What kind of Solutions for the upcoming Deepfake video problem might be effective?
Qualify Deepfake videos in scale have arrived and are probably one of the biggest existential dangers we face, along with climate change, nuclear and biological weapons, superbugs causing an epidemic, and a massive solar flare like the Carrington event of 1859 (plus of course getting hit by a huge asteroid). We can add Deepfake videos (enabled by AI technologies) to the list of technologies where our ability to manage the risk created by the technology is not up to par with our mastering of the use of that technology for a wide variety of purposes.
Biggest Technology risks: https://www.weforum.org/agenda/2017/01/what-emerging-technologies-have-the-biggest-negative-consequences/
In the security world I work in, here is one way to classify risks:
Deepfakes are a particularly difficult problem because people trust videos a lot more than text. Even if you could educate people effectively in mass, the sheer volume of videos and the psychological effect of seeing a video is hard to combat, and this virus can basically result in the short circuiting of our collective brain.
Whenever dealing with a complex issue like security, a good way to understand the problem is to pretend you’re evil, and think about how you would achieve a certain objective. For example:
OBJECTIVE:
Sway public opinion to give power to somebody and take power away from somebody else (e.g. sway an election)
STRATEGY:
1. Create multiple fake videos of a prominent person saying things or doing things they didn’t do.
2. Post them to social media through various channels, and provide a nice clickbait catalyst
3. Use many different videos and many different channels (e.g. fake accounts, unsuspecting friends).
4. If you’re really good and have the resources (and we already know some entities do), flood the Internet with so many fakes nobody knows what is real.
5. Augment or bundle with fake news
So what are some solutions to protect us against this virus?
1. Outlaw the technology or require those creating it to put a watermark in the videos (something Congress has proposed). Ha ha ha!
2. Train the population to only look at videos from trusted sources like CNN, BBC, etc by going directly to their website. Even funnier, ha ha ha ha ha!
3. Analyze all the videos posted for evidence of fakery (with deepfake detectors), contact social media companies to take them down. Ouch, it pains me to think of where that solution fits in the above matrix.
4. Then find all the people who posted fake videos, where they got them from and then ban them from social media and/or throw some of them in jail. Ugh…this isn’t going to work either
5. Force social media companies to require videos to have digital signatures from known sources in order to PROMINENTLY display them as TRUSTED along with the Logo of the source. (Similar to how Twitter has an indicator next to celebrity accounts to indicated that these are verified). We’re getting warming, but this is really difficult.
#5 is not a great solution but seems ultimately a lot more effective than the others. People watch videos largely on social media platforms and these are easy to post fakes on vs. a traditional media channel which controls their videos and at least has some semblance of integrity forced upon them due to the need to manage their brand image. So let’s play a game and think about what it would take to pull #5 off. But first let’s keep something in mind:
Humans are terrible at managing risk unless:
A. They are directly affected personally (i.e. they have an incentive to do a good job and ensure results)
AND
B. They are wise (i.e. have relevant experience)
AND
C. They have the resources to stop or prevent problems
AND
D. They have the resources to monitor #3 and make adjustments
AND
E. Whoever is in charge is aware of the problems and strongly supports the program
To pull this off:
- You would need to design something effective, for example:
a. Videos would have a digital signature tracing their source
b. That source would need a certificate from an authorized entity (e.g. a Certificate Authority today issues certificates to websites to ensure identity authenticity over https, something people do every day with web browsers).
c. Videos from well-known trusted sources would be prominently displayed (e.g. with a green border and a logo, and those from unknown sources would have a flashing red border with no logo).
2. Congress would need to pass legislation FORCING social media companies (e.g. Facebook and YouTube) to do this.
a. You would need to prove the impact to Congress (e.g. after a hacked election) to get the ball rolling
b. It took us about a decade to get the FDA to start taking the security of medical devices more seriously, and this only happened after devices like Pacemakers were hacked in a lab. And they are just getting started, it will probably take another 5–10 years before reasonable security is widespread. Does Congress really take this seriously? Can we afford to wait five or ten+ years?
c. Who’s going to pay for all this?
d. The problem from above — Congress is (A) not directly affected personally, and (B) not wise. (Well actually perhaps they are wise if you consider that their objective is simply to get re-elected and the easiest way to do that is simply do what the donors and lobbyists want).
3. Congress would also need a program to MONITOR the effectiveness of the solution and force adjustments to be made as required
a. And who’s going to do that, and do it well? The very thought of it, makes me want to pour a glass of whiskey…in the morning before my breakfast.
And this doesn’t even take into account that even if the above is done effectively, many people will simply ignore whether a video is trusted or not and believe what they want to believe. How many people engage on links that are clearly marked as sponsored as if they were true? How many repost fake news from sources that clearly don’t have a known brand because it confirms what they think and it gives them something to bitch about and get attention?
We also don’t know to what extent certain tabloid news sources will continue to thrive. Conspiracy theorists love to exaggerate and make stuff up, and it’s big business. Can the branding of news (once we have the digital signature solution in place) solve this problem, or will certain companies with a strong brand still disseminate fake news as a business model? If that’s the case (and there is evidence that it is), that we truly are screwed. Democracy requires an educated population, no way around that one. And education has a much different meaning in today’s complex world vs. when Jefferson first penned that statement.
I fear that our capacity as a species to handle this kind of problem might be our final undoing. I hope not. But if the predominance of reality & shock TV, the effective use of issues like abortion & gay marriage as a distracting political football, and our problems with actually effectively dealing with issues like climate change is any indicator, we might just be following the intelligent species extinction rulebook that may to be the driving force behind the Fermi Paradox. But we can take comfort in one thing, we may not know to what extent it’s happening, and we can live our lives in bliss full of plenty of entertainment and whiskey. I actually personally recommend Cannabis, it’s more cost effective and has less side effects than being a drunk.
