I don’t understand how Touch ID can only be accurate to 1 in 50k when everyone on the planet has a unique fingerprint. I think Apple have made that up to make Face ID seem more secure than it actually is.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Face ID Appears to Fail at Telling Apart Brothers Who Aren't Twins in New Videos [Updated]
- Thread starter MacRumors
- Start date
- Sort by reaction score
they look nothing alike. you must be smoking that
crack !![doublepost=1509823118][/doublepost]
What? Did you watch the video ?
it clearly shows the issue Take it back. Let someone who appreciates tech get it at a discount.WTF !!! I care, I paid $1000
Ummm, a lot of people care when the "new" tech is apparently less secure than the "old" tech...
I get it. Apple advertises how security conscious they are, but because you don’t care about security, no one else should either.
There is literally millions of people trying extremely hard to make Face ID fail. The surprising thing is that People are actually making a big deal when they manage to make it fail less than 1 day after release.
Especially when we know that 1) its not 100% perfect and 2) the security gets stronger the more you use it.
And I guarantee that this guy did the initial Face ID setup with his glasses on. If I'm not mistaken you'll get better results if your initial setup is done without glasses, hats or anything else obstructing your face. Facial hair doesn't matter die to IR.
Edit: Just noticed that they are wearing the EXACT same model glasses. Different colors but the same model.
I guarantee this guy did the initial setup with his glasses on. Which would explain why the failure occurred when his brother put his glasses on. And anyone that says that these brother are not borderline twins are crazy.
Especially when we know that 1) its not 100% perfect and 2) the security gets stronger the more you use it.
And I guarantee that this guy did the initial Face ID setup with his glasses on. If I'm not mistaken you'll get better results if your initial setup is done without glasses, hats or anything else obstructing your face. Facial hair doesn't matter die to IR.
Edit: Just noticed that they are wearing the EXACT same model glasses. Different colors but the same model.
I guarantee this guy did the initial setup with his glasses on. Which would explain why the failure occurred when his brother put his glasses on. And anyone that says that these brother are not borderline twins are crazy.
Last edited:
they look nothing alike. you must be smoking that
[doublepost=1509823118][/doublepost]
What? Did you watch the video ?
To be fair dumastudetto is someone who resolutely refused to believe the X anything less than perfect, so he would certainly suspect in cases like this.
Apple’s not going to make something up that could easily be fact checked.
Why are people so intent on pissing all over FaceID? Don’t like it then get a phone that uses fingerprint sensor instead.Its not the adding the glasses thats the problem. Its that they didn't setup the phone properly. Your supposed to scan your face without the glasses first, but they setup the phone with the glasses on. That probably messed up the measurements the phone made.
We engineers are constantly looking for and thinking about faults. That's not negativity. Quite the contrary, it's what I call fun. Constantly thinking about ways something could fail is what makes it better. It's like a puzzle, and finding a solution gives us a tremendous amount of satisfaction in the end. My whole work day is about constantly navigating around problems. That doesn't mean we're always negative in real life. I don't think the brothers in the video are super angry, they seem to be smiling. However, there's always someone on the internet who blows everything way out of proportion, angrily attacking public faces of companies, and everyone else. The rest of us can decide if the product is worth the money or not, and move on with our lives.
I've read the first page ...
I think the drop of the iPhone beyond the camera when the 2nd younger brother tested tells me another iPhone X was used. Curious if the icons differed though.
I think the drop of the iPhone beyond the camera when the 2nd younger brother tested tells me another iPhone X was used. Curious if the icons differed though.
It doesn't really work that way, at least not in most common face recognition algorithms.
For example, one common algorithm consists in training a neural network (NN) to generate a 128-dimensional array of numbers for a given person's face. The key is that the NN generates approximately the same 128D vector for the same face regardless of the lightning, facial expression...etc.
So if you have two twin brothers and one random stranger faces and you calculate this 128-dimensional numbers (called encodings) for each of their faces you'd have something like this:
Twin 1: [1.0, 4.2, .... , 7.8]
Twin 2: [0.9, 4.5, .... , 7.2]
Stranger: [1.9, 3.2, ..., 6.5]
Twin 2: [0.9, 4.5, .... , 7.2]
Stranger: [1.9, 3.2, ..., 6.5]
Then imagine you scan Twin 1's face again and the system tries to determine if it belongs to Twin 1 or not. Let's imagine the encoding for the new face looks like this:
Unknown Face: [1.2, 4.0, .... , 7.1]
To calculate the "likeness" of it being Twin 1 you'd have to use some sort of algorithm that could be as simple as using the Euclidean distance between Twin 1 and Unknown Face's encodings. Let's imagine the distance between that unknown face and Twin 1 is the real number 94.3.
You could use 100 as a thresold and conclude that if the distance is under 100 the face probably belongs to Twin 1. But then, maybe, when Twin 2 tries unlocking the phone with its face the calculated distance could be under 100 too, so maybe the thresold should be lower.
How lower? Well, over time you'd get dozens of new scans of Twin 1's face, so you could average them over time (thus updating the 128-D face encoding) but more importantly you would also get a sense of how much the encodings change throughout the day, and you could choose a better thresold so while most scans of the owner's face are accepted, other people would become gradually less likely to be accepted as the encoding gets better over time.
TL;DR: While Apple obviously uses more sophisticated algorithms than what I mentioned above, as a rule of thumb for both facial recognition and neural networks security and accuracy gets better over time (to an extent) as it gets to know you and how much you change over time.
They've done a second video where you can see it's the same phone all the way through
Yes, I think this is where the attention should be... establishing if indeed FaceID does get better over time and with use.
I have a twin, though... so I'm screwed.
Why do we need to establish this? Apple already explained it...countless published articles have already explained it. The only people who don't understand how the technology works are the people who haven't actually done the research. Those are the same people who are posting inaccurate information about the techonology.
Dude its science! But seriously it makes sense if you consider how much harder it is the replicate someones face versus fingerprint. Also Touch-iD doesn't have one crucial element that Face ID has. Machine learning.
Also, read my first post on the subject
https://forums.macrumors.com/thread...ns-in-new-video.2084599/page-13#post-25391709
Do they show the setup, do they show how many times the phone was used prior to the other brother using it, do they show if they setup without or with glasses on? That all matters.
The only people that are stressing this are the people that didn't buy the phone. Most of them are still having trouble getting over the price.
Oh sorry I think we're talking about different videos, i was referring to the one where the little kid can open his significantly older half brother's phone...
Or, Face ID and voice ID in tandem. All 3 would be even better.
[doublepost=1509825446][/doublepost]
I gave him credit for tandem authentication. Most at this website would.
1 in a million. given 7.6 billion people in the world. This means there are 7,600 other people in the world that could look like you...obviously your family members are a good start. This isn't proving anything.
He said his dog looks like him, not the other way around!!! When are people gonna pay attention?!