field of AI, the papers' researchers who have departed Google, and where they are now."
And then they list a huge number of heavy hitters who've left Google.
https://www.businessinsider.com/goo...as and Noam,The Wall Street Journal reported.
"Pichai has said that Google continues to hire top researchers in the AI field, and there's no doubt the company still retains much strong talent. Still, many of the major contributors to Google's most seminal AI papers have left to join competitors like OpenAI or start their own firms. Daniel De Freitas and Noam Shazeer, two researchers behind Google's large language model LaMDA, left the company, frustrated by its hesitation to release a chatbot like ChatGPT, The Wall Street Journal reported.
That frustration over Google's slow movement has been corroborated by other former Google researchers who spoke to Insider. One former employee said that at such an exciting time for AI, startups offer the opportunity for researchers to have more ownership over their work and possibly make a greater impact. Below are some of the most notable papers in the field of AI, the papers' researchers who have departed Google, and where they are now."
And then they list a huge number of heavy hitters who've left Google.
https://www.businessinsider.com/goo...as and Noam,The Wall Street Journal reported.
Upvote
1
(1
/
0)
Apple hasn't been on the bleeding edge of technology in recent memory if ever. They'd generally preferred waiting for things to be more mature then jump on that technology. This really just seems to be par for the course for how Apple does things.
Also there is the fact that as of yet nobody is really making money in the AI space much like body is really making money in the voice assistance space so Apple probably isn't ready to throw a bunch of money at AI until there is some better understanding how how/if they can actually make money off of it.
Also there is the fact that as of yet nobody is really making money in the AI space much like body is really making money in the voice assistance space so Apple probably isn't ready to throw a bunch of money at AI until there is some better understanding how how/if they can actually make money off of it.
Upvote
-16
(0
/
-16)
Settings -> Siri & Search -> Siri Responses -> {Prefer Silent Responses, Automatic, Prefer Spoken Responses}
Upvote
6
(6
/
0)
If Apple decides to buy their way into the LLM space, they'll probably buy some small-ish startup whose fundamental tech is solid and can be integrated into their products (a la PA Semi).
Upvote
2
(2
/
0)
So for the past few days I had Bing chat do some language translations from English to Vietnamese and English to Spanish. Not single words or phrases but paragraphs and headers. We were converting a number of documents. So I gave it to our staff to judge the quality of the translation. They asked me who translated and to minimize bias I said Google translator. But they said Google is usually awful and that these are really good.
I even did some blind side-by-side comparison where we took the revised English version of a document and then presented 2 Vietnamese versions to the staff: 1) the original Vietnamese version we had that was written by a native speaker and 2) the AI translated version of the new English version. And the vast majority gravitated toward the AI document for almost the entirety. It was a shocking win. I figured there would be wins here and there for various sentences. But no, the AI version almost won every sentence.
I did the same with the Spanish version but we don't have a lot of Spanish speakers on staff so I thought the sample size was a bit too small but they too were very impressed by the AI version.
That's when I spilled the beans and said AI. they were so surprised that AI could write so well in foreign languages. Because from what I understand, almost all the training was in English.
I even did some blind side-by-side comparison where we took the revised English version of a document and then presented 2 Vietnamese versions to the staff: 1) the original Vietnamese version we had that was written by a native speaker and 2) the AI translated version of the new English version. And the vast majority gravitated toward the AI document for almost the entirety. It was a shocking win. I figured there would be wins here and there for various sentences. But no, the AI version almost won every sentence.
I did the same with the Spanish version but we don't have a lot of Spanish speakers on staff so I thought the sample size was a bit too small but they too were very impressed by the AI version.
That's when I spilled the beans and said AI. they were so surprised that AI could write so well in foreign languages. Because from what I understand, almost all the training was in English.
Upvote
4
(4
/
0)
Do you really think many people are going to look behind the scenes at how a LLM produces the results it does?
Using this logic using a search engine should never be trusted, or any closed source software.
Besides, these are people asking for job ad ideas, then picking and choosing the bits they want. On the side, I’m doing my own fine tuning. And the worst thing that can happen is it spits out a bad ad.
Upvote
1
(1
/
0)
There was a bbc report on this ( maybe not bbc ..sorry it was a podcast and I can’t remember which) and they tried a bunch of translation tasks with less common languages or uncommon online like Norwegian and Danish as its common for Norwegians and Danes to use English online, and then reviewed them with native speakers and the results were surprising excellent getting the tone and syntax to sound very natural.
They even tried it on more obscure dialects and while not as good still found it superior to anything else they had tried.
It was still just as likely to get basic facts wrong and make bizarre claims but could produced surprisingly smooth translations
Upvote
1
(1
/
0)
I mean, Siri is a porn star...
...there's a big boob porn star who goes by Siri...
...I mean, um, so I've been told...
Upvote
-9
(0
/
-9)
OrangeCream
Ars Legatus Legionis
It wasn’t long ago that Apple didn’t have a CPU either.
In the grand scheme of things Apple has a few years to figure out how to use LLM to sell more iPhones.
Upvote
7
(7
/
0)
Or it could be that Apple want to spend time working out where AI can be used to improve the end user experience beyond the 'dog and pony show' being put on by the big silicon valley players to entertain Wall Street and boost their share price. New technology goes three three phases: novelty, flogging it to death and then finding sensible ways to use the technology in existing products. We're at the novelty phase but soon we'll move into the 'flogging it to death' phase where AI will be rammed into everything even if it doesn't make sense. I would say that Apple will start launching it once the third phase kicks in - using AI where it makes sense and has material benefits to the end user beyond it being a box that is ticked for the sake of filling up the specifications sheet.
Upvote
15
(15
/
0)
Apple has undoubtedly just asked ChatGPT how a corporation with assets north of $300B, and $50B in cash, should target the AI market. Results will follow.
Upvote
-7
(0
/
-7)
I would say that's notably different from 1+1=x. My problem isn't with talking about the fact that unassisted LLMs can't do significant mathematical operations, that's fine and it'll be very interesting once they can do so, especially if we can use interpretability to see exactly how. My problem is with the persistent pattern among critics of LLMs to give hypothetical examples of supposed failure that aren't true. It's misleading, and I see people who've taken these faked "examples" literally very frequently. It's good to turn a critical eye towards new tech, just don't use faked examples to do it. Yours is great, it illustrates an actual failure and doesn't over-exaggerate the magnitude of the failure.
It is solved in actual use, through the specific feature I noted. It's called ChatGPT Plugins, there's a waitlist to access it that you can find by Googling "ChatGPT Plugins".
Upvote
-2
(1
/
-3)
Not to mention the very serious legal questions regarding copyright and defamation law that these products have brought up. Apple has time to let the dust settle.
Upvote
1
(1
/
0)
I have not been impressed with LLM AI. In one prompt, I asked ChatGPT 3.5 to write a condolence letter to a person who contributed money in the name of the deceased. The AI wrote a lovely letter to the deceased.
Tonight, I asked ChatGPT 4.0 whether an opossum is conscious when it plays dead. It answered yes, it enters a catatonic state and no, that it is fully conscious and can still bite and scratch. Looking at the sources, it included three pest control companies’ websites and Ripley’s Believe it or Not. Not one science website.
LLM is great at blathering without concern for facts. It literally has no idea what it is saying. It merely sounds confident and knowledgeable. Maybe it’ll make a good replacement for Tucker Carlson on Fox News, but Apple isn't missing the boat here.
Tonight, I asked ChatGPT 4.0 whether an opossum is conscious when it plays dead. It answered yes, it enters a catatonic state and no, that it is fully conscious and can still bite and scratch. Looking at the sources, it included three pest control companies’ websites and Ripley’s Believe it or Not. Not one science website.
LLM is great at blathering without concern for facts. It literally has no idea what it is saying. It merely sounds confident and knowledgeable. Maybe it’ll make a good replacement for Tucker Carlson on Fox News, but Apple isn't missing the boat here.
Upvote
14
(14
/
0)
In the meantime, if any of you would like to play around with chatGPT on your iPhone, Macstories has a shortcut for interacting with it. 
https://www.macstories.net/ios/s-gp...through-mode-better-homepod-support-and-more/
https://www.macstories.net/ios/s-gp...through-mode-better-homepod-support-and-more/
Upvote
1
(1
/
0)
Upvote
-6
(0
/
-6)