Is it too cynical of me to say that people will happily accept a fake image or video if they want it to be real?

@howtoappearincompletely9739

2 ай бұрын

Not at all. That's a truism.

@danielduncan6806

2 ай бұрын

No. It is an accurate statement. Fact of the matter is, none of this is real to begin with. It is all about perception. There is no evidence anyone can present that definitively proves all of this is real. Even if you are standing in front of me, and I reach out and grab you with my hand; that is still not definitive proof you exist. Because I am still perceiving you with my brain's senses. And I cannot rely on YOUR perception, because your perception is just as unreliable as my own, as is anyone's. So yeah, your reality is whatever you want it to be; if you want to perceive a "deep fake" as reality, well, then that is your reality. "deep fakes" are as real as any other. If this frightens you... "The only serious question in life is whether to kill yourself or not." - Albert Camus That is _probably_ the only thing that is real. Do you stay on the ride all the way to the end, or do you exit early.

@piemaster310

2 ай бұрын

I would say yes, because "too cynical" to me means that you're sacrificing nuance for simplicity

@Monkey-fv2km

2 ай бұрын

That's exactly the right amount of cynical.

@mobilephil244

2 ай бұрын

Nope it isn't.

Everytime some new internet/tech security comes out, it's like someone sent yet another "unpickable lock" to lockpickinglawyer.

@FourthRoot

2 ай бұрын

Not even close. TONS of internet data is secured with open cryptography standards that have existed for decades. Emails are signed this way, for example. As do cryptocurrencies, any website with https as the prefix, etc. etc.

@cameron7374

2 ай бұрын

Ehhh, it depends. There is actual, proper good security which, if implemented right, is unbreakable without either ridiculous amounts of time on a supercomputer or revolutionizing mathematics. And then there's "Methinks we should just train an AI to detect the AIs", based on no good proofs apart from "it seems to kinda work."

@FourthRoot

2 ай бұрын

@cameron7374 Exactly. By the way, the way these deepfake programs work is that they train two AIs, one to generate fake images, another to detect fake images, then pit them against eachother and iterate until they both stop improving.

4:26 Sabine hairstyle 4:44 slightly different Sabine hairstyle We need those credentials now!

@lz43p15

2 ай бұрын

Yes, different hairstyle and for more of one month, every day the same shirt. Is Sabina a fake?

@oskarskalski2982

2 ай бұрын

​@@lz43p15 is always thought that he German-style corny jokes are indication that she is a robot;).

@red.aries1444

2 ай бұрын

@@lz43p15Her shirt has more or less become her trademark, she may have several identical ones. 🙂 But on the other hand she now only does short videos and I think she records all the videos for one week in a row, so she has enough time to clean her shirt in the meantime.

@soerren5393

2 ай бұрын

@@red.aries1444 we'd call her Ein Emsiges Bi(e)nchen

@lz43p15

2 ай бұрын

@@red.aries1444 It may be, but it's not very convincing. It's a mystery like dark matter.

"It's trust all the way down" - this is how SSL certificates for websites already work. With "content", though, I think certificate revocation could get problematic, Orwellian. Imagine an authority invalidating a war documentary from the root certificate, and public libraries being forbidden to carry "fake content". We saw versions of this behavior from 2020-...

@darrennew8211

2 ай бұрын

The problem is that the SSL only tells you that someone with the key signed the content. It's currently just linking the DNS record to the web site response. It keeps you from claiming you are delivering that web site when it belongs to someone else, regardless of content. It isn't possible to algorithmically prove that any given content hasn't been manipulated somehow. The best you could do is what SSL does now, which is "CNN claims this isn't a deepfake," but we already do that. Putting data in an image that says it *is* a deepfake won't work at all, because that can just be removed.

@LettersAndNumbers300

2 ай бұрын

It just validates source, not truth.

@darrennew8211

2 ай бұрын

@@LettersAndNumbers300 It doesn't even do that.

@bmatt2626

2 ай бұрын

@LettersAndNumbers300 I see what you're saying, but join any political forum and count how many people currently make that distinction. When Grandma's free antivirus pops up a warning about your video she's not going to go sifting through the certificate chain for clues, she'll just watch CNN's version. It concentrates power at the CA.

@Marque734

2 ай бұрын

The processes of revocation (or maybe leaving it intact but adding a remark) should and can be transparent. Invalidating also doesn't need to mean that it is forbidden, just that viewers show a proper remark that it can't establish trust. Just as your browser does on http websites.

Coevolution is taking place between KZread and AdBlock right now, and we are the battlefield it's playing out on 😂

@Thomas-gk42

2 ай бұрын

That's right, the plug in that shows me the dislikes on YT works and works not at the same time 😂

@AndrejMejac

2 ай бұрын

Evolution has already brought us uBlock Origin.

@thesenamesaretaken

2 ай бұрын

​@@Thomas-gk42Ah, I see they found a use for quantum computers then

@Thomas-gk42

2 ай бұрын

@@thesenamesaretaken 💯

@hyouki8529

2 ай бұрын

I thought youtube already gave up, haven't seen a disable adblock message in ages

"It would be credentials all the way down." Perfect.

@emifro

2 ай бұрын

Just like TLS

As a native Floridian, that dog riding the shark pun was NEXT LEVEL 😂😂😂 I had to pause the video to catch my breath 🫁🎉

@Thomas-gk42

2 ай бұрын

She´s a diamond

Someday, perhaps when its too late, all humans will hopefully realize that ethical thinking and ethical actions are the only real solutions to such problems.

@cameron7374

2 ай бұрын

And what do you do when someone isn't being ethical?

@sukhmandersingh4306

2 ай бұрын

​@@cameron7374 Teach them, quarantine them, fine them or jail them. Ethics will be the new law but i dont know if we can come up with ethics when there are so many unethical people who have something at stake on this world

@paihobbes8680

2 ай бұрын

And what about those with Flexible-morals and Ethics, posing as though they actually have proper Morals and ethics (virtue signalling and gaslighting) whom have more power (greater reach) begin to lock up people who have counter morals/ethics/values for not having the same?? Remind you of anything that happened recently?? See, morals and ethics are relatively useless and equally dangerous without critical thinking and deductive reasoning. And conversely, analysis is devoid of compassion without Morals and Ethics. Balance is required

@cottawalla

2 ай бұрын

Like all other "all humans" projects up until now, that would just become another tool for manipulation.

I loved the outtake at the end. 😀 Re the actual topic, it seems inevitable that the noise will drown out the signal before too long.

@DrDeuteron

2 ай бұрын

What signal?

@JonathanMaddox

2 ай бұрын

Ma Ma Ma Maaax Headroom.

@PSnaptic

2 ай бұрын

This problem already exists. Content credentials is a way to minimize noise by allowing people to easily identify non trustworthy sources. It's one tool. It won't make people think... But it will allow those who do to better inform those who don't.

@2ndfloorsongs

2 ай бұрын

If people think it's all noise, they'll think everything's fake and fakes won't have any affect anymore. It's already the case that nude pictures of celebrities, even if they're real, are assumed fake. Paparazzi and blackmailers are out of luck. Most people already assume that anything a politician says is a lie, so deep fakes won't have any effect there. And trusted news stories are believed to be fake by at least half the population whether it be left or right. Bottom line, I think, is that fiction overcame reality a long time ago and deep fakes won't change much

By the way big thanks to you sabine. You are one of my official news channel nowadays, it is so nice to actually have daily updates on things that really matter for humanity.

This is not a deep fake. - René Magritte.

@likebot.

2 ай бұрын

avant ou après la guerre ?

So images will be digitally signed? Maybe browsers and user agents could enforce policies around multimedia, as they already do with domain certificates, CORS, CSRF and the like. That way, a popup or tooltip could display over an image if there's no verified source, if the source is not related to the site it's displayed in, etc.

I imagine it will be quite expensive to have authenticity in this brave new world of ours...

@johnwang9914

2 ай бұрын

Well, it's a matter of trust, the better known content providers would voluntarily sign their images with redundant encrypted watermarks so you can identify what content are from the well known providers, won't stop Fox News though. Then there's having publicly available image modification software also digitally stamping their images so that casual deep fakes could be easily identified and traced perhaps image viewers could even display that a deep fakes is suspected on a banner. You can't guarantee that all images would have the appropriate watermark stamps but at least you have a way of identifying which content is from a provider you trust and which was made with commonly available deep fake software or image manipulating software. It's all software so the only costs are development and certificate registration costs.

@wanderingquestions7501

2 ай бұрын

Yes. It is ultimately about money

@AttilaAsztalos

2 ай бұрын

...and also quite difficult whenever we're talking about some kind of event captured by a random member of the general public who just happened to be there at the time - now did those cops really shoot that guy in cold blood, or is it just a deepfake...?

@danielduncan6806

2 ай бұрын

The real question is, was ANY of it ever authentic to begin with? Are the signals your brain is sending to your consciousness actually real?

@alieninmybeverage

2 ай бұрын

Wym. It's only 5 dollars for a ✅️

Ministry of Truth is coming...

@willyburger

2 ай бұрын

BBC, CBC and NYT. Perfect candidates for that.

@07Flash11MRC

2 ай бұрын

Well, then not much will change...

@zasteray73

2 ай бұрын

Our AI overlords will tell us what to think about that🙄

@stewartsiu1960

2 ай бұрын

@@OOL-UV2 Bias doesn't have to be zero for there to be protection.

@deth3021

2 ай бұрын

​@stewartsiu1960 it does when the opossing opinions are illegal. Otherwise it is just propaganda.

I love your deadpan humour.😁 Makes every video that much better than just a science article, and much more memorable because of it.

Ha, credentials all the way down! That's just gold.

Ooh, an outtake at the end! I love those 😮😍

Very special video today thank you.

I love your sense of humor. That dog riding a shark joke made my day.

This seems like it will just be opted out of if the certificates represent any real increase in the size of image uploads. I think there's also a fair amount of discussion that needs to happen around "what does editing an image entail?". Making a resized thumbnail certainly is editing an image and tuning colors in an image is certainly editing it as well, and there's a whole bunch of issues about trying to draw the line between an inauthentic image between "resizing and retouching are okay" and "materially changing an image is not okay" because they both represent an inauthentic image.

@dmitripogosian5084

2 ай бұрын

But what is authentic image to start with ? Even photo cameras already do processing, and always did starting from the first one.

@nemanjatrivic9505

2 ай бұрын

I guess cryptographic signature can store information such as: this image was resized by x:y ratio. Then the viewing software like browser can verify if this edit check out.

@evvveeeeeeee

2 ай бұрын

We don't need this discussion. It is not philosophical in nature. The problem is people release fake imagery, not edited. Meaning that the intent was to use this tech to fool people, not improve image quality. The solution is simple - AI has the potential to be used as an informational weapon. Investigate claims of deepfake, decide the intent and then provide punishment to those whom used it like swatting and revenge porn. Impose a prison sentence and 5 years post-prison probation where access to tech is violation of probation - this is akin to computer crimes

@nieznanysprawiciel

2 ай бұрын

Yes, but the problem is that normal editting process includes much more changes like color balance adjustments etc. We don't see uneditted photos normally. And you can't just store history of changes, because algorithms used by editting software aren't public (and pixel perfect precision is necessary for cryptographic signatures verification)

@darrennew8211

2 ай бұрын

Also, any certification that gets added in later to say "this image is bad" can be removed as easily as it's added. What you would need to make this work is a certification that the image is authentic.

perfect timing, a nice video along with a nice meal.

@FiNiTe_weeb

2 ай бұрын

same lol

One option people often seem to forget is: ignore it. It feeds on your attention.

@dykam

2 ай бұрын

That's something you can do, but it is a useless thing when it comes to dealing with the problem in society.

@CrniWuk

2 ай бұрын

You can do that and if it works for you great. But as a social problem, that won't work. Fakes already influence society on certain levels. Like imagine a mob forming due to some fake content and they start to riot burning down your business in the process. How do you ignore that? You can't. In other words, the damage still hits us all what ever if we ignore the issue or not.

@sdfsfmnsdkfsfdsfsldmfl

2 ай бұрын

I think you underestimate the consequences. You literally can't trust everything you see and read unless you directly see it

@deltaxcd

2 ай бұрын

@@sdfsfmnsdkfsfdsfsldmfl I woud say that you should trust people (sometimes) not evidence if somebody provided image to support their claims then treat that image as illustration for their claims. As i I say that I found a bigfoot in the forest and provide a photo then treat that photo just as illustration what I saw not as evidence that I did saw it.

Very informative, well done, Sabina. Thanks. Peace 😎 ✌️. Perception

Except most social platforms strip out metadata from images on upload, making this approach completely useless in practice.

@kazioo2

2 ай бұрын

Until US gov and EU enforce legally every social media to use it. Add device ID registration and voila - you can now track every citizen easily for posting a photo online.

@sdwone

2 ай бұрын

@@kazioo2 Well... That's usually how it goes right? They give us Freedoms... Which then gets abused by a handful of unsavoury individuals... Which then means that our toys either get confiscated altogether! Or we get supervised when playing with them! We will ALWAYS suffer due to the actions of a few! That's Freedom 1-0-1! We abuse it... Then they take it away!

@pafnutiytheartist

2 ай бұрын

@@kazioo2 Possible, but also not good

_Which sources are we to regard as trustworthy and why? And how are we supposed to check trustworthiness?_ The technical answer to this has already been given by SSL & TLS. With certificates and certification authorities. Which can certify the identity of the original creator, if the creator so wishes. But whether you'll trust this creator will remain at your discretion anyway.

@swancrunch

2 ай бұрын

So if the creator is CNN and it created lies using deepfake, you still have no way to discern the reality of the video. Welp, welcome to the dark age, where no proof exists and ad humanem is the only arguement.

I swear as far as my favorite physics and science/tech personalities go, Sabine just keeps climbing the charts! Such a nice diversity of subject matter. In fact if you listen to Penrose go on about his greatest breakthrough theories they always seem to include some tangential link to for example neuroscience. A great physicist I think spends quite a bit of time outside the privacy gate of their field, wandering about before returning. Everything is physics after all.

You speak the truth Sabine

Who's watching the watchmen, that's the point.

Upgrade your deepfakes to shallowfakes. 👍

@ThatOpalGuy

2 ай бұрын

just reachable without scuba fakes

@charlesbrightman4237

2 ай бұрын

'Free Speech': Range, where is the line? Absolutely nothing allowed -> Line -> Absolutely everything allowed. And who decides? And by what authority do they and they alone get to decide? And what about absolute truth reality of what is really true? "IF" people (or AI's) are speaking the truth, then they are speaking the truth. And 'if' truth is censored, then society is living a life in lies.

@alieninmybeverage

2 ай бұрын

​@@charlesbrightman4237 I think it's fair to say there is no good solution, and perhaps even that there are only "less-bad" solutions given our collective lack of foresight and/or discretion. But also, that we only have such options makes it all the more important to choose the less bad. The worst options are quite sinister and difficult to fix once taken.

@charlesbrightman4237

2 ай бұрын

@@alieninmybeverage Agreed. But who and/or what decides what is censored and possibly even punished? And possibly even punished even though the US Constitution, the highest law here in America, says it's citizens have 'free speech'? For example also, what about the real apparent truth? a. Many people believe that 'God' truly exists, even though there are various versions of God and various versions of whom supposedly is God's chosen. But yet the analysis seems to indicate that God does not even actually exist except for as a concept alone. So, who gets censored and who is allowed 'free speech' and who is allowed to even speak the real apparent truth to society? b. Many people also believe that the singular big bang theory is really true, even though the analysis indicates that it is not really true. So, who gets censored and who is allowed 'free speech' and who is allowed to even speak the real apparent truth to society? c. What about 'climate change' and what the apparent real truth is concerning how the climate might actually be changing? Just a certain narrative allowed to be spoken to society or the real apparent truth spoken to society? d. And even what modern science claims about certain items. How could modern science even progress if certain ideas are not challenged? Let the real apparent truth fall where it may?

@alicemiller8031

2 ай бұрын

Is the faa responsible for stsrship's communication loss? Faa delays have made elon want to cram more missions into one flight. It's said the payload door test did not close. That could account for the communication loss during reentry: hardware failure from heat recirculation zones Space flight supporters should not vote for Biden

Great video, loved the humour

So basically, in few years we will be back to pre-internet era. Or wait pre-television era.

5:30 The Elon Musk shade is real 🤣!

2:48 Encrypted or signed?

@diamondq

2 ай бұрын

It's signed, although I can forgive the slip since signing a document is just encrypting a hash with a private key.

Thanks.❤

Content origin tracking built into the file metadata is long overdue imo, but it wont actually fix the issues of discerning truths from fiction. It will however help determine source trustworthiness like SSL does for web domains, sure it can be manipulated, but its better than no verification at all

TY Sabine. An excellent dive into "is this real?" That is the first fundamental question the physicists ask. It is a meta-question. And in science it is a very valuable question because it is followed with, "how would I know?" Not all that long ago the authenticity of a signature on a check was known to experts who relied on several technics comparing a questioned signature with known true signature. Unfortunately, even the fake signatures today are good enough to deceive even an expert in fake signatures. There is no easy solution. It all goes back to the authenticity of the verifier and can it be faked.

@cameron7374

2 ай бұрын

I mean, digital signatures can be un-fakeable, as long as the person doing the signing doesn't give out their secret key.

@user-wg2vw3mz1v

2 ай бұрын

Maybe we can finally move away from idiotic "representative" democracy where we vote for which dumb-looking buffoon will screw us over to direct democracy where we vote for the laws, regulations, and policies are best for us as a whole.

Perhaps people will begin abandoning the Internet except for basic services like shopping and email. Do we really need Twitter anyway?

@Nordlicht05

2 ай бұрын

I also thought about it what will happen if many bail out. But we will see how bad the bads get 😅

@cameron7374

2 ай бұрын

But the problem doesn't just stop outside of the internet. Lets say you want to vote. To make a good decision, you should have at least some idea of who the candidates/parties are and what they stand for. You can't just go and meet all of them in person all the time, so you'll have to rely on and trust some form of second hand (mis-)information. A while ago, a video of a politician giving a speak pretty much meant that that person actually said those things but that is slowly falling apart and it can affect TV broadcasts and newspapers just as much as the internet.

@Nordlicht05

2 ай бұрын

@@cameron7374 yep that counts too. But newspaper and classic television especially are for most people now obsolete and is dying. It is possible that the internet will change drastically in how it is used if this gets out of hand. People now may say we were lied before everything became digital but that's a new quality.

@heckinbasedandinkpilledoct7459

2 ай бұрын

I’ll just do all my serious research at the library since internet users by and large don’t care about truth

"..just a normal day in Florida." Love it! Love your videos. Thank you for all your hard work.

@Thomas-gk42

2 ай бұрын

😊😂

Great video Sabine! It's an important issue, nevertheless it made me laugh few times. BTW are you performing some kind of Change Blindness Experiment on us?🤔

This strongly feels like another big at corporate control of information. If this was simply a method to claim ownership of data, then it could potentially be used to work towards recognizing sources more effectively.. but I really doubt it isn't going to come with fees, which means it's just a way to silence those without enough money. My biggest fear with this is that it potentially finishes eliminating the gold standard of photographic proof. Who cares if you get photographed doing something? It's fake because they don't have the credentials.

@dmitripogosian5084

2 ай бұрын

Well, providing verified information (or at least signed) probably has some monetary value

@MRL8770

2 ай бұрын

It's probably going to end up with some closed technology like a certified chip that you can embed in a device such as camera and that chip is going to sign the image with a registered key that's fused in it. If the camera is determined to not have been modified, the signed photo is valid.

@aaronsj80

2 ай бұрын

The idea has been around a while. Check out C2PA. Modern processors are already being built and sold with the ability to sign files with what computer created them. AI images are just a convenient way to make the idea sound good. Potentially, this could be used by modern operating systems to blacklist any content created by certain computers, even if the file has already been on your hard drive. It could just get an updated blacklist and poof, no more file that the powers that be don't want you to have. Whether that's the way they are going to go with it or not, who knows?

😂😂 maybe Trump and Biden dancing together isn't a deep fake!!!!! 😂😂❤

@ThatOpalGuy

2 ай бұрын

biden would be the lead in a real photo.

For video, you could always integrate GPS, time, and accelerometer data into the video, which would disprove many types of fake videos, like fake badgecams.

So, in the end, it's not about telling if an image/video is deep-faked, but to tell if it's genuine. This means that we have to introduce this signing pretty much everywhere: in camera, scanners, phones... anything that can take an image should be able to say something: "indeed, this image was taken by me" ("me" as a device). But then... nothing stops me to create a deepfake, put it on the screen (or print it) and take a picture. Also we have to rely on the certificates chain of trust, what we have now in any device (or create a parallel one?). Then we have to include in the signature a hash of the image itself so that can't be tampered or the signature applied to a deep-fake. It's a horror.

Don´t do that, you look much better than K.R.! Interesting information and wonderful entertainment on a nevertheless serious problem.

Sounds like a last grab at relevance from a bunch of media outlets that nobody trusts and are dying out to competition in the online sphere.

@kf5338

2 ай бұрын

Yep. There are some countries/warzones where there are only independent journalists. This will give politicians a way to dismiss such reporting - they will just say "it's not verified, it can't be true". I hate this world.

@FourthRoot

2 ай бұрын

Creating an open cryptography standard to allow people to prove their videos are real is a grab for relevance? How?

@id01_01

2 ай бұрын

​@@FourthRootIt doesn't prove it's real, just "made by this person"

@U20E0

2 ай бұрын

@@id01_01 true. The actual thing this does is make it possible to track who made any "real" image.

@kovatembel

2 ай бұрын

I understand some political tendencies are probably alienated by media and don't trust them. But considering the future where anyone at home can fake any story, video, etc. , to the point that anything that is seen through a screen is useless, is that alternative better ?

Sounds like a great idea.

Every quantum of friction ... lovely

Most of the bizarre AI images still aren't as bizarre as the real photo Leonid Brezhnev kissing Erich Honecker.

Oh good. More push for "authoritative sources" from the corporations we love. There's always a good excuse to pull the ladder up a little higher.

@frankcl1

2 ай бұрын

The way I see it, it will just be something similar to the HTTPS certificates, where anyone can own one, not just corpos

@AAjax

2 ай бұрын

@@frankcl1 that's by the grace of non-profit organizations, not google. Thankfully administrating https is just as simple as plain http, and it doesn't require more overhead, right? Unfortunately no charitable orgs have stepped into code-signing certs, and google recently turned up the virus scanning in Chrome to 11, handing out false positives to open source projects like candy.

@Godfrey544

2 ай бұрын

@@frankcl1 is it finally time to bring out the cyberpubk slang?

Dr. Sabine this is a good topic to keep discuss, thanks for covering. So Credentials seems to address the authenticity of real media, but how about identify fake media. I was thinking governments could pass laws requiring any AI touched media be flag with its own "fake" credentials (for automated systems detection) but also AI generated media must be made obvious by the observer (warning labels visual displayed if "fake" credentials set, for example). Because we will have authenticated media, user / personal media (as it will be too expensive to get Cr credentials on all your personal media), and fake media, on the internet.

Sure, been pumping out a lot of videos

The web (https) already works on a “chain of trust” basis, so content credentials should work at least as well and may also help restore revenue streams for legitimate journalism.

@DensityMatrix1

2 ай бұрын

It’s better to be skeptical. I would rather distrust most images, than trust an image with a signature. Namely because nation-state level actors can compromise chains of trust.

@sdwone

2 ай бұрын

@@DensityMatrix1 Bloody heck!!! Distrusting most images?! That sounds like HARD bloody work! Could you imagine distrusting most of the Internet!? Well actually I do distrust most of the Internet! But that's not a problem, because I only checkout a SMALL number of sites on a regular basis anyway! And ignore the rest! Anyway, regardless it's coming! Whether we want it or not! And since we already have HTTPS, which has been successful, for the most part, then Hell! Why not do something similar for content!

@thesenamesaretaken

2 ай бұрын

The first half that comment seemed fair, but the second part made me wonder if it was a subtle joke.

@DensityMatrix1

2 ай бұрын

@@thesenamesaretaken CNN or Fox: we have this picture of Putin/Trump/Biden eating a baby, it’s legit the certificate signed by the NSA says so.

So, will it evolve into something like the Social Credit System they have in China? If you do not have a high enough score all your online communications and content will be discredited or outright ignored. I wonder how you can get your social credit score higher? and what would cause it to be downgraded. (Not really, I have a pretty good idea what you would have to do to have a high score. Pay your taxes on time, and not write anything negative about anyone that has a high credit score. Certainly, no negative posts about the platform you are using, the government, etc.) What a dystopian nightmare we could be plunging headlong into.

This could be a problem for open source image & video tools like GIMP, etc.

1:07 I have picture of Biden and Trump riding unicorn 🦄 on my phone 😂

Interesting. So basically, we will have a trust-based system like we now have with SSL certificates. Instead of questioning the content, we will be questioning the creator. And since SSL certificates are quite reliable, this should work fine, too.

@SabineHossenfelder

2 ай бұрын

It's very similar indeed, but I don't think there'll be anything similar to a "this connection is not safe" warning. Rather, any user will be required to make up their own mind about who to trust.

@jokerES2

2 ай бұрын

It's also notable that problems with certificate authorities are abundant and have been so far as long as they've been around. That's been the history of Internet Explorer and Chrome in a nutshell - a rotating set of CAs that slowly give in and eventually must be removed because bad actors take advantage of a lack of diligence.

@ungainlytitan1460

2 ай бұрын

I think it will be more like the rise of trusted publishers in the eighteenth century and yeah, you will have to pay for trustworthy news.

@foolmoron

2 ай бұрын

@@jokerES2 Yeah certificates are less of a solution and more of a system that needs to be constantly maintained, forever. But if like SSL it's relatively little effort to maintain, and (more importantly) zero effort on the user's part, then it can work well.

@kayakMike1000

2 ай бұрын

That is true, it's not precisely certificates, but digital signatures. Certificates create a chain of trust where some top level authority signs another certificate. If you trust the top level, then you trust the next link in the chain.

To bad we are losing trust in “State Media”

I worked creating gans for a period and thought about something like this cr initiative. It is a very needed solution for our times. They must have a signed hash version of the content and then be able to compare it by the device u are using. Similar to what your browser does for verifying websites authenticity..

HaHaHa! The part I laughed at the most was where she said you could maybe trust an image with Content Credentials from the BBC, of all places! You'd make a better stand-up comedian than a 'scientist,' Sabine!

ChatGPT, how do I defeat deepfake credentials?

Propaganda doesn't even need to use deep facts. Look how many people believe Trump said stuff the media said he said

@oskarskalski2982

2 ай бұрын

So what exactly was wrongfully reported about what he said?

@avsystem3142

2 ай бұрын

The media I consume doesn't just make claims about what Trump said, it shows the videos of him saying it. And they aren't deep fake.

@patrickr9416

2 ай бұрын

Yes, can you please give a list of the false quotes?

@osmosisjones4912

2 ай бұрын

@@oskarskalski2982 call Dead veterans su cars and losers. Telling Republicans to not solve boarder crisis so he can run on it . And the pea pea tape

@osmosisjones4912

2 ай бұрын

@@patrickr9416 calling fallen veterans su cars and loose . Telling Republicans not to solve the boarder crisis so he can run on it. The steal dossier the pea pea tape

It will ensure the source is who you think it is, its a good thing

What a talented Binchen - the end is just priceless - but still missing a natural vibrato. Did you not practice enough? But you could fake it. 🙂

@Thomas-gk42

2 ай бұрын

For a new music video?

@soerren5393

2 ай бұрын

@@Thomas-gk42 oh, yes - in the pasticcio Echo and Narcissus she would be able to sing all roles. From Hera to Charon. She's my tragic hero 🙂. I'd even pay the obol for the passage.

There are ways to embed information on an image starting at the camera and then all the alterations made in the image including users and what was done. A slight color or exposure correction isn't a lie depending on the image and what it's being used for but it would be nice to know. BTW, I have no idea of the ways that could be made but AFAIK there's watermarking techniques for digital files that are somewhat robust.

Meta data is really easy to strip from an image, sometimes accidentally during the editing process.

@Andreas-ov2fv

2 ай бұрын

It is, but then that picture without metadata would be suspect and unreliable.

@thesenamesaretaken

2 ай бұрын

Moreover it's just good practice to strip metadata from any photo you share on the internet. If the idea is that eventually every file holds metadata about its history then whistleblowers can be ignored if they do not provide the metadata that unmasks themself.

@darrennew8211

2 ай бұрын

@@Andreas-ov2fvNope. That would be true if the certification was "this has not been manipulated." It doesn't work if the metadata says it *is* a deepfake.

@FourthRoot

2 ай бұрын

Obviously, but you're missing the point. You can't fake the metadata and images without the metadata will be suspect by default.

@FourthRoot

2 ай бұрын

​@@thesenamesaretakenIt is possible to preserve certain metadata while redacting other data.

One potential way to certify photos and films could be a blockchain embedded & hardcoded into the camera, which adds the metadata into the fotos taken. Once you click on a displayed link, you would be guided back to the camera, which is certified and owned by e.g. BBC and manufactured by e.g. Leica. Blockchain could be a really good option to trace fotos & films back to the source!

Sounds like a money grab. Someone will have to be paid for the credential verification

Image credentials should use blockchain for distributed consensus about their origin

There needs to be one universal watermark for all AGI

Unless the encryption happens online (in which case, goodbye offline image editing) I give it 2 days (a week tops) before the encryption is cracked and whatever CR desired can be generated. We've seen this before with DRM on music, games, etc. It never holds up and for this one there is a lot of motivation.

I don't see why cameras and image editing software like Photoshop couldn't be issued certificates from a trusted certificate authority. Similar to how websites have certificates for https. The camera would hash a representative portion of the image and the CA issued certificate, attach it to the meta-data, and then re-hash everything again. If the image is edited in any way, the hash doesn't match, and the image viewer (whether on desktop or online) shows an icon indicating the image is not from a verified source. The difficulty here is that some editing is considered acceptable. Color correction and such. This would require programs like Photoshop to have algorithms that judge how much an image has been modified and either re-sign, or deny the edited image. I think this may also issue in a new era of making the general public more aware of traditionally fake images (such as in marketing, etc.)

Credentials can be personalized significantly by using hardware information: processor type (and serial number?), GPU type (and serial number?), network details (name, IP address, etc.), etc. This information can be encrypted using a private key such that only the associated public key can decrypt and validate it. This digital signature can be included in the metadata as an authenticity signature. It is true that software can always be written to avoid including this authentication -- today's software are already in this condition -- however, the AI engines can be programmed to include it as steganography within the images they create as well as matching metadata. In fact, intellectual property laws should already require that image capturing and creation software include authentication protocols to protect the IP of image creators... but then, none of our politicians create such content.

This works for commercial services like OpenAI etc. but not for open source projects.

This sounds extremely easy to bypass. Just screenshot the image, crop it, post it again without metadata. Sites could blacklist all files without metadata, but that would also mean there will be people creating softwares to fake the metadata, while also punishing people who's software of choise is not yet creating the wanted information when exporting. To not count the problem that many use older software licenses to avoid paid mandatory subscriptions, only to get told they are now forced to use the latest software just to get a damn certificate to post an image or a video. Everyone lose but the scammers.

@dmitripogosian5084

2 ай бұрын

It is not a question of bypass. It is that a person who wants to check verification to trust the material, will have a way to check it. And if he can't find metadata (stripped), well, he will not trust it. So lack of metadata will be equated with untrusted source. And those who don't care will not check credentials to start with.

@thesenamesaretaken

2 ай бұрын

I feel it's more about conditioning people to disregard any media without a signature. So if you suspected Trump was secretly meeting Putin somewhere in town, you'd better have a digital signature on any photographs you take or else it's fake news.

@davidrush4908

2 ай бұрын

​@@dmitripogosian5084Stripping the meta-data is easy. So is completely replacing it. Make an AI deep fake. Replace authenticated meta-data with digital camera meta-data. Edit it in a signing app. Increase the contrast imperceptibly, cut a few pixels off the edge, etc. When you save it will have meta-data indicating it is essentially authentic. Fake becomes real.

This was my idea: Every creator needs to sign their images. AI naturally doesn't have certificates. Photographers certificates can be chained to their agency/media company or a federal register. If we get to see a signed picture with a properly chained signature, we can assume it's trustworthy. If it comes out, somebody misuses their signature to push faked content as real, their certificate get invalidated or a remarked gets added via some online service. If an intermediate authority gives out too many certificates to people who don't follow their rules, their certificate is revoked or "stained" with a remark (Image viewer could show something like: "Source is knows to spread AI generated content as real"). This kind of proplem is not really new.

My name's Forest, Forest Gump. My momma always said life was like a box of chocolate. You never know what you're gonna get.

Weird, why would BBC and NYT act against themselfs? Or maybe they don't want competitors to their specialty?

@dmitripogosian5084

2 ай бұрын

Signing under the news as trustworthy was always a part of business model for news agencies. They just want to make more of it. In the world where the other part - collection and distribution of news are getting away from them.

Public shaming of ppl who intentionally try to manipulate an audience.

@judewarner1536

2 ай бұрын

Bring back the stocks and a ready supply of rotten tomatoes.

"... or just a normal day in Florida." INSTANT LIKE

@Thomas-gk42

2 ай бұрын

She's simply great

@WhoTheLoL

2 ай бұрын

Absolutely!

Just maybe though. Yes...

Why not hash the image file (secure hash) and encrypt the hash with a private key? Then, assuming the corresponding public keys are published online by those who care about their original content, so we decrypt, obtain the hash, and rehash the image to check they match. This doesn't seem too hard. If you want to edit an image, you can then rehash your new version and encrypt with your own private key, along with the URL of where to find the public key. Public-key publishing is as easy as putting it up on your website. If the credentials are missing, then the content is deniable. If the public key associated with the file cannot be decrypted by the private key associated with the file metadata, it is fake. If the private key is not at the URL for the claimed publisher, it is fake.

So is @4:43 a deep fake or just a later insert where you forgot to brush your hair

I don't think Deepfakes are going to fool me! By the way you were great in 'There's Something About Sabine!'

I doubt that web browser apps will force users to click a button to learn whether an image's credentials are trusted. I expect (optional) whitelists & blacklists, and (optional) automatic alerts by the browser if there's no credential or the credential isn't whitelisted.

⚠️ It's better to believe in anything, than on what BBC says 😅😅😅

What is stopping someone from just using software without the credentials, then just plopping the modified file into a software that assigns credentials? All this seems to do is add a way to track existing public images.

A good start would be for websites to list the picture's credentials - name, date, place, etc - underneath each picture.

@FourthRoot

2 ай бұрын

And how are they going to know that information without content signing?

@QuentinStephens

2 ай бұрын

@@FourthRoot Due diligence, like licensing the photo from the photographer.

@FourthRoot

2 ай бұрын

@@QuentinStephens But how will they be able to verify that the photo wasn't a deepfake passed off as real by the photographer?

Sabine. Could you create an episode about the current state and future prospects of fertilizer production that doesn't rely on fossil fuels?

It just becomes an opportunity to censor. The positives IMHO don't outweigh the negatives.

Maybe there should be a Dogma 95 for still pictures?

What's the difference between deep fakes and people shilling corporate products?

Digital signature from a accredited issuing entity. Is the way HTTPS work's.

'Deepflakes and credentials '. I shook the box and looked to fine print; ' Made from local ingredients not less than two percent'.

"Indistinguishable." Yeah, no. We're not far, but we're definitely not there yet.

@olafsigursons

2 ай бұрын

We should begin to think about it before we are. It won't be long.

Embedding a public key into the image and using a public private key scheme for authentication is the obvious choice here.

@goodfortunetoyou

2 ай бұрын

Blockchain or some distributed database for the reverse image search?

@dmitripogosian5084

2 ай бұрын

I bet it is something like that.

I hope that committee didn't get too much funding for this idea... Being able to trace the source of the image all the way back to its creator even when being edited etc. would be a far better idea, might even be an actual real usage for blockchain...

@vladimirdyuzhev

2 ай бұрын

Remember whistleblowers. Those folks wont be happy if any image can be easily traced to the author.

@modal-kkurt4986

2 ай бұрын

First media outlet could be the one to put it on the blockchain. Doesn't have to be the original source. Probably not worth bothering as media doesn't tend to have any teeth anymore anyway. ;)

this is basically SSL/TLS but for image identity.

@shinjirigged

2 ай бұрын

which i should say that I'm for; because the issue with deepfakes is not who believes them, it's how little work is involved in creating them vs much work is involved in fact-checking. So an SSL-like cert that verifies that an image was taken by a particular piece of hardware is very valuable. Because while it will not eliminate the noise of fake media, it will allow us to better identify authentic media at a lower cost.

@shinjirigged

2 ай бұрын

also it's already an arms race.

@vladimirdyuzhev

2 ай бұрын

And note that state actors are known for issuing fake certificates in the past.

Let's start a fund for a new outfit for Sabine. 😉

I see future OS snapshots that include metadata, or flag the screen cap as possibly fake.

same coevolution issue lol