Apple Reportedly Reduced Face ID Accuracy to Speed Up iPhone X Production

[deanthedev]

I’m still buying the iPhone X for my family. Made-up rumors don’t affect my purchasing decisions.

It is a amazing how fans of a company are unwilling to accept they did something negative, no matter what. The article makes sense to me. There were reports earlier that they could not produce a high enough yield on these parts that have to be mated to work together. Then all of a sudden its reported that the problem is solved. Now this article says that they loosened the spec on the parts, which would mean to me they loosened how tightly the parts needed to be paired. I would fully expect from this that the accuracy of FaceID was reduced. As someone who's worked in manufacturing, this all seems logical. There is a concept called Design for Manufacturability, that they apparently didn't follow very well in the design of X. What is unknown is what impact the reduction in accuracy will have on the user. It could mean that its less secure but in turn has less false fails. It could mean its harder to pass and has more false fails. It could have no impact at all. I will say that no one outside Apple's engineering department will ever know the answer to that question, since no one will have access to an X made to the original spec, or if they do they won't know it. The bottom line is going to be reviews from people who have the final product. I still stand by my position that removing TouchID for FaceID was a dumb design move. Now that I've had the Note 8 for a while and it has multi factor, I like how quickly I can unlock by the race that occurs... first one wins.

It’s absolutely amazing to me that people will automatically accept negative rumors about Apple.

 

[DoctorTech]

A company reduces the accuracy of their key feature just to be able to up their production? I don't believe this for a second. It would be utterly stupid if they did this.

I was a quality manager for a medical device company for several years. Things are not always as black and white as they seem. There were times when we would receive parts from a supplier that did not meet some characteristic on the part's print. We had a Material Review Board (MRB) comprised of several engineers who, on a case by case basis, had to decide whether to...
1) send the parts back to the supplier (and possibly shut down an assembly line)
2) rework the parts in-house to make them meet the print
3) issue a Temporary Engineering Deviation (TED) to allow us to use parts in production.

The main factor in the decision of the MRB was that we could not in any way compromise the function of our products in terms of performance or reliability. Most of the time when we issued a TED it was to use a part that only had a cosmetic problem (the color of the material was slightly off or the surface finish wasn't quite as sooth as it was supposed to be). There isn't enough information in this article (assuming it is true in the first place) to say whether Apple is compromising performance or reliability. However, it always looks bad to say you are opening your specs to use parts that would not have met your original requirements.

 

[phpmaven]

It is a amazing how fans of a company are unwilling to accept they did something negative, no matter what. The article makes sense to me. There were reports earlier that they could not produce a high enough yield on these parts that have to be mated to work together. Then all of a sudden its reported that the problem is solved. Now this article says that they loosened the spec on the parts, which would mean to me they loosened how tightly the parts needed to be paired. I would fully expect from this that the accuracy of FaceID was reduced. As someone who's worked in manufacturing, this all seems logical. There is a concept called Design for Manufacturability, that they apparently didn't follow very well in the design of X. What is unknown is what impact the reduction in accuracy will have on the user. It could mean that its less secure but in turn has less false fails. It could mean its harder to pass and has more false fails. It could have no impact at all. I will say that no one outside Apple's engineering department will ever know the answer to that question, since no one will have access to an X made to the original spec, or if they do they won't know it. The bottom line is going to be reviews from people who have the final product. I still stand by my position that removing TouchID for FaceID was a dumb design move. Now that I've had the Note 8 for a while and it has multi factor, I like how quickly I can unlock by the race that occurs... first one wins.

Well if you work in manufacturing then you know that just relaxing the specifications on one part doesn’t necessarily mean that accuracy would be reduced. If they said "let’s reduce the tolerances needed for the laser module from plus or minus 2 microns to plus or minus 4 microns because it works perfectly well even if it’s off by 4 microns. Our manufacturers are having a really hard time meeting that tight tolerance and it doesn’t need to be that tight.", does that automatically mean that accuracy is reduced? Obviously not. All this story means, if it’s even true, is that they realized that the tolerances didn’t need to be as tight as they though for the module to work perfectly.

This isn’t about blindly supporting Apple, but about not making wild assumptions based on the rumor that they relaxed the tolerances on a part. Relaxed tolerance <> Reduced accuracy.

By the way, I was a CNC machinist for many years. Just so you know where I’m coming from.

 

[MarkB786]

Drop your phone = PIN code from then on.

 

[Tech198]

That's one way to speed up production, make it less accurate..

Apple probably had a very good reason and knew it still was enough not to compromise security even though it meant slightly lower.

On one i'm surprised Apple did this, because Apple is all about perfection.. but on the other hand, iPhone X would probably never come out then..

You can be sure this story will be reported to death with the tagline "FaceID not as secure as people think" by the media... Anything to get a story.

Its secure still..

 

[DevNull0]

"Face ID is designed to work with hats, scarves, glasses, contact lenses, and many sunglasses. Furthermore, it's designed to work indoors, outdoors, and even in total darkness."
Total darkness? You mean it unlocks before the screen is turned on and lights up your face?

What do you think it uses light from the screen for?

What's more impressive will be if it does work well outdoors in bright sunlight.

 

[jbizzybeetle]

For the record, this is a MUCH better and non-clickbaity title.

https://9to5mac.com/2017/10/25/bloomberg-face-id-iphone-x-production-quality/

Mayo is much more balanced as a commentator in context on the Bloomberg article.
As a former, 25-yr journalist, I can say today’s online standards are poor and no-name sources would not previously have been allowed to totally drive a story.
That said, I also realize how desparate writers are to see their material read and how little good iPX info made available to them.
On the humorous side, yippee!
Drive away those potential buyers—more chance for me!!!

 

[verniesgarden]

This won’t backfire at all....

 

[brian3uk]

Didn't read everyone else's comments but this only makes me believe my first impression of this phone was right. It's basically a beta phone being sold at a premium price. Next year will be much better and likely return to a lower price point. This years phone's resale value will tank and everyone will be mad.

 

[mdellepi]

By the way, do we know when the change has been made? As far as we know, the 1000000:1 figure at the keynote could have already been the post-relaxation one.

Also, the correct wording used by Bloomberg when actually describing the rumor is "Apple relaxed some of the specifications for Face ID". "Relaxing" is not "reducing". It means that, if before they required all produced dot projectors to differ by no more than a certain value, now that value is larger. As stated by Bloomberg, we have no clue about "how much the new specs will reduce the technology’s efficacy".

 

[Smartass]

This is just clickbait on Bloomberg's part IMO. Unnamed source said Apple reduces quality.

Baseless accusations I think. Just trying to get a few ad views from the hype/rumours.

And what kind of source do you think it should be? Something like:

"We reduced FaceID accuracy to speed up our production"

- Tim Cook

Ain't never gonna happen.

 

[Rafterman]

I’m still buying the iPhone X for my family. Made-up rumors don’t affect my purchasing decisions.




It’s absolutely amazing to me that people will automatically accept negative rumors about Apple.

Because more and more of them are turning out to be true. This isn't the first sign that the X was rushed to production.

 

[Val-kyrie]

To those who dismiss this article a priori: (1) this article is too detailed to be a baseless accusation and (2) Apple's refusal to comment on the story lends validity to its claims.

If Apple believes a change in the tolerances for the components related to FaceID will not adversely affect its function and will simultaneously increase production, then I have no trouble believing Apple would issue such a change to the iPhone X production. It is simply logical.

The only way to verify this is with independent testing once the iPhone X is released.

On a personal note: this smacks of a compromised first-gen hardware product. I have learned the hard way to listen for the warning signs and avoid first-gen designs, especially from Apple who tend to push the envelope too far. The prudent move is to wait.

 

[UK-MacAddict]

And that mantra very nearly drove them into the ground.

No, a bloated line up of crap products such as the Newton and continued rehashing of the Apple II is what nearly brought them down.

Steve came back and trashed the whole line up and focused on a few top quality products which transformed Apple into what they are today.

 

[SwissGuy93]

OK, finally making progress.

No doubt you are correct in that the neural net makes a de facto template to which the pattern of dots serving as the input are compared. We also know from the keynote that the template is dynamically updated (probably each time you use FaceID there are incremental changes to the template). It can also be assumed that the template captures spatial correlations among the infra-red dots, so can handle partial obscuring of the face. The question is how the template matching process is affected by variability with respect to the user wearing one or more pairs of glasses or contacts. As it stands now, we don't know. That is all I am saying. And I note so far as I can tell nobody seems to be able to find an actual statistic in the Apple literature that specifies what the false positive rate is. Any engineer working on this problem would use signal detection theory to describe how well the system works. Apple has the information, they just haven't released it. Again, that makes me have a healthy scepticism.

The false positive rate is 1:1'000'000 and Apple always communicated that quite clearly. The false negative is not communicated, but that wasn't communicated for the Touch ID either. Well, healthy scepticism would be able to tell, that Apple released everything they could in order to not give too much away to the competition.
Apple will never give detailed infos about how it handles glasses etc. because they assume that it just works and that the information would be not necessary and therefore even giving to much information to their competition.
Again, I would recommend that you look at the white paper. There are a lot of information in there, and I think some of your questions will be answered.

 

[iLLUMI]

Says nothing about accuracy. I may be wrong but it seems like all that Apple did is loosen some of the tolerances of the dot projector? Maybe they’re just accepting more dead ‘dots’ allowed per dot projector? Who knows.

Maybe there's a dead dot doctor that can detect the dead dots on the dot projector to make it project dots better?


But seriously, I guess if the projector works with a fairly high success rate then that's possibly going to be acceptable enough.
I'm really curious now to see what real world usage is going to reveal. Looking forward to having a play when these hit the Apple stores.

 

[SwissGuy93]

What group has the ability to prove the accuracy isn’t 1,000,000:1? How do you even quantify that?

With math. The probability was most certainly based on mathematic calculations and test cases during the development. According to Apple over a Billion datasets have been used to test Face ID.

 

[Tech198]

To those who dismiss this article a priori: (1) this article is too detailed to be a baseless accusation and (2) Apple's refusal to comment on the story lends validity to its claims.

If Apple believes a change in the tolerances for the components related to FaceID will not adversely affect its function and will simultaneously increase production, then I have no trouble believing Apple would issue such a change to the iPhone X production. It is simply logical.

The only way to verify this is with independent testing once the iPhone X is released.

On a personal note: this smacks of a compromised first-gen hardware product. I have learned the hard way to listen for the warning signs and avoid first-gen designs, especially from Apple who tend to push the envelope too far. The prudent move is to wait.

"Apple's refusal to comment" perhaps they also don't wanna cause panic in everyone either by announcing anything.... Apple's very good at deterring away fro what it thinks would cause a stir with competitors..

 

[AdonisSMU]

"Apple overcame its production challenges by quietly telling suppliers they could reduce the accuracy of the face recognition technology to make the iPhone X easier to manufacture."

Quietly scratches the iPhone X off list of possible purchases.

I've got a bridge to sell you.

Lets be honest. Apple made this decision before they demoed the device. You are not going to be changing a production process on the fly. Hardware is not software...although people like to treat them similarly.

 

[Macalway]

What they probably did was to change the accuracy requirements for one or more of the sensors involved in the technology.

Makes you a little pissed off, whatever

 

[Chupa Chupa]

Yes, but others are quickly piling on clickbait headlines, and Bloomberg jumped to some conclusions as well.

Here's the source "accuracy" evidence:

To boost the number of usable dot projectors and accelerate production, Apple relaxed some of the specifications for Face ID, according to a different person with knowledge of the process. As a result, it took less time to test completed modules, one of the major sticking points, the person said.

So it depends on exactly what specifications were relaxed in order to speed up the tests.

Sure. All true. None of that changes anything I wrote though. Unnamed sources are an accepted part of journalism as-is the reporter reporting more than "just the facts." It's up to the reader to attach a level of validity to any article. One hopes the news org. did some kind of verification but that's not a guarantee these days. But even the lack of that is not an indication of the article attempting to be a hit piece vs just trying to gather eyeballs. Pack journalism has long been part of the news game too so expected that once one outfit reports something attention-worthy others will follow.

Ultimately, though, as-said, I doubt this report, flaws and all, is dissuading early adopters. I know I'm still all-in. I have two weeks from shipping to me to test it out, so no risk on my part. It's hard for me to accept that Apple (or any high-end company) would sacrifice their brand by significantly reducing effectiveness or security of something as vital as Face-ID. Apple has staked it's future on it. If it fails early on because of this "lower" standard it's quite a lot of hat eating. I don't think Apple cares to binge on hat.

 

[UK-MacAddict]

Actually it drove them to produce the iMac, the iPod, iTunes, the iPhone and iPad. Now we see profits before products, its my opinion that they used to believe if you created a great product, the profits would follow.

This is very true. The product lines are getting to bloated yet again which was their downfall the first time around. Now I know this time they are highly unlikely to get to that point again due to the large cash pile they have but they should stop having so many variations of the same product.

It was much better when there was one iPhone which had all the best technology inside and everyone paid the same price and got the same product. Now with the iPhone, iPhone Plus and iPhone X there’s always a trade off depending on which one you buy. The same is true if the Mac and iPad line ups.

 

[brian3uk]

And what kind of source do you think it should be? Something like:

"We reduced FaceID accuracy to speed up our production"

- Tim Cook

Ain't never gonna happen.

Why is it so unrealistic? Reducing the sensitivity of the sensor a little bit sounds exactly like something Apple would do to get to the "it just works" point.

 

[SwissGuy93]

I don't think Apple needs any help to put this phone in a bad light, they'

Agreed, we really have no idea if this is an issue or not.

Why are hating the phone so much? I think it's funny that if Apple hadn't launched a OLED bezel less screen then they would have been called out too.
And the whole "Touch ID on the back" is not an option and you know it as well as I do. When the rumours were swirling that Touch ID would maybe be moved to the back everybody was against it. I don't think I read a positiv comment about it at the time. Then there was the no-Touch-ID-rumours and now you want it on the back? Sorry, but Apple can never please everyone. And if you want a standard top bezel (like the Pixel 2 XL, Samsung S8, Note 8 and LG V30, which all practically look the same) and Touch ID on the back it would look like just another Android phone.

 

[deanthedev]

This is the perfect news article for the Apple haters.

- We aren’t going to be able to look at sensors in a phone (for example, from an iFixit teardown) and spot any irregularities.
- We aren’t going to be able to take early production iPhones and compare them with later versions to verify if there are any differences in the sensors.
-We won’t be able to extensively test newer and later iPhones with enough faces to be able to verify that one sensor is more or less reliable, accurate or faster than another.

In short, there is absolutely no possibility anyone will ever be able to verify this rumor. In spite of this people will be able to claim that we’re getting an inferior version of FaceID (the signature feature of the iPhone X and the future of authentication for Apple) in our phones. They will follow this up by transferring the burden of proof to others asking them to “prove a negative” (prove Apple didn’t do this). Since nobody can provide any proof they can claim victory.

I don’t think I could have come up with a better rumor if I tried. It’s basically perfect.