I don't think these cameras are high enough definition to allow this?
I assume it uses the face mapping technique, they would have to train a model using peoples passport photos/whatever biometric data is stored in your passport chip.
that would be fairly easy I'd assume to train a model that can tell them all apart but the image data from a cctv camera is not going to be good enough surely since you would need a straight on capture of the face, any distortion or shadows/light on the face would likely destroy the possibility for a match.
but then unless you have some photo identity that the government issued for you, how would you be on the system.
they would have to force everyone in the country to get their face on a system? that's like ordering the whole country to go and give fingerprints.
people would tell the gov to do one surely
all machine learning agents and what not are basically programmed bots they don't self learn.(okay too a small extent they do, because they choose which preprogrammed action to take based on the current state, and over time they get better at this, but they still have to be told what that state is)
it's like making a bot to organise photos/files by alphabet or file size and claiming it's "AI" because it does a small amount of math
the dota bot reads memory and cheats that's how it beats humans it will also have less input lag than a human, per pixel accuracy etc.
it doesnt even visualise the data, it has no screen image of the game
there's much more impressive bots people have built with machine learning that rely on an image grab of the screen and regular inputs a human is restricted to. they tend to perform on the same levels as humans once they are trained.
but it would be more efficient to just code a regular bot and less time consuming....
we are at-least 50 years away from anything resembling AI
machine learning AI is basically a calculator
I assume it uses the face mapping technique, they would have to train a model using peoples passport photos/whatever biometric data is stored in your passport chip.
that would be fairly easy I'd assume to train a model that can tell them all apart but the image data from a cctv camera is not going to be good enough surely since you would need a straight on capture of the face, any distortion or shadows/light on the face would likely destroy the possibility for a match.
but then unless you have some photo identity that the government issued for you, how would you be on the system.
they would have to force everyone in the country to get their face on a system? that's like ordering the whole country to go and give fingerprints.
people would tell the gov to do one surely
Throwing AI at any machine learning bot created so far is a massive stretch of what AI means.People aren't going to be thrown in prison on the basis of AI though, they can simply flag up images for humans to review.
all machine learning agents and what not are basically programmed bots they don't self learn.(okay too a small extent they do, because they choose which preprogrammed action to take based on the current state, and over time they get better at this, but they still have to be told what that state is)
it's like making a bot to organise photos/files by alphabet or file size and claiming it's "AI" because it does a small amount of math
the dota bot reads memory and cheats that's how it beats humans it will also have less input lag than a human, per pixel accuracy etc.
it doesnt even visualise the data, it has no screen image of the game
there's much more impressive bots people have built with machine learning that rely on an image grab of the screen and regular inputs a human is restricted to. they tend to perform on the same levels as humans once they are trained.
but it would be more efficient to just code a regular bot and less time consuming....
we are at-least 50 years away from anything resembling AI
machine learning AI is basically a calculator
Last edited: