Deepfakes: Nothing Is True, Everything Is Permitted

.
H+ WEEKLY

.
Issue #402

"Never trust anything you see on the internet" - Abraham Lincoln

The history of deepfakes starts in 1997 with Video Rewrite, created by Christoph Bregler, Michele Covell, and Malcolm Slaney. It was the first-ever computer program to fully automate the process of facial re-animation. The next improvement in quality came in 2001 with the active appearance model.

Nothing much happened in the space until the introduction of generative adversarial networks (GANs) by Ian Goodfellow in 2014. GANs consist of two neural networks - one network is generating an image, and the second one evaluates the image. When an image is rejected by the second network, the first one learns from it and adjusts itself. Repeat this process a couple thousand or million times, and you get an AI that can generate images almost indistinguishable from real ones.

The AI community took on the idea of GANs and started experimenting with them. In 2016, Face2Face has been released. One might put this as the beginning of the modern deepfakes. Face2Face was the first program to allow a real-time face swap in videos. It made waves in the AI community and inspired more research into face swaps and deepfakes.

The general public first learned about deepfakes in 2017 with this infamous Obama deepfake video from Buzzfeed. From there, the deepfake technology became better, cheaper and more accessible. If you have a good enough computer, you can run one yourself. If not, there are online services available.

Like every technology, deepfakes can be used for good and for bad. For example, BBC used deepfakes to hide the identities of Hong Kong protesters in their documentary. Deepfakes found their use in the movie industry. A good example of that is Disney, which used old-school (before GANs) deepfake technology to bring back Peter Cushing (who died in 1994) to "play" Grand Moff Tarkin in Rogue One. The same technique was also used in the last scene of the movie to recreate young Carrie Fisher as Princess Leia.

And just to illustrate how good and accessible GAN-based deepfakes are, just three years later after Rogue One was released, one person was capable of creating a better deepfake Tarkin than the entire Industrial Light and Magic.

These days, however, deepfakes are portrayed as a tool for misinformation and abuse. 4chan users used a deepfake voice service to make celebrities say racist, transphobic, and violent things. We have had fake Zelensky and fake Putin calling to surrender or offer peace talks respectively. And just recently, a well-known streamer got caught using a service offering deepfaked porn videos with female streamers.

It's not just famous who are the target of deepfakes. It is very easy to deepfake someone into revenge porn, put that on the internet and destroy someone's life and career. All that is needed is enough photos of the target person. Microsoft has an AI that can clone any voice from a short, 3-second-long sample. Just one voice note and someone can pretend to be your friend or family member.

The question then is, how do we detect deepfakes? The best summary I've found is this paper published in February 2022 analysing different detection methods, from deep-learning-based to blockchain-based. Some of the examined methods report a 98% detection rate but are highly sensitive to the training data. With some tweaks, the detection rate goes to 99% and solves the sensitivity problem.

The battle between deepfake generators and deepfake detectors will the going on. What we can do in the meantime is to be sceptical about anything we see on the internet, and double-check if what we see is real.

Share

Forward

MORE THAN A HUMAN

▶️ Super Soldiers (32:28)
Science fiction is full of enhanced super soldiers but how likely are they to become a reality? Isaac Arthur explains how super soldiers could be created (from genetic engineering to cybernetic enhancements to cloning), which human traits to augment, and the possible problems with creating a caste of enhanced soldiers. I like the closing thought of this video - maybe instead of focusing on creating super soldiers, with all problems they come with, it would be a better idea to focus on creating super citizens.

These prosthetics break the mold with third thumbs, spikes, and superhero skins
There is a substantial effort put into engineering prosthetic limbs that perfectly match the natural limbs. But there are some researchers and artists asking why should we match the nature? Why not go the other direction and make prosthetic limbs that are expressive or add new capabilities to the human body? This article quotes Victoria Modesta, a bionic pop artist - “You should be able to experiment with not just your wardrobe but your limbs, your power, your everything.”

One Third of Americans Would Use Genetics Tech to Make Their Offspring Smarter, Study Finds
National Institutes of Health researchers have found that if you offer people who may conceive using in vitro fertilization a polygenic screening or CRISPR-style gene editing to increase their kids’ chances of getting into a top-100 ranked college, 28% will accept gene editing to make their kids smarter and 38% will choose the test. But there is one important assumption: researchers told the respondents both options are free and safe.

ARTIFICIAL INTELLIGENCE

Lockheed Martin's AI flew F-16 fighter jet for 17 hours
Lockheed Martin announced in this press release that their AI flew an f-16 fighter jet for over 17 hours as a part of their VISTA (short for Variable In-flight Simulation Test Aircraft) project. "VISTA will allow us to parallelize the development and test of cutting-edge artificial intelligence techniques with new uncrewed vehicle designs," said Dr. M. Christopher Cotting, U.S. Air Force Test Pilot School director of research. "This approach, combined with focused testing on new vehicle systems as they are produced, will rapidly mature autonomy for uncrewed platforms and allow us to deliver tactically relevant capability to our warfighter."

Inside the ChatGPT race in China
ChatGPT made a smashing entrance not only in the West but also in China. According to this article, the Chinese were impressed by how well ChatGPT understands the nuances of traditional and pop culture in China. Meanwhile, Chinese tech giants such as Baidu and Alibaba announced they are already working on their own versions of advanced AI chatbots, while smaller companies are exploring how to incorporate chatbots into their business. However, as this article points out, it will take time. It took Baidu 18 months to release a homegrown answer to GPT-3. The US ban on selling state-of-the-art chips to China does not help here either.

Netflix Brags That It Used AI to Replace Human Animators In New Anime
Netflix Japan tweeted they have been using image-generating AI in a new anime. This did not go well with the fans, seeing how Netflix is resorting to AI in an industry notorious to overworking and underpaying animators. And since the show must go on and there is not that many people willing to go through anime hell, the AI-powered image generators might be more and more popular in the industry.

The original startup behind Stable Diffusion has launched a generative AI for video
Runway, the company behind text-to-image generator Stable Diffusion, released an AI that generates videos. Named Gen-1, this new AI can transform existing videos into new ones by applying any style specified by a text prompt or reference image. “We’ve seen a big explosion in image-generation models,” says Runway CEO and cofounder Cristóbal Valenzuela. “I truly believe that 2023 is going to be the year of video.”

ROBOTICS

▶️ Small Drones & Loitering Munitions in Ukraine - The terrifying rise of cheap precision (1:08:39)
Perun gives another excellent analysis of how drones - from off-the-shelf Mavics to larger kamikaze drones like the Iranian Shahed and Russian Lancet - have been used in the Ukraine war. The analysis touches on the cost-effectiveness of drones, how they are being deployed (from scouting missions to direct strikes), countermeasures, how they changed the behaviour of soldiers in combat zones ask what lessons we can take from the terrifying rise of cheap sensors and cheaper precision on the battlefield.

Amazon’s Delivery Drones Served Less than 10 Houses in their First Month
According to recent reports, Amazon's drones have delivered packages to less than 10 houses in their first month. Compared to other drone delivery services, which made thousands of successful deliveries, this number is laughably small. However, as this article points out, context matters. Amazon's MK27-2 drone is way heavier than its competitors, making it fall under stricter FAA regulations and restricting where can it fly. Amazon has announced in November 2022 a new, lighter MK-30 drone, which is scheduled to enter service in 2024.

▶️ DroneCase: The DIY Phone Case That FLIES (Not a Scam) (8:49)
Nicholas Rehm shows how he built a case for his phone which is also a drone. With it, he can leave the phone in the air and it will not fall, allowing him to make some cool photos or videos. The project is open source, so if you are interested in building your own drone case, you can do that.

BIOTECHNOLOGY

From Lab to Market: Bio-Based Products Are Gaining Momentum
Propelled by government investment and shareholder demand, manufacturers are pushing to get bio-based products into the marketplace. In the US, President Biden launched a $2 billion biotechnology and biomanufacturing initiative. Similar initiatives are happening in Europe, with the EU proposing new rules to require all product packaging to be recyclable and possibly reusable by 2030. Such initiatives could propel the bioeconomy and bio-based products to grow to more than $7 trillion by 2030. This article gives some examples of how such products might look like, like a leather-like material made from fungi or homes 3d printed using sawdust.

This issue was brought to you by our awesome patrons whmr, Floris, Eric, Andrew, dux and Tom! You too can support the newsletter on Patreon.

Thank you for subscribing,
Conrad Gray (@conradthegray)

If you have any questions or suggestions, just reply to this email or tweet at @hplusweekly. I'd like to hear what do you think about H+ Weekly.

Follow H+ Weekly!

You are receiving this email because you signed in to H+ Weekly newsletter

Our mailing address is:

H+ Weekly

London

London, London EC4 2AA

United Kingdom

Add us to your address book

Unsubscribe Update subscription preferences

This email was sent to you
why did I get this? unsubscribe from this list update subscription preferences
H+ Weekly · London · London, London EC4 2AA · United Kingdom

nbspactive appearance model infamous Obama deepfake video Hong Kong protesters Grand Moff Tarkin young Carrie Fisher two neural networks realtime face swap good enough computer accessible GANbased deepfakes The AI community

Deepfakes: Nothing Is True, Everything Is Permitted - H+ Weekly - Issue #402