The Incredible Creativity of Deepfakes - and the Worrying Future of AI | Tom Graham | TED
HTML-код
- Опубликовано: 18 май 2023
- AI-generated media that looks and sounds exactly like the real world will soon permeate our lives. How should we prepare for it? AI developer Tom Graham discusses the extraordinary power of this rapidly advancing technology, demoing cutting-edge examples -- including real-time face swaps and voice cloning -- live from the TED stage. In conversation with head of TED Chris Anderson, Graham digs into the creative potential of this hyperreal content (often referred to as "deepfakes") as well as its risk for exploitation and the new legal rights we'll need in order to maintain control over our photorealistic AI avatars.
If you love watching TED Talks like this one, become a TED Member to support our mission of spreading ideas: ted.com/membership
Follow TED!
Twitter: / tedtalks
Instagram: / ted
Facebook: / ted
LinkedIn: / ted-conferences
TikTok: / tedtoks
The TED Talks channel features talks, performances and original series from the world's leading thinkers and doers. Subscribe to our channel for videos on Technology, Entertainment and Design - plus science, business, global issues, the arts and more. Visit TED.com to get our entire library of TED Talks, transcripts, translations, personalized talk recommendations and more.
Watch more: go.ted.com/tomgraham
• The Incredible Creativ...
TED's videos may be used for non-commercial purposes under a Creative Commons License, Attribution-Non Commercial-No Derivatives (or the CC BY - NC - ND 4.0 International) and in accordance with our TED Talks Usage Policy: www.ted.com/about/our-organiz.... For more information on using TED for commercial purposes (e.g. employee learning, in a film or online course), please submit a Media Request at media-requests.ted.com
#TED #TEDTalks #AI 
Наука
Yup we’re screwed
Yup 😅
So screwed.
Or blessed, it all depends on what we do with it
They said the same thing with splitting atoms. We're still here aren't we?
@@PBFoote-mo2zr well guess we have to put our big boy pants on and keep on keeping on. It's inevitable that we will always find a way to weaponize anything that could be weaponized, including artificial intelligence.
So we are talking about “Human” connection, emotions, socialisation and all the fancy jargon… to talk about (and sell) a “machine” generated modified content. That too from people dead or alive, with nearly no distinction. Nothing about this says “human” experience really. The legal and privacy issues aside, imagine the psychological distress this could generate in people. As much as I appreciate this video from point of awareness, it barely addressed the massive downsides of this technology. Something even the creators themselves don’t comprehend (or don’t care?)
Exactly the feeling I had watching this. I don't know why they're laughing about this. There's nothing funny about this at all...
Value of human life will hit an all time low. Why worry about your ACTUAL aging, perhaps ill, grandparents when you can just can a more convenient AI version for when you want to "connect."
This!!! When he mentions creating AI generated interactions with family who may pass on… that’s essentially inhibiting natural reactions to grief and day to day communication with others
I've never seen a pitchman less confident in what he was selling.
Yeah you can do a lot of negative, but I *think* we can do good things too. LOL we actually are fucked. Just wait until the majority of population can't tell what is real and the fear takes over.
@@alexpavalok430 tbh I fail to believe any of this positives. This technology will have 99% of profoundly dangerous and negative impacts, and maybe 1% of entertaining/useful. Even the family example about grandparents, while some might find useful, is just terrible for your mental health. Jesus, what about imprinting your ex or dead fiancee onto fucked up AI generated videos? It feels like we are already living in Black Mirror series :(
We're in a bit of a pickle. Things are moving fast and social media has already had a pretty destructive impact on our society. Now, with these new technologies spreading like wildfire, it's going to be tough to regulate and control. There will likely be some pushback from tech advocates, which could make things spiral out of control. It's definitely a confusing time to live in. It's clear that our children will grow up in a completely different society. The breakneck pace of technological advancement, combined with the devastating effects of social media, mean that the world they inherit will be almost unrecognizable from our own.
We shouldn't regulate this, regulation is made for things which we absolutely have to regulate and the it should be done with grace and a nudge not totalitarian control, which seems to be what a lot of people are advocating for right now.
It's already unrecognizable. We just don't understand that because we are living it.
I'm keen to let it run its course, keeps things interesting. The information should stay free as meant to be. Regulation will just cause issues & pushback is inevitable.
Obviously. Time waits for no one. Your children will get used to it like how the genz are most familiar with technologies but not the generations before them.
We are on the brink of the death of genuinity, honesty, and sincerity.
And the death of appreciation for pure talents and raw human skills.
Genuinity? I see we have already passed the point of no return on the English front.
The idea of saving your loved ones after they've passed, or even worse, while they're still alive, is going to create some super awkward relationships
Photographs already capture one's appearance, videos one's likeness and now AI models one's mannerisms.
Why will this not simply be considered in the future as merely a more advanced recording in time/space?
@@BenjaminMeasures because you can't develop a social relationship with a photograph
@@recks1151 A "social relationship" with an AI model is fantasy. Such fantasy has existed long before technology.
I'm pretty sure people have fawned over characters in photographs/phonographs/paintings/tales.
I ask again- why is a 3d hologram, or AI model any different in this respect?
It is early, but people are starting to develop social relationships with AI, it no longer is fantasy and it's clear that future models will make this even more prominent. SnapchatAI is a good example
@@BenjaminMeasures because the ai will develop conversation and “personality” that would otherwise only be in a single brain’s imagination.
We’re so fucked. There’s no possible way the positives of this outweigh the negatives. Tom talks about owning our data but that’ll never happen.
It’s impossible. Regulation will be too unwieldy and impossible to monitor. We are at a very low point in moral standing and the adversarial standoff is encouraged and rewarded.
AI has already stolen everyone's data to train on. That notion is ridiculous. ",We need institutions to identify authentic media" wow, as if that has already worked for social media which is a lesser technology...
I totally agree
@@lilliangraham9850 I mean think about how companies currently sell our data. Like the things we search or like or whatever. That’s OUR thoughts/interests in essence, and companies gather that information and sell it to advertisers. Why aren’t we getting 80% of those sales? I mean we’re the ones doing the work of liking a post or sharing a video etc. idk how much companies pay for data but we’re basically doing the work, then a middle man is telling advertisers what we did and they’re getting paid for that!
Human experiences shouldn't be 'decoupled' from anything.
We should be able to control our image across any platform and search engines. Once our photo is on the internet we now have no control over who uses it and how it's used. While I like this technology, it needs to be used with great restraint and responsibility.
What kind of idiot uploads their pictures on the internet?
Yeah, it's a good point. Most data on the internet was created, unknowingly powering AI companies. Even if the data and content is/ was public. Public data does not necessarily equal training data. I can definitely see platforms suing OpenAI or other AI companies, when the advent of certain AI services, that were trained on said platform, may cause the platform to fail.
@@Freakei even before AI, companies have been farming our likes/interests and sell them to other companies so they can market more effectively. Why aren’t we being paid for that information? There’s no way that the newer AI industry will use restraint or responsibility.
@@slimeruthlessroyals1026 good Point. I don't expect AI companies to do that at all, hence the prediction that companies may sue, even if their Data was Public to begin with. If we take the right to be forgotten (EU) into account, even individuals could sue.
Just a few things :
1. What will be the future of evidences for crime investigation? I don't think we are supposed to adapt to such a situation where we cannot trust anything.
2. He talks so fondly about having images/avatars of your loved ones so similar to the actual person and being able to have 'extensions' like speaking another language or even existing after the person's death. Am I the only one to find it tremendously violating the essence of a person being a person? It's really creepy and it's beyond the existential crisis that most of us have at some point during our lives and already struggle with.
I am with you
30 years ago the internet promised a bright future. Today's 'innovators' are all promising a troubling future. Then why are we doing it???
fun
Reality is just a playground for the capitalist
How can the internet promise anything. It's not a human
@@kahyui2486 Is English not your primary language?
This is terrifying
This is the so frightening, we are so screwed.
“It’s not within our control” great. 👌
There should be a law regulating deepfakes to have mandatory identifiable software signatures by law that can be used to distinguish between what is real and what’s altered.
What's stopping you from faking them also? It would have to be something like public key cryptography. So owner owns something and picture owns something that can be compared by anyone, but then comes quantum computing, however there are already propositions of some new ways of algorithms to make it secure, that is hard for quantum to hack.
Trump, Biden everyone we see on Tel-lie-vision the question is how comes they are allowed to expose it?
People should be able to own their image and voice. It’s a gross invasion and this kind of thing needs to be regulated immediately.
It is nice to see that TED is hosting people that talk about AI. The invisible AI revolution is here I guess.
What would you do if you have the options now as a younger generation, I guess we have to start doing a lot more education.
how comes the TED Talk's corporation is allowed to expose the lies??
@@timetowaketheheckup123 can you be more specific in your criticism? You are discrediting yourself
@@jirivlcek3288 exposing the fact tat we see alot on mainsteam TV and the media!
and the door opens for bad actors going to be a nightmare
When these people talk about all the positives of AI, without even thinking of the impact of it, I hear it in the voice of John Hammond from Jurassic park
This technology comes from a place of “could”, rather than “should”. I envision very few positive applications for the technology, but very many nefarious ones.
People w/ facial or behavioral or vocal differences could "wear" an avatar of themselves as a social prothesis.
Already actors (James Earl Jones, for one) are capturing their likeness/voice to secure continued income for their families.
People w/ young children & terminal illness will have a way to be present for them.
Many therapeutic uses like troubled teens could see themselves as successful/happy adults to visualize a positive future.
That's what people said about computers, and the internet.
My dad had me when he was older, after his passing I began to see pictures of him when he was younger and boy that had a profound impact on the way I saw the world, so I can see some advantages to this.
Funny how in the end he never answers a real thought solution to the downsides of AI but prefers to only focus on "the good and the wonderful things", most likely they still haven't thought through the solutions because they don't see yet how big of a problem this could become before it starts being really useful, beyond entertainment. Food for thought.
there's no problem, courts will throw out of the door alleged evidence, lying detector machines will be perfectioned, people will be declared innocent or guilty on this basis, plus jury plus judge, plus evaluation of the history of the accused person.
@@fredzacaria the thing is, that system doesn't even work correctly now, how can we expect to be able to make it work when there are more complex factors involved
I agree with u 💯 I have had debates about this with many ppl on the internet and most of the positive things they think of revolves around entertainment and cinema, but that’s stupid coz we can’t put entertainment over sanity
@@Random-mj5np
What this tech does is lower the bairer to entry, now all you need is powerful consumer hardware to do what was once only possible with a big budget and VFX.
It's not that this was impossible before AI, it was, it was just costly, and usually only seen in big budget video production. All the AI does is automate the process and lower the cost.
The is the way it changes our world, not bc it can make a realistic convincing fiction, that was already possible beforehand, but now more people can do the same thing.
"Technology moves faster than our ability to civilise it"
Great to see these kinds of talks on TED. Thank you!
how comes the TED Talk's corporation is allowed to expose the lies??
downsides are far more that the upsides.
same for guns
@@deepfakescoverychannel6710 please shut your mouth
I trained my own AI model yesterday which makes photos of me in any location with any clothes. It took 30 minutes. We are already at the point where nothing we see online can proven to be real for sure
Post truth
@@PBFoote-mo2zr That's A LOT easier said than done and not realistic when the entire method of training AI is imitating human work.
@@PBFoote-mo2zr nah. The models will be better at faking it than detecting a fake. The tech will get so good it won't be possible to distinguish without some sort of token associated with it. Welcome to the world of everything being an NFT...
@@PBFoote-mo2zr i think that this is one of the pros of the blockchain and of nft-style verification of ownership. it will help to trace back who made what and how; you could have a level of verification to say it was digitally changed in the exported file as part of the metadata, which automatically embeds as a kind of “verified checkmark” to say it is fake or real and how in a twitter-community notes style
@@PBFoote-mo2zr you dont already live this way? rip to me
Hey Chris Anderson, fantastic talk! I laughed so hard that my pet parrot started imitating the Mona Lisa! 😄
By the way, any chance you could translate the talk into Brazilian Portuguese? I'd love to show it to my mom, who's convinced she can spot a Deepfake from a mile away. She'll either be impressed or think it's some next-level prank from her soap operas! Obrigado! 🇧🇷
If you want to know what this tech will/wants to do, track who is financing it. You will understand what they "really" want from it besides "fun"!
The technocracy only depends it's grip on society
Those are the same people who fund everything lmao. People have been trying to boycott nestles for decades.
Can't beat it join it
What software do they use to create these videos?
We're going to have to have OTPs as people when communicating with others to verify that who we are speaking to is really them.
Thanks comrade
We probably have already given away these data of ourselves sometime ago and not even know it.
Never did I think that I would be against technological advancements 😶
It’s exhausting how much our world is changing. You just want to say stop to all of this. Can’t we all just live? It’s like we’re constantly being challenged to accept the new normal. When will normal just be normal again?
Never. Technology's Pandora's box was opened probably the moment the atom bomb was made. As long as our economy is profit driven, the new normal will be forced on us, even if we don't want it.
한국어 자막 감사합니다..
잘 볼께요
Technology is both intriguing and perplexed to minds like mine that are 😳😳 at this moment. Are we being replaced?? Just a thought
Yes, it's evolution baby.
Traditional photography will be making a comeback as something that can't be faked so easily. If it's not on film you can assume it's generated.
But why can't AI learn how a film photography looks?
I was always suspicious that TC is an artificial construct
21st century isn't the extension of previous ones. We are seeing things that previous generations even didn't think of them. I am certain, IT/ technology will bring more!
Thank you for the video.
That's amazing, and worrying 😢😅
I think if we consider that digitalization has been super fast and things like cyber security or digital infrastructures already today have trouble keeping up, this whole Deepfake topic has come wayyy too early 😅
7:24
Advocating for industry standards that prioritize ethical AI practices is essential. Let's work towards the establishment of clear guidelines that promote responsible AI development, preventing the misuse of deepfake technology for malicious purposes.
There's no need for this.
This is very concerning. The adult film industry (who’s name shall not to be spoken) was a key industry in developing the internet through use of online transactions: credit cards, online banking and shopping, this is going to blur the line of what is child exploitation as the models maybe consenting over 18’s but to all intense and purposes the facial features aren’t how does that fit in with the law🤔
This is really spooky
We need to grasp the “Power Of Now” (to quote Eckhart Tolle). To live in the moment - not chained to the past. And definitely not in a future that doesn’t exist. Obviously there are brilliant minds behind this tech. But for all that intelligence- how stupid and ignorant to claim that “decoupling human experience from where it happened and the moment in time that it happened” will allows us to understand ourselves better. What a joke. The hubris that it took to build this tech is out of control. Worse, the cavalier way in which these companies approach the immediate and immense dangers of deep fakes is utterly absurd. They will suffer the consequences of their creations.
But sadly, we will too.
What is the actual benefit of this and how is the downside not 100x in magnitude? So we can create synthetic versions of ourselves. Talent can scale itself more easily. So we can make videos more easily and efficiently. Great. A more efficient way to create amusements for ourselves and others. The potential harms from this are incalculable.
What's to say if one cannot fully own a web address not owning your own image could also be leased out. Food for thought.
Nobody wants this.
hey i want it
To us we need to fight and win !!!
When you consider how many people that will believe a picture or video that is poorly done and obviously fake…yeah, this is worrisome
We should definitely be doing our homeworks regardless if humans are afraid or not
The singularity will arrive sooner than you think.
One thing, I couldn't imagine, is this, what they are doing on stage, happening only our screens or what.. I couldn't understand 😬
Only on screen
They always want to say mental health is an upside and they are crazy… it’s too dangerous decoupling human experience
In a world dominated by pictures and films and with an audience with very short attention spans this technology has some catastrophic potential. On the downside were doomed anyway.
It just makes me scared. Democracy is screwed. We can not have an informed public anymore.
This is my very first time to see dcreativity of ai for deepfakes.
And it's scarying me for sure
The governments should just make this illegal. There are far more downsides to this than upsides.
great for ultra low bandwidth high res video chat
Regulation will be irrelevant because of this.
I'm concerned he never answered the question: can we copywriter our likeness or not? That's the only concern I have. There would have to be legislation to reverse the TOS we all signed when we signed up for Facebook, Instagram, SnapChat, TikTok, Twitter, RUclips and any other place that allows us to upload photos or profile photos. 🤔 The horse has already left the barn, we are going to have to move quickly to enact some basic protections going forward.
Imagine politicians using these
I am sure they already are.
We're all going to look back at the day we heard AI became a topic because there's no going back and AI is a printing press of moment in time!
The printing press wasn't used to bomb two cities in Japan.
All time is now.
Besides having fun I'm not convinced that there are positive benefits to this technology. Creating loved ones to talk to will not help you deal with grief, it will just pull you away from the living world completely and consume you. Using anyone's likeness without their explicit and informed permission is crossing a boundary. People have a right to their autonomy. It's very different than creating a fictional character for artistic expression. This is not someone who looks like Tom Cruise or exaggerates his mannerisms. IT IS TOM CRUISE. And he has a right to say no I don't want to participate in this Ted Talk.
Why?
Death creates legacy....
Immortalizing celebrities like Tom Cruise means no new actors will surface ever again as we keep consuming "dead but famous" actors forever
This is the type of guy who would take the blue pill.
So we'll have to copyright the data that constructs our model to have a right on our own image (hopefully for free). What if someone copyrights someone else's model? I hope you won't be able to, and if it won't be possible, why would you have to copyright it in the first place? What about dead people that are well documented in videos? I guess the owner of that media would have the copyright in the first place, right? 100+ year media can't be copyright so I guess we'll have a bunch of old time stars resurrecting soon...
Gotta love the awkward silences during the "jokes", like everyone in the room is aware of the disastrous potential this tech has. But like the guy said, there's nothing to do about it. Quite an interesting future we have ahead of us.
Well put. Hey lets open Pandora's Box and see what happens.
Why though?
It was amusing to listen to the man discuss a scenario where one could communicate with a loved one by utilizing a different representation of them. Fortunately, both the host and I shared a sense of unease towards the concept, mirroring my own initial reaction. - with the aid of ChatGPT.
Put Anna May Wong in "Good Earth" movie! Please! (She was the Chinese-American star who was passed over for this iconic movie because of anti-Asian racism.) Please someone do this!
This man is dangerous😮
so what's happening from 4:40 to 5:10 ? Is the AI mimicking the woman's body language and copying it into the man's body? I didn't quite get it.
What was on the screen!?
The reveal:
This presentation never actually occurred.
Counterfeit people. This should be much more worrying than Chris Anderson acknowledges.
if history wasn't over before, it is now..
The worrying future of book reading. Read a book on the subje. Don't let these guys pre-chew your thoughts!
The first time I saw these deepfake videos I got very worried. The creator is crazy, this will be used by many evil people in various evil ways.
They don’t have a god damn clue about the nature of our reality if they think you can simulate your loved one instead of dealing with the grief and their passing. Wow hope everyone really hears what they are saying as it’s kinda cross and mechanical
This is scary
We are designing an isolating pod world.
Now that you’ve created a “original” video to be reference by the authentic you… what happens if you have a twin or a dobbleganger? Will the first person who does it get the ultimate right?
Disturbing… don’t you think? It’s about deception.
If life was a purely linear series of sunshine and lollipops, you might never know that you are built for recovery. Isn’t it far more interesting to know you are adaptable?
Your very bones are resilient. When one breaks, your body has coding to create the cells to rebuild you from the inside.
Practice your powers of resilience in the face of life's ups and downs. Leave the binary of ones and zeros to computers. You are worthy of a far more interesting life.
Well damn
People nervously laughing at the demise of the human race.
TLDW from your friendly neighborhood AI:
AI developer Tom Graham discusses the extraordinary power and potential risks of rapidly advancing AI-generated media. He demonstrates cutting-edge examples, including real-time face swaps and voice cloning. The conversation with Chris Anderson, head of TED, explores the creative potential of hyperreal content, often referred to as "deepfakes," as well as its potential for exploitation. They also discuss the need for new legal rights to maintain control over our photorealistic AI avatars.
(You're welcome, humans)
I'm screwed, none of this leaves the uncanny valley for me. Takes me right out of reality. Very distracting.
That means there could come a day that video and audio evidence might not hold up in court anymore 😅 You could also bring artists back from the dead - literally
Having rights and rules about the usage of your image is great, but most of the fake/creepy video's will be made by teenagers who do not care and the damage will be done before they're removed.
Since the beginning of “accepting terms and conditions” of installing any software on a computer or using a cloud service, the vast majority of people do not read the terms and click “I accept “ almost immediately they see the chance to do so. We will give up our rights to our photos, voices and videos and we will no longer fully own this data and the big Tech companies will own these rights. We want and like free services and that’s how they will get us.
Welcome to Orwellville, what does it mean when deep-cheap-fakes punk us 24/7?
Me, I’m reviewing Small Squad Tactics of the Butlerian Revolt.
*只能說小玉走得太前面...用錯方式了*
Ah…so much I wanna say about this….while this “may” be “innovative”, but this is really evil….
We are living in a Simulation!
We are creating a simulation for ourselves.*
Terrifying