Ben Lovejoy: I suspect Apple limits Face ID to one person because it would otherwise be too slow

Ben Lovejoy, 9to5Mac, pulled a quote from this Mashable iPhone X review:

One important limitation of Face ID: It only lets you register one face. That may strike many as unnecessarily limiting since Touch ID lets users register up to 10 [sic] fingerprints, but Apple says it found the number of people who register more than one person’s fingerprints is miniscule.

Ben continues:

The idea that hardly anyone registers more than one person’s fingerprint didn’t ring true to me, and our poll shows that it’s not true for 9to5Mac users at least.

At the time of writing, the majority of our readers have more than person’s fingerprints registered for one or more of their iOS devices. Almost half (48.98%) have one other person registered, and a further 6.85% have more than two people.

I would certainly agree with this. I have my entire family “fingered” on my iPhone 6s, and the Touch ID performance is still very fast.

Ben’s take on adding a second face to the Face ID registry:

My guess is that doing all these checks for more than one person would make face-recognition noticeably slower than Touch ID, and Apple was concerned that reviewers and consumers alike wouldn’t respond well to that. That, I think, is the real reason Apple limits Face ID to a single face.

Interesting perspective.

  • lkalliance

    Yeah, the “miniscule number of people” statement seems disingenuous. For one, on introduction Touch ID allowed a maximum of five fingerprints, now it allows ten. I suppose that someone could choose to register each of his ten fingers, but still it seems odd that that would be the reason for the increase.

    Secondly, how would Apple know this? Fingerprints are only stored locally on the Secure Enclave, and not synced to iCloud. Apple shouldn’t know the configuration of any iPhone’s Touch ID settings, should they? Or is that bit of information (how many fingers) synced while the actual fingerprints aren’t?

    • How would Apple know this? My guess: Something more likely to be accurate than a web poll on a tech nerd site.

      Rolling dice would be more accurate than trying to apply tech nerd data to the market at large, but I suspect Apple found something much better than that. Maybe a survey?

      • A survey of whom? Your response is pure speculation. You don’t know that Apple did anything more than make shit up.

        • You’re assuming they made shit up and then lied to cover it.

          That’s lizardmen conspiracy level stuff. A survey would be cheap for Apple.

  • John Voliva

    The readers of 9to5 and Loop are not “casual” Apple users. They are the power users — so taking a poll of the readers will lead you to skewed results, unless you are looking for power user information.

    I wholeheartedly believe Apple’s claim that a minuscule of people have only one person registered on TouchID. In my small circle of family members, we have about 15 people who own iDevices. No one has anyone else’s fingerprints registered. I would wager to bet that is the same for upwards of 99% of iDevice users. I bet if I would reach out to co-workers and colleagues (all in the pharmacy field), I would be hard pressed to find one or two people out of the couple hundred I know that would have more than one person’s fingerprints registered.

    We too often get trapped in the bubble and forget about the non-tech savvy consumers who make up the overwhelming majority of the installed base.

    As for Apple knowing this, dollars to donuts they did a consumer poll, using an outside agency specializing in user feedback.

    • Your belief— your word — is pure speculation. And your use of your small circle friends as a sample set doesn’t disprove 9to5 Mac statements. Since I don’t know you, I will ignore your statements.

  • I would put $1,000 on Apple having better data than you or 9to5mac, Dave.

    • lkalliance

      I generally agree, but where would Apple get the data? Fingerprints are stored in the Secure Enclave and remain on the device, not synced to iCloud. Is the data on total number of fingerprints stored in iCloud backups (just not the fingerprints themselves)?

      EDIT: Posted this before reading John Voliva’s response above, and that makes some sense.

    • Why would you do that? That’s an irrational statement.

  • john doofus

    Seems plausible. If so, we may be waiting a little longer for FaceID on the Mac, and maybe the iPad too.

  • Michael

    I believe that Apple is highly likely to have an accurate read of how many people regularly have more than one person registered via TouchID.

    And “slow” isn’t the right theory, IMO. If there’s a technical problem, it would be issues with accuracy for training the neural network in the “bionic” part of the A11 when there’s more than one face to learn. This feature isn’t just simple static coding. The neural net has to adapt to meet the spec, and having more than one thing to learn is a much larger hurdle to clear than just adding a second face.

    • Argh. I was beaten by more than seven hours here. 🙂

    • And “slow” isn’t the right theory, IMO.

      Agree. Lovejoy need only to go back to Apple’s gamete for a generalized answer. Slower, inaccurate, a timeline that didn’t allow for the feature to be perfected…whatever the reason, Apple solved the FaceID customer experience for one user (note how its competitors have done).

  • lkalliance

    In an imaginary future…

    …Apple allows multiple Face ID faces. It also adds a “bare bones” login type, where you designate certain apps or data that will not be shown. You designate certain fingerprints and certain faces to unlock to bare bones instead of fully. Now I register my own face twice, each time making a different face (sticking my tongue out, arching an eyebrow, having the right eye closed, making fake lips). Look at it one way: full unlock. Look at it the other way, bare bones unlock.

  • Just to add my anecdotal information:

    Both my wife and I have iPhones. Neither of us have our fingerprints registered on both devices. We do know each other’s passcodes. Both of us are fairly tech savvy nerds (I do IT work; she’s a graphic designer.)

    Take it for what it is worth. However, I’m sure we aren’t the only families like this, so add us in with all the non-techie masses, and I suspect Apple is correct in their assessment. Face ID for the owner and the passcode for the rest seems fine for now. Just as they added more fingerprint capability, they will add more face recognition as the tech grows.

  • franksspam

    What I don’t understand is, why are we wasting all of this time trying to figure out if someone has more than one person’s fingerprints on their Touch ID device? Why aren’t we asking if they have more than one fingerprint encoded at all? I have four of my fingers stored, and I guarantee that just because they come from me, they are completely different from my other fingers. I know this because I use my fingers that are not encoded when I want to turn on the screen to see notifications that have popped up since my last look at my phone. My phone NEVER opens when I use one of my non-encoded fingers on Touch ID. So, the normal operation of Touch ID is for it to be able to recognize or not recognize different fingerprints all day long.

  • rick gregory

    “…and our poll shows that it’s not true for 9to5Mac users at least.”

    Honestly, every tech writer should have to pass some basic stats courses. Tip: a self-selected group on a site that’s not representative of the general population has no statistical validity. Polls like that mean nothing about the general populace and they’re suspect even for the site’s population since the sampling isn’t random even within that.

    • lkalliance

      Seems like he didn’t try to indicate that it applied to the general populace.

      • rick gregory

        He wrote an entire article on the basis that it was an issue without really addressing the fact that the poll means nothing.

        Yes, he notes it… but my overall point is that the article – and all articles based on “a poll on our site indicates…” logic – has no meaning outside of the people who came to the site, saw the poll and decided to take the poll. You can’t derive anything from those kind of polls. At all. So why try to use such a poll as justification for his opinion?

        If he wants to write it as an opinion piece, by all means go ahead, but don’t try to buttress arguments with data that is inherently meaningless.

        • lkalliance

          But of course the poll means something. If it is representative of the people who visit his site, then it’s likely representative of most people actually reading the article.

        • Are we to assume that Apple’s sampling is any better? Did Apple sample across all ethnicities, encomics classes, cultures, etc. Since we can’t know, should we also take Apple’s statement with a grain of salt.

          • Obviously yes, by surveying across their customers instead of a web site. Do you have any idea how statistics work?

          • You haven’t provided evidence of anything.

          • rick gregory

            You… don’t seem to understand how professional polling works. Go read up, then we can talk.

          • How do you know it was professional?

          • rick gregory

            Yeah, Apple just grabbed a couple of high school kids and asked them to set up a site with Polldaddy. Right. Apple wouldn’t, if they did a poll, use a pro. Nah….

            OK, I think we’re done here.

  • komocode

    i could definitely see something related to multiuser sign in with multiple people registered in FaceID in the future

    • rick gregory

      Why? Phones are personal devices. on iPads it might be useful since I can see those being shared a lot more but I just don’t see a use case for multiuser accounts, etc on your phone. Aside from the kid-parent issue where parents want access to their kids’ phones, how many cases can you think of where one person needed to be able to sign in on another person’s phone?

      • komocode

        yes i meant ipads.

      • David Stewart

        The kid-parent issue goes both ways, kids on parents’ phones, parents on kids’ phones. Add in the husband/wife sharing and a typical family alone provides a significant volume of use-cases for multi-user accounts on iOS devices.

        • rick gregory

          But that assumes people routinely share their phones and I’ve not seen that. I’m not saying it doesn’t exist, but when people act like it’s normal and therefore Apple must be wrong or lying about the fact that they see a small percentage of people do this. The thing is, if only a few percent of users do this, that’s a LOT of actual people given the number of iPhones sold… but it would still only be 2 or 3 or 4 or 5%.

          • David Stewart

            I’m sure Apple has solid numbers on this. I’m also pretty confident Apple doesn’t care so much what people are doing now so much as what they will be doing in a year or two. They also know they can push people into behavior with their design and implementation. Having a parents phone be able to have a “kid” mode that hides all the important things, but displays some videos or games seems like a pretty solid feature people would use if done well (not too dissimilar to the parental controls in iOS now). Likewise sharing an iPad would be much nicer with separate accounts for each user. As iOS becomes more and more an alternative to macOS, user accounts seems inevitable.

  • David Stewart

    It should be relatively easy for Apple to only monitor for the primary user, but have a button on the lock screen that lets you tell it to use a secondary Face ID profile. This would probably push Apple to support full mulit-user system, so there is a good deal of plumbing to be done behind the scenes to make this work as well as it probably should.

  • Thunder le ouf

    It should be relatively easy for Apple to only monitor for the primary user, but have a button on the lock screen that lets you tell it to use a secondary Face ID profile. This would probably push Apple to support full mulit-user system, so there is a good deal of plumbing to be done behind the scenes to make this work as well as it probably should.

  • I doubt it’s speed, honestly. I see a couple more likely problems, like training the right neural net or low capacity (based on the size of the data).

    • David Stewart

      My guess is they would need to train a different neural network for each face. I imagine the architecture is designed to run the detected face through a single network, running more than one network in parallel probably isn’t part of the design. They can train multiple networks, but swapping in the parameters for each network would certainly add a delay to face detection.

      • Yeah, but there’s no way it would double the time (for example), and even if it did doubling wouldn’t bring down the speed of recognizing the first face enough. That’s just the snazzy hat on the top of the problem. The real problem is juggling the neural nets.

  • It’s almost as if Apple is just making shit up.

  • brisance

    The conclusion is wrong, and reflects Ben’s misunderstanding of how neural nets and supervised machine learning in general works.

    FaceID generates a mathematical description of your face during enrollment. No one has access to this description, even Apple, because it is stored on the device.

    Now the time comes to authenticate. FaceID scans the face that is presented. The mathematics comes into play again; this time you’re wearing glasses/a beard/whatever. Obviously not the same face that was enrolled, but you’re allowed to access the phone. Because FaceID generates an output with a certain confidence threshold (between 4.5 and 5 sigma, based on their 1 in 1 million statement), and if the sample is within the threshold, it is considered authentic. Obviously it doesn’t take a lot of time to determine since it scales linearly O(n).

    Note that which features are weighted more heavily, how the features are calculated etc are not revealed, since these are expensive to acquire and train and can be used to defeat the system, and so usually end up being trade secrets.