(Music)
-------------------------------------------
Qu'est-ce que c'est ? - TRIVIAL GURIPA #5 in Normandy - Duration: 2:32.Hi folks!
We're at Sword Beach,
at the Ouistreham- Rivabella Beach.
Here, bathers go by some remnants of the war,
scarce in the region.
Do you know what they are?
We asked this question in social media.
Yes, they're the obstacles known as Dragon's teeth.
In earlier episodes, we talked about different obstacles and traps
that the Germans laid along the beaches of Normandy.
Among them, the Dragon's teeth
weren't used as much, and only a few remnants still stand today.
They are big blocks of concrete in a pyramidal shape,
that worked as an anti-tank obstacle
or to stop any vehicle in general.
Usually, they occupied a very long line
and were distributed in several rows.
Land mines and barbed wire were also laid in between them.
During WW2, they were used by both sides,
but are usually associated with the German one
and were employed particularly on the Siegfried Line:
the defense system that protected Germany's west border.
Despite their symbolic name,
in reality, the Dragon's teeth weren't as fierce as planned.
It wasn't hard for the Allied engineers
to make way through them
either by destroying them or by burying them.
But don't worry folks,
to see them, you don't have to go that far.
You can just visit Sword Beach,
where you'll find these fine examples.
Their use in the Atlantic Wall can also be seen.
So now you know,
the next time you come to sunbathe
or swim at Ouistreham-Rivabella
don't forget to visit these remnants of the Battle of Normandy.
If you liked this trivia,
hit like, share, and follow us in social media.
Until next time, folks!
-------------------------------------------
A-Trak And Eli Gesner On Their 'Impossible' Young Thug Video 'Ride With Me' - Duration: 8:39.A-Trak, with Falcons, has produced a really great song with performances by Yung Thug,
and 24 hours.
Hey everybody, my name is Eli Morgan Gessner, and I am the style editor here at Uproxx.
And today, we have the wonderful story of a young Canadian boy who took his OCD and
bad posture, and used it to become one of the greatest DJs of all time.
Ladies and gentlemen, it's my old friend, DJ A-Trak.
Hi.
You're a DJ.
You were, a five time?
World champion, yes.
DMC world scratch champion.
It wasn't all DMC.
I started scratching and messing around with DJ'ing at thirteen, and then at fifteen I
was world champion and I kept entering more battles and I accumulated five of these world
titles.
Fifteen years of age! What a revelation!
I didn't see much sunlight from the age of thirteen
to fifteen.
We all grew up listening to sort of classic rock, and, you know, through like Beastie
Boys and Cypress Hill, that was like the funnel to get into hip hop.
I really fell in love with hip hop around ninety-four with Wutang, Biggie, that whole
era.
So I'm listening to hip hop, right?
And I'm hearing scratching on records.
I tried playing the piano but I wasn't very good at it, or I just, it didn't feel like
my instrument.
And, one day I tried scratching a record on my dad's record player, and I discovered
a knack.
I showed my brother and his friends one day, and they were like yo, what the hell, you
can scratch?
We can't scratch.
What the hell?
And, I started practicing every day, and I would come home after school, practice, have
dinner, do homework, after dinner, and then that was my day, every day.
I had this sort of general idea of making a skate video.
I ended up making the video with them.
The concept, being that because of our relationship, back at Zoo York, in the nineties, I on one
hand, was like oh, let's use the old hi-8 video camera that we shot all the original
Zoo York footage with.
It's the impossible video.
Thug showed up, on time, but there was a lot of hanging out time in the van.
At one point, you opened the door and said, guys we're losing sunlight.
They ran right out.
We did the video.
For this song, I hit you up, it was sort of like hey Eli, this is your world, can you
help me figure this out?
And then, you hit me back with this super duper, dream reply of like, here's a test
cut of like twenty seconds of this secret footage that no-one's seen before, that I
just happen to be digitizing, how about we do that for your video.
And, I still have the camera, and we filmed Thug with that camera.
And I'm reading my email and I'm like, brain explodes.
Okay, okay, okay.
Can I call you Eli?
One thing that me and you have been talking about, and this is the current state of media
and culture and music, which I've been a little bit like, uh, I don't know guys.
It seems like, the idea of we're trying to make something original, has become secondary
to I know ya'll like this so here it is again!
Like, that's kind of where I feel there's a shortcoming in culture.
It's not like, oh, there's clearly Biggie Smalls, and there's clearly Tupac, and there's
clearly De La Soul, and that's clearly Public Enemy.
But you've always been more optimistic about it all.
What I would say is, that, in fact, right now in hip hop, there's something for everyone
for sure.
And maybe you're referring to what a lot of people would call maybe soundcloud rap?
Or just like a certain form of sort of druggie, very free kind of abstract rap?
But there's that, but you can also go and listen to some rappidy rap too.
Neil Soul is back!
Like, there's something for everyone, for sure.
So, I don't know.
And by the way, when you were saying that, you know, maybe that type of rap...
You seem to be hinting that it's less original.
That's not, that's not really what I'm getting at.
And I'm sure like anything, it's like people being like, you're a DJ.
Like, on the radio DJ contest?
Yeah, now people would be like oh you're a rapper, do you sip lean, and is your name
Little something.
But by the way, part of what's cool, even about that scene, is that being weird is celebrated,
and it took hip hop a long time to break through to that.
Then when people were like, oh this worked for that guy?
I'm going to do something quite similar.
But I don't even think people approach it with that much of a sort of cheapish sameness.
I think it really is that legitimately that, this is a culture.
And you know, if someone is listening to a certain kind of music and then they have aspirations
to make that music, maybe their first couple of records will sound like what they listen to.
One of the differences between now and the nineties is that that first song that someone
makes, the whole world gets to hear it, whereas back then you had to get a record deal and
you wouldn't hear it until they honed their skills.
Yeah.
A lot of the guys that get dismissed now for being samey, a year later developed their
own identity.
It's just that the removal of the gatekeepers with everything just being posted onto soundcloud
right away, you get to see that development stage.
And you know what, another thing that's fascinating by the way, is a lot of the rappers who seem
to have basic skills are actually a lot more skilled then they give off the impression,
and they choose to make this kind of rap because there's an immediacy that is super punk-rock
and undeniable.
So, it's funny hearing some of those rappers who might get popular from having a song where
they're just going amasasavasasaamagavasasana, amagadasavanasadada over like a distorted
808.
And then, you can, you can interview one of these guys, and he'll be like, oh but I also
have, oh what's the term...
It's not backpack rap.
I'm trying to think of the term.
Lyricist rap?
But it's funny, like I saw an interview with XXXTentation, where he was like, oh I have
like earl type of raps.
Because earl sweatshirt is, you know, is lyrical.
And then, I've heard those records and it's true.
So, a rapper who might be known for blown out distorted, and kind of make ignorant records
is also fully capable of making rappidy-rap records.
And so when some of the old heads will just say, like oh what happened to the skills.
It's more...
It's deeper than that.
It's a conscious decision to make records that translate in a live setting.
And by the way, live rap is blowing up too.
Let's talk about this.
Me and you both have had the distinction of working with Mr. Kanye West.
Yours was far more successful than mine was, but, he was smart enough to pick you.
Damon at that point in time was starting a sort of off shoot imprint that was supposed to literally
be a rock label.
And, the day before Kanye saw me in London, Damon Dash saw me.
And he wanted me to DJ for Samantha, maybe not knowing that she was actually a DJ as well.
And Dame kept trying to pair us for about twenty four hours until I met Ye.
He was really heavily trying to be like, Samantha, this guy A-Trak, he's a really good DJ.
He's going to DJ for you.
It's going to make your show cool.
And she was like, Dame, I'm a DJ, I know DJs, but she didn't really want me to DJ for her.
And, I was just like, wait, I think I know your brother, and this whole thing.
And then, next thing you know, I met Ye and he was like, you're going to DJ for me.
And then, as it turns out, we're having this conversation at this whole Rockefeller thing
shindig.
And so Dame's also somewhere, and Samantha is also somewhere.
And Kanye and I are making are talking and we're making our master plan.
And he's like, da-da-da, I'm going to take you on tour, and the crowd's going to do this,
and say that, take this guy's number, he's my manager.
And then Dame sees what's going on, and I'll always remember this.
He screams, he goes, Samantha!
You see what's happening?
Kanye is about to hire A-Trak!
That's why he's Kanye West!
I was trying to have you work with A-Trak.
But Kanye West is stealing A-Trak from you right now, Samantha!
It's so Dame.
And both she and I knew that this Kanye thing was probably best for everyone.
What up, this is A-Trak, and you're on Uproxx.
-------------------------------------------
🔴Live hi #Roadoto 70 chillen mit euch cool Dienstag Stream - Duration: 35:07.-------------------------------------------
Diferencia Entre Nada y Nadie | #EspañolReal - Duration: 10:27.-------------------------------------------
FACE 2 FACE: El Shaarawy and Pellegrini interview each other! - Duration: 8:07.- Hey, Zio! You good? - Hey, Zio! Yeah, all good.
Joseph asks: "Why does everyone call you 'Zio' [uncle]?
Right. And why do I call everyone Zio?
It started off in my last year at AC Milan.
There was this security guard at Milanello.
He and I used to call each other 'Zio'. I can't remember exactly when it started.
Then when I joined Roma I just started calling everyone Zio.
That's how it all began.
Marco asks: "Your signing was announced with a video of you playing FIFA.
"Do you actually play at all and are you any good?"
Yeah, I do play from time to time
but not that often.
We don't really get that much time.
I'm no world-beater but it's the taking part that counts, right?
Ok, my turn, Zio.
This is from Stewel92Fan.
They call you Il Faraone. What nickname would you give me?
- Zio. Ah, come off it. That's too easy.
Zio... Zio's good.
Matteo wants to know how much you celebrated the first goal against Chelsea.
And who scored it?
Who could it be?
So, did you celebrate?
I sure did, especially as I never thought you'd score from there.
The things you expect least are always the best.
But yeah, I did celebrate loads.
1-0 straight from the whistle. What was it, 38 seconds or something?
They've shown your goal thousands of times all over the place
so I've even memorised the second you scored.
Valery asks: "How did you feel after scoring that first goal against Chelsea?"
- It was... - Unexpected!
No, it was... an amazing feeling to score straight away
with less than 40 seconds of the clock.
- 38. - 38, exactly. Less than 40.
It boosted our confidence and for me personally it was an unforgettable goal.
Anico asks: "What's Stephan's most annoying habit?"
is when we have an hour or two to rest
and you want to play snooker or have the TV on with the volume whacked right up.
I'm there trying to get some rest so I go: "Ste, turn it down."
And 15 or 20 mins later you actually turn it down.
Anico asks you the same question. What's my most annoying habit?
Your worst habit is that as soon as you come into the room
you lie down and want everything switched off.
Lights off...
You have to rest.
You always want to go straight to sleep!
Whereas you want the volume up on full for half an hour!
10 or 15 minutes yeah, then we can rest afterwards. But you don't.
Nicolas says: "Did you practise any other sports when you were a kid?"
Yes, I did. I did swimming for a couple of years, from the age of six to eight.
Why are you laughing?
What's up?
You just make me laugh.
I can't help it.
Question five is from Lenny.
What did you think of the fans at the Olimpico against Chelsea
The support was fantastic.
They were behind us for the entire 90 minutes and never stopped.
It's always great to see the stadium full in Rome so it really was amazing.
Especially when you score.
- They shout your name. - When they shout your name. That's a real rush.
- A real thrill. - Yeah, a real thrill.
Lollo asks you: "What's the atmosphere like in the Roma dressing room?"
It's great. Very chilled.
It's full of good lads and we have a laugh too.
- Yep, I can second that. -You know that yourself.
El Shaarawy's the only problem. Otherwise it's great.
Michael says: We've seen a new gesture when you celebrate. Can you explain that one?
Right, so my new gesture...
This one, right?
I did it for my best friend.
- Whose name is? - Manuel. He's in America now.
We came up with it last summer.
He said to me, "When I can't come to the stadium cos I'm in America,
would you celebrate like that for me if you score?"
I promised him I would so that's what I do after every goal.
- It's for him. - Nice.
Roberta wants to know
what's your favourite film or TV series?
My favourite film, well it's my favourite genre, but this film in particular, is Law Abiding Citizen.
It's brilliant.
As for TV series, at the moment I'm watching Narcos,
which is nice but I don't have a favourite TV series.
This is from Giulia. Can you cook?
If so, which is your speciality? Chicory?
- I'm sorry Giulia but I can't cook. - Yep, I can confirm that...
At best I could rustle up some plain pasta.
Me too. That and chicken breast.
Yeah, I can do chicken breast in the pan. You just turn it over.
Valeria wants to know what football skill of Lorenzo's you'd like to take off him?
Get the list out your pocket.
What would I take off you?
The number of years left in your career, since you're four years younger than me.
- You can't say that! - Why not?
Because it's impossible.
Come on Zio, that doesn't count. You have to say something else.
I'd like to go back four years.
- But that's impossible. - Yeah but it's ok.
- Something that by training... - No, I've already answered.
If you could be a video game character, who would you be?
I've always been crazy about Batman.
He's a video game character.
Batman is cool.
He's got that armour and an amazing car.
All black. It's cool.
I love Batman.
This is for both of us from Gabri, who's got quite an imagination.
What does he say?
What would you do if a seagull flew into your house?
What would I do if a seagull came into my house?!
A bat came into my house once when I was in Milan.
My brother and I were there with the towels trying to shoo it away.
Trying to kill it?
Because it kept going around in circles and we didn't know what to do.
- Did you open the windows? - They were open but it kept going round the room.
You could have just let it be.
What about you?
- Seagulls are big, though. - Yeah, they are.
It's not a bat; it's a big thing.
What would you do?
I'd open the windows and then leave the house.
Go for a walk and hope it's gone when I come back.
Maybe he just popped in for a look.
Ok, thanks Zio.
See you all soon.
A big hello to all our fans.
- Ciao. - Ciao.
Shooting – you can't shoot with your laces.
You got lucky with that goal against Chelsea.
Did I laugh too much?
-------------------------------------------
This is what happens when you do EDG in 2 minutes! - Duration: 3:59.-------------------------------------------
YOUTUBE IFŞALANIYOR DİSS TRACK F.T / - Duration: 2:12.-------------------------------------------
Best Gym Hip Hop Workout Music 2018 - Svet Fit Music - Duration: 37:21.Svet Fit Music
-------------------------------------------
YENİ AKIM YEŞİL UZAYLI (Dame Tu Cosita) 👽 - Duration: 3:51.-------------------------------------------
Optimization Tricks: momentum, adaptive methods, batch-norm, and more - Duration: 10:16.Deep learning is closely related to mathematical optimization.
What people usually mean by optimization is to find a set of parameters that minimize
or maximize a function.
In the context of neural networks, this usually means minimizing a cost function by iteratively
tuning the trainable parameters.
Perhaps the biggest difference between pure mathematical optimization and optimization
in deep learning is that in deep learning we do not optimize for maximum performance
directly.
Instead, we use an easier to optimize cost function on a training set and hope that minimizing
that would improve the performance on a separate test set.
We already talked about how optimization works in its simplest form in the earlier videos.
We pick a loss function that we want to minimize, do a forward pass given a mini-batch of inputs,
take the derivative of the loss function with respect to the weights, update the weights,
and iterate.
This is how vanilla stochastic gradient descent works.
There are some tricks that we can make this process more efficient.
First, let's focus on the update rule we had.
This update rule tells us to take a step towards a lower loss value guided by the gradient.
This usually works fine but can be a little slow to converge.
One useful trick that we can do is to add a velocity term to this update rule, which
helps our updates gain momentum towards minima at every update.
This is called the Momentum algorithm.
The momentum algorithm is simple and straightforward.
We define a parameter that accumulates a weighted average of the past gradients.
Then use this parameter in our update rule.
In other words, we add a weighted average of past updates to our current update.
The value of the momentum parameter determines the weight of the past updates in this weighted
average.
As a result of accumulating the gradients, the updates get larger when the algorithm
repeatedly gets gradients having a similar direction.
You can imagine the current state of the weights as a ball moving down on a surface defined
by the cost function.
The ball gains momentum as it moves towards the basin even if it hits some small pits
on the way.
This usually gives a quicker path to a solution as compared to plain stochastic gradient descent.
As you may recall, another factor that determines how big the step size will be is the learning
rate.
It's usually a good strategy to start with bigger steps and decrease the step size as
we get closer to our target.
You might think the magnitude of the gradient should shrink over time anyway, but that doesn't
happen in many cases.
The norm of the gradient might even increase, while the loss still keeps decreasing.
So, it's common to set up a schedule, such as exponential decay, to decrease learning
rate over time either in a continuous way or by taking discrete steps.
In the plain version of stochastic gradient descent, the choice of learning rate might
have a crucial impact on the performance.
There are several methods that set a separate learning rate for each trainable parameter
and adaptively adjust the learning rate to decrease a model's sensitivity to the initial
learning rate.
AdaGrad algorithm decreases the learning rate faster for the parameters that have large
gradient components and slower for the ones that have a smaller gradient.
RMSProp algorithm also adaptively tunes the learning rate for each parameter in a similar
way but uses a moving average of gradients to make the optimization more suitable for
optimizing non-convex cost functions.
Another algorithm that uses adaptive learning rates is the Adam optimizer.
Adam stands for adaptive moment estimation.
It attempts to combine the best parts of RMSProp and momentum optimizers.
In practice, Adam and RMSProp both work well.
In addition to the optimization algorithm, the model architecture also has a big impact
on how easy it is to optimize a model.
Many successful models owe their performance to their architecture rather than the choice
of the optimization algorithm.
You can check out my earlier video on designing neural networks to learn more about how model
architecture can facilitate the optimization procedure.
You can find it in the Deep Learning Crash Course playlist in the description below.
One challenge in optimizing deep models is the internal covariate shift problem.
When we update the weights in one layer in a deep neural network, we update them assuming
that its inputs would stay the same.
However, the distribution of the inputs might change every time we update the weights as
the previous layer parameters are updated.
In deep models, even small changes in the early layers get amplified through the network
and cause a shift in the distributions of the later layers.
Changes in the input distributions make it harder for the following layers to adapt.
This problem is called the internal covariate shift problem.
A technique called batch-normalization makes it easier to optimize deep models by normalizing
the outputs of the hidden nodes right before they are fed into an activation function.
The first step of batch normalization is to subtract the batch mean from every output
value and divide it by the batch standard deviation.
This gives us a zero-mean unit variance distribution at the output.
The second step uses scaling and shifting parameters to allow the variables to have
any mean and standard deviation.
These scaling and shifting parameters are trainable and learned during training.
Essentially, the second step can undo what the first step does.
You might ask what's the point of normalization then?
The answer is that in practice the second step doesn't really undo the first one.
It's true that the variables are allowed to have an arbitrary mean and standard deviation
both with and without batch normalization.
The difference is that when batch normalization is not used, the distributions are determined
by a cascade of parameters.
On the other hand, batch normalization parametrizes the mean and standard deviation as trainable
parameters, which makes the distribution shifts manageable during training, resulting in faster
convergence.
Once the training is complete, global statistics that are computed during training are used
to normalize the activations rather than the batch statistics.
In this way, the inference becomes independent of the input batch during test time and we
don't need a batch of samples to run inference on a single sample.
Optimizing deep models involves iterative processes that require some sort of parameter
initialization.
The way we initialize the parameters can have a big impact on a solution that a learning
algorithm achieves.
For example, if this is how the cost function looks like, projected into a single dimension,
and the weights are initialized on the wrong side of a hill like this then the model would
converge to a local minimum although there are much better solutions on the other side
of the hill.
In practice, though, cost functions like this are very rare.
In higher dimensional space, local minima are not very common and it's likely that there
is a way around hills like these.
Why are local minima rare?
Think of it this way: for a point to be a local minimum it needs to have a smaller value
than its neighbors in all axes.
If we have a single dimension, the odds of observing such structures is not very low.
But what if we have a million parameters.
Then, for a point to be a local minimum it needs to have a smaller value than all of
its neighbors in all one million directions.
How likely is that?
If it does happen it's likely that it already has a very small value that can be considered
an acceptable solution.
In deep learning, we usually care about finding a good solution rather than finding the global
minimum.
Another type of critical point is a saddle point, where the cost function gets a minimum
value in some directions and a maximum value in some other directions.
Saddle points are likelier to be observed than observing local minima since it's harder
to get a value that is smaller than its neighbors in all directions as compared to a being a
minimum only across some directions.
If the norm of the gradient becomes very small during training the problem is likelier to
be a saddle point than being a local minimum.
Let's go back to weight initialization.
The initial state of a neural network can have an effect on how fast the model converges,
how good the converged point is, or if the model converges at all.
So how should we choose the initial values?
There's no definite answer but there are some heuristics that might help.
Initializing the biases is usually easier.
It's usually safe to initialize them to zero or a small number.
However, initializing the rest of the parameters to zero or another constant is not a good
idea.
If we initialize all parameters to the same value, then they will all get the same updates
during training and end up learning the same features.
Ideally, we would prefer each neuron to learn something different to be useful.
To do that, we need to initialize the weights in a way that breaks the symmetry so that
each neuron gets a different update during training.
Initializing the weights randomly usually works fine although it doesn't guarantee an
absolute asymmetry.
A very common initialization algorithm is the Glorot initializer, also known as the
Xavier initializer.
Glorot initializer randomly samples the weights from a uniform distribution where the range
is determined by the number of inputs and outputs to keep the activation and gradient
variance under control.
This is the default initialization method in some frameworks.
The scale of the initial values matters because if they are picked from a very narrow range
of small numbers then the units might not be different enough to learn different features.
If the scale is too big, then the gradients might grow exponentially as the weights get
multiplied from one layer to another.
This problem is called exploding gradients and is more prevalent in recurrent neural
networks than the types of neural networks that we have discussed so far.
So far, we have focused only on feedforward neural networks.
In the next video, we will discuss what recurrent neural networks are and how they work.
That's all for today.
Thanks for watching, stay tuned, and see you next time.
-------------------------------------------
Oil Pastel Painting For Absolute Beginners By my Son Hrithik - Duration: 1:25:25.-------------------------------------------
J-PLA는 국내 최초의 뮤지션으로 YouTube World Music Chart에서 톱 30에 진입했습니다.|조회수8.212.910 - Duration: 2:34.-------------------------------------------
L25:Compiler Design Tutorial, Bottom Up Parser,LR(0) Parser, SLR(1) Parser,Simple LR Parser In Hindi - Duration: 26:57.
www.universityacademy.in
www.universityacademy.in
www.universityacademy.in
www.universityacademy.in
www.universityacademy.in
www.universityacademy.in
-------------------------------------------
How To Post Youtube Video On Facebook with Large Thumbnail [Bangla] | - Duration: 2:36.
How To Share Youtube Video Link In Large Thumbnail On Facebook
post fb
large thumbnail
Lets go
open your videos
-------------------------------------------
Dupion Silk Sarees from Flipkart - Duration: 3:24.-------------------------------------------
HITTING A TRICKSHOT I WOULD NEVER HIT AGAIN - Duration: 1:15.Right here
Cuz you have the flag Oh in it I
Sort of go out that was another no scope for a quick scope I won't be so bad
Duck I actually hit this time oh
Shit yeah
Yeah, don't judge me hmm. I
Thought for sure you gonna get like just uh no scope I was about to be like I was doing
Yo, so uh game reviews I will be streaming tomorrow around maybe a little bit earlier two hours earlier
But from now, so if you just hop In the stream. I'll add you tomorrow
Yeah, I'm not adding you I'm sorry
I'm not gonna. Add I was gonna play with it. We could play her. Yeah, sorry
-------------------------------------------
China's Freedom-Crushing "Social Credit Score" - Duration: 4:52.Is your privacy under threat?
Has the age of privacy come to an end?
Are you sick of Facebook's
lack of respect
for your privacy?
If you think
your privacy's threatened,
be glad you don't live in China.
Chinese government wants to give
every citizen a score
based on behavior.
Purchase history,
political leanings,
and social interactions would be used
to calculate a person's trust score.
Facebook and Twitter
are banned in China,
so people use
So the government spies on that
round-the-clock.
The state also monitors
the Chinese version of Amazon,
called Alibaba.
Alibaba is by your side
24 hours a day,
seven days a week.
Why should we care what the Communists do?
We're not in China.
Li Schoolland came to America
30 years ago.
Here's Li when she was
16.
She survived the
Great Leap Forward,
the Great Famine,
the Cultural Revolution.
Her parents were doctors so
they
and she were
re-educated.
And this was to teach you
not to be fancy?
The repression is over.
It's all better now.
Today in China,
if you tell friends about certain books,
your message will be blocked.
Even innocent sounding phrases
are censored.
So I understand
the titles
of novels like Animal Farm,
Brave New World,
but Long Live the Emperor?
Well he's President forever.
They can't even talk about this teddy bear.
Winnie the Pooh?
And now, another step
more subtle than just
banning things.
The state will monitor what you say
in social media
and assign you
a social credit score.
That will tell them how trustworthy you are.
The government says
this will allow the trustworthy to roam
everywhere under heaven,
while making it hard for the discredited
to take a single step.
There's gonna be this new
social credit score.
Some American governments
already do something
similar.
The LAPD can scan tens of thousands of
license plates.
Los Angeles police now practice
predictive policing.
They pay a company called Palantir
to analyze social media,
trace people's ties to gang members,
and predict the likelihood that
someone may commit a crime.
After searching over a hundred million data points,
Palantir displayed an impressive
web of information
on one burglary suspect.
People like that.
They think it makes them safer.
I would like to know
that there's a trust score so
I can know who's trustworthy
and who's not.
Sounds sort of appealing.
When government does gets involved,
bad things can happen.
What happens if you have a low score?
If they really don't like what you say,
they lock you up
and torture you.
They didn't allow me to sleep.
I was kept in a small room
and saw no daylight for half a year.
But that's China.
Why should
we be afraid?
Get out of my life!
In America, every week
on YouTube,
Twitter,
Facebook,
I challenge people in power.
Trump does make things up.
I say these things and
no one punishes me.
So far.
Không có nhận xét nào:
Đăng nhận xét