3 Summaries of the Web Report

1ST: Disruptive Technology

Technological advancement made it convenient to deliver messages in different areas of communication. It also plays a huge role in the development of Disruptive Technologies. An innovation originating from the low-end or new market foothold that are capable of affecting the existing markets and value networks. Usually, this is a result of adaptation and modernization.

One example of a Disruptive Technology. The word “deepfake” is coined from merging the words “deep learning” and “fake” (Brandon, 2018). Its mainstream original use was applied to public personalities that have many digital image materials available to train in order to develop a fake video or photo overlay of the same person into an existing piece of media product, completely out of context.

Batir (2021) announced the most fundamental algorithm employed by deep fake AIs is Deep Autoencoder, which has two components. The encoder tracks the major facial structure of the overlay image to identify its fundamental facial feature, allowing the decoder to align the superimposed faces together and reconstruct the original image’s face over the overlay face to give it the body of the original image. This is a type of Artificial Intelligence used to create an altered image, audio, or video and is being created by using two Artificial Intelligence or AI algorithms namely, the generator and the discriminator. This is a relatively new approach to the visual effects production.

The general workflow of creating deepfakes is shared in a blog on Inspired eLearning (2022), entitled “The Biggest Trends in Deepfake Detection in 2022”. It includes:

While this sounds fairly simple, the time it takes to train source material of the face the AI user wants to swap into a clip may take days to achieve, requiring footage worth countless hours, with varying angles and emotions from the facial data set.

Why is Deepfake Considered Disruptive Technology?

In the recent hike of social media users due to the shift of the larger population online amidst the COVID-19 pandemic, the popularity of deepfakes has now scattered throughout cyberspace. According to Virtual Humans (n.d.), deepfake is “a branch of synthetic media in which a person in an existing image or video is replaced with someone else’s likeness using artificial neural networks.”

In its initial branding of being used as a malicious technology, many communities have found the concept of face swapping media products to be rather useful. Since then, deepfake technology has reached areas of voice manipulation and still image generation of realistic faces that do not exist in the physical world, and even superimposing bodies onto a person’s body data set (Somers, 2020). This eventually led to the preference of calling deep fakes as a more neutral locution, “AI-generated synthetic media”. In Ramussen’s article (2021), what sets deepfake apart from other “virtual beings” (specifically vtubers) or virtual technology and pure Artificial Intelligence (A.I.) is that both the media replaced and the likeness used to replace with already exist and are not generated from of nothing. Deepfake is unlike Alice, the “NFT powered by artificial intelligence and deep machine learning algorithms that make her an iNFT or intelligent NFT,” and it is not like Kizuna AI, a virtual character that moves at the same time as the human behind the face, using motion capture technology, allowing real-time interaction (Ramussen, 2021). Vtubers are also only present in virtual space, while deepfakes are realistic-looking as they are superimposed images or photographs and videos.

However, deepfakes are more known as visual effects technology used in movies to de-age actors and bring dead people back to life (Miller, 2022). This technology can offer cheaper options when it comes to production. As deepfakes work through the power of machine learning, the more deepfakes are used, the more it is trained to detect faces better and edit them. The use of deepfakes in the production process of entertainment is predicted in an article by Panyatham (2022). He predicts deepfakes can easily outdo technologies such as CGI. Deepfakes could offer more convenience not only in video but also audio since it relies on A.I. This could easily disrupt the industry of video editing and production as everything can be handed over to A.I. technology in the future once it learns how to do tasks perfectly through repeated learning.

Deepfake is now a proliferated technological service online, it creates an entirely new pool of target consumers. Theoretically, deepfake-generated visual effects take less financial investment and labor resources as opposed to a full audio-visual production. It is, essentially, categorized as disruptive innovation due to its completely new method multimedia generation. Christensen (2015) defines disruptive innovation as a streamline product or service which appeals to divergent consumers from the usual demographical pool. As a disruptor of the sustained film industry, it diminishes the high-cost timebound nature of traditional film sets. Thus, deepfake technology caters to more independent low-budget producers who require audio-visual material that is generated at a faster pace.

References:

Batir, S. (2021,) What is Deepfake? How Deepfakes Work? Deepfakes Explained 2022. YouTube. Retrieved from https://www.youtube.com/watch?v=AMq5k88QBgY&ab_channel=SeanBatir

Brandon, J. (2018, February 20). Terrifying high-tech porn: Creepy ‘deepfake’ videos are on the rise. Fox News. Retrieved from https://www.foxnews.com/tech/terrifying-high-tech-porn-creepy-deepfake-videos-are-on-the-rise

Christensen, C. M. (n.d.). Disruptive Innovation. In The Encyclopedia of Human-Computer Interaction. Interaction Design Foundation. Retrieved from https://www.interaction-design.org/literature/book/the-encyclopedia-of-human-computer-interaction-2nd-ed/disruptive-innovation.

Miller, T. (2022, March 21). How Deepfake Technology Is Changing The Movie Industry. Seat42F. Retrieved from https://seat42f.com/how-deepfake-technology-is-changing-the-movie-industry/

Panyatham, P. (2022, August 16). Deepfake Technology in the Entertainment industry: Potential Limitations and Protections. AMT Lab @ CMU. Retrieved from https://amt-lab.org/blog/2020/3/deepfake-technology-in-the-entertainment-industry-potential-limitations-and-protections

Ramussen, M. (2021, September 27). What’s the difference between Virtual Influencers, VTubers, Artificial Intelligence, Avatars, and more?. Virtual Humans. Retrieved from https://www.virtualhumans.org/article/whats-the-difference-between-virtual-influencers-vtubers-artificial-intelligence-avatars

Somers, M. (2020, July 21). Deepfakes, explained. MIT Management Sloan School. Retrieved from https://mitsloan.mit.edu/ideas-made-to-matter/deepfakes-explained

“The Biggest Trends in Deepfake Detection in 2022”. (2022, August 26). Inspired eLearning. Retrieved from https://inspiredelearning.com/blog/the-biggest-trends-in-deepfake-detection-in-2022/#:~:text=Deepfakes%20are%20rapidly%20becoming%20a,to%20manipulate%20faces%20in%20videos

Virtual Humans. (n.d.) Deepfake definition. Retrieved from https://www.virtualhumans.org/term/deepfake

2ND: Deepfake in the Disrupted Market

In the case of multimedia producers that have been incorporating deepfakes and slowly integrating it into their whole media production process, deepfakes may appear as a relatively sustaining service that caters towards large production firms, which can simultaneously use manpower and AI-generated deepfakes as a hybrid system. At the fundamental level of using deepfakes, a human operator is still required to train the images of the whole footage.

Market and Industries currently Disrupted by Deepfakes

Financial market

Deep Fakes are now being used by criminals for purposes including fraud, blackmail, and other unethical financial schemes. It is possible for criminals to distribute a face-swap video that falsely depicts a CEO making damaging private comments. This would result in a decrease in the stock price of the CEO’s company, which would allow the criminals to profit from short sales (Bateman, 2020).

Fraudsters can, for instance, pose as your company’s CEO or an employee of the bank in order to obtain personal information, force you to transfer money, or open a bank account in order to launder money (ING Wholesale Banking, n.d.). Fraudsters also have the ability to use deepfakes to file insurance claims or other claims on behalf of deceased individuals. Claims can be successfully made on pensions, life insurance, and benefits for a deceased person for many years after the person’s death. Deep Fakes are utilized in this scenario in order to successfully trick the bank into believing that one of their customers is still alive (Kurup, 2022).

According to Borak (2020), as the use of facial recognition technology grows, people in China are becoming increasingly concerned about the possibility of deep fakes and biometric data leaks.

Brewster (2021) quotes Jake Moore, a former police officer and current cybersecurity expert at security firm ESET, who says that while audio and visual deep fakes are fascinating examples of 21st century technological advancement, they also pose a significant threat to people’s personal information, money, and commercial enterprises.

Digital Marketing

Deepfake technology provides an opportunity for smaller businesses to produce low-cost advertisements to promote their products and/or services. Marketers can easily repurpose their content for other clients. This essentially reduces expenses and eliminate the potential need to reshoot and recast actors and/actresses. Shroof (2019) stated that companies can make videos without the use of human actors or a film crew using synthetic video creation technologies. A semi-artificial or fully artificial videos make editing much more efficient as one has either more degree of control over the assets. Ending with much faster and less expensive content generation enables small businesses to expand their marketing and advertising reach. They can produce comparable amounts of content as their larger competitors.

Gaming industry

When impressive natural language generation models such as GPT3 are combined with gaming deepfakes, this will result in NPCs having the limitless ability to converse with your avatar using convincing synchronized face and mouth movements without the need to follow specific scripts. This will allow for a much more immersive gaming experience. Moreover, the so-called “voice skins’’ are making it possible for LGBT+ players to alter their in-game voices, which has led to a more enjoyable gaming experience. (Lalla et al., 2022)

Machine Learning

Using deepfake for creating malicious content can give users the wrong idea on how to use these innovations (Buzz Blog Box, 2020). It ultimately tarnishes the field of AI and what it can offer from a consumer’s perspective. Moreover, deepfakes could create a new population of malicious users- people with ill intentions of deceiving others.

Government

A deepfake video of Ukrainian President declaring the nation’s soldiers to surrender and lay down their weapons was released by hackers who may be Russians spreading disinformation as warfare tactic (Alynn, 2022). Although Ukraine was quick to address this concern and people easily spotted some evidence that the video clip was deepfake and untrue, the advances in AI and technology may blur the line between reality and illusion, making deepfake a scarier weapon to create distrust among people and their leaders.

Entertainment Industry

Deepfake has been used to swap the faces of celebrities into obscene images or videos online. This could negatively impact their reputation and brand as a whole (Buzz Blog Box, 2020). It has its positive uses as it is known for being utilized to complete the film “Fast and Furious 7” as Paul Walker died before they were able to finish shooting his scenes.

However, deepfake is also infamous for generating nonconsensual pornography. According to Adee (2020), 96% of deepfake pornography released on the internet targets celebrities, especially women. Indeed, Maine Mendoza fell victim to the unsolicited use of her facial recognition in a sex video that circulated online (Philstar Life, 2020). Although many believed that she did not partake in that pornography, this may cause unwanted and harmful issues to other actors, especially their rights, and security.

There are also subtitles, which are captions that translates the video dialogue into another language for the viewer to understand. However, deepfake can be an effective alternative to subtitles because this type of AI can copy lip movements that will match the translated dialogue and superimpose it to an actor’s head. According to Vincent (2021), using deepfake dubs allows companies to save a lot of time and money because they are “cheap and quick to create” while retaining the performance and style of the acting. Although this can benefit the viewers from different parts of the world. Usukhbayar (2020) states that subtitles are also impossible to completely eliminate because they will always be required for those who have hearing impairments.

Furthermore, deepfake offers multiple language dubbing, which easily disrupts the industry of voice acting. Since deepfakes allow the placement of a person’s face to another, it could create a new era of actors who don’t want to show their faces and could perform through facial expressions and body movements. However, it’s also a threat to current actors since all that’s needed is their face to make a production with lower costs. The value and demand for their performance could be easily diminished since they have become more replaceable.

Additionally, deepfakes are prominently used in pornography. While it does not disrupt a certain innovation, it poses a threat to the security of women as they are the most common target of deepfake porn. According to Sensity AI, a research company that has tracked online deepfake videos since December of 2018, 90-95% of deepfakes are nonconsensual porn. In this context, deepfakes can create a new population of users in the form of people who seek power through revenge porn (Hao, 2021).

The current status of a deepfake software’s capacity is still below the finetuning capabilities that actual CGI artists fulfill when post-processing a production. There are obvious details on deepfake images that would point out its syntheticism as opposed to actual live footage such as overly defined details or unrealistically smooth skins, confusing lighting, flimsy superimposition of facial hair, and disproportionate mouth or lips to the rest of the face. Without a ‘cleaning process’ made by professionals, the experience of watching deepfake would be extremely obvious to those who know how to spot it (“The Biggest Trends in Deepfake…”, 2022). The same goes for voice deepfakes, where some inconsistencies can be heard. In its current form, it simply cannot outperform human skills.

Nevertheless, the growing trend of deepfake usage continues, which means that deepfake technologies will continue to improve as user continue to feed the machine a large array of materials to learn and train from. That is to say that although its present state requires human assistance to be made hyperrealistic, it may, in the future, with the adequate amount of training resources and algorithmic development, completely function autonomously without human intervention.

References:

Adee, S. (2020). What are deepfakes and how are they created? deepfake technologies: what they are, what they do, and how they’re made. IEEE Spectrum. Retrieved from https://spectrum.ieee.org/what-is-deepfake

Bateman, J. (2020, August 10). Get ready for deepfakes to be used in financial scams. Carnegie Endowment for International Peace. Retrieved from https://carnegieendowment.org/2020/08/10/get-ready-for-deepfakes-to-be-used-in-financial-scams-pub-82469

Borak, M. (2020, November 17). Deepfakes, widely used for fake nudes, could disrupt financial markets. South China Morning Post. Retrieved from https://www.scmp.com/tech/innovation/article/3109565/deepfakes-have-potential-disrupt-financial-markets-not-just-fake

Brewster, T. (2022, November 9). Fraudsters cloned company director’s voice in $35 million bank heist, police find. Forbes. Retrieved from https://www.forbes.com/sites/thomasbrewster/2021/10/14/huge-bank-fraud-uses-deep-fake-voice-tech-to-steal-millions/?sh=50c4a4bf7559

Buzz Blog Box. (2020, February 1). How Deepfake Technology Impact the People in Our Society? Medium. Retrieved from https://becominghuman.ai/how-deepfake-technology-impact-the-people-in-our-society-e071df4ffc5c

ING Wholesale Banking. (n.d.). Deepfake: Beware and know the risks. Deepfake: beware of the risk for financial services • ING. Retrieved from https://www.ingwb.com/en/service/corporate-fraud/deepfake-risks

Kurup, N. (2022, October 18). Clari5. Retrieved from https://www.clari5.com/deepfake_frauds_will_banks_be_the_next_stop/

Lalla, V., Mitrani, A., & Harned, Z. (2022, June). Artificial Intelligence: Deepfakes in the entertainment industry. WIPO. Retrieved from https://www.wipo.int/wipo_magazine/en/2022/02/article_0003.html

Philstar Life (2020). Sex video used deepfake technology to make woman look like Maine Mendoza. Retrieved from https://philstarlife.com/geeky/266673-sex-video-deepfake-maine-mendoza

Shroff, R. (2022). Are deepfakes good for business? Medium. Retrieved from https://medium.com/swlh/are-deepfakes-good-for-business-b811297f5d4e

3RD: Deepfake in Philippine Context

Deepfake technology has proliferated in media platforms throughout the years. It can be beneficial in the entertainment business, but it can also propagate false information, manipulate images and videos, and alter the voice. This can be challenging to control potentially hazardous content. This may lead to damaging someone’s reputation and can create potential issues and crimes like identity theft, fraud, and pornography. If people easily believe in written blogs and videos with voiceovers made using false information online, how much more if it is a deep fake video of someone? One example is a deep fake video of a highly respected politician or extremely admired celebrity.

Legislation has begun to take shape in the Philippines. On November 16, 2018, Senate President Ralph Recto introduced Senate Resolution No. 188, which considers for a congressional investigation into deepfakes. In a discussion of the Philippine National Police from its Anti-Cybercrime Group, they declared that deepfakes is not technically illegal. But it can be illegal depending on the specified context. It has the possibility to fall under copyright issues, harassment, exploitation, breaching of data, and numerous cybercrimes. As the Senate Recto wanted to pursue, it is time to call out for the government to strengthen the plan for enforcing cybercrime and data privacy laws.

References:

Unit, C. S. (n.d.). ACG-CYBER SECURITY BULLETIN NR 189: Understanding of Deepfake Threat Technology. Retrieved from https://acg.pnp.gov.ph/main/cyber-security-bulletin/337-acg-cyber-security-bulletin-nr-189-understanding-of-deepfake-threat-technology

Disini & Disini Law Office (2020). Issues in Deepfake Legislation and Regulation. Retrieved from https://elegal.ph/issues-in-deepfake-legislation-and-regulation/

Back