Although not all uses of deepfakes are malicious, some of this content is directed towards the path of extortion or scam. What do experts warn?
Although the scope of Artificial Intelligence (AI) is unlimited, its ethical margins are not. Even the Organization for Economic Cooperation and Development (OECD) proposes regulations on use in favor of citizens. It thus proposes applying the principle of transparency and monitoring equity and privacy. However, there are already threats to the responsible management of the technological tool: deepfake is one of them.
Deepfakes are videos, images or audio that aim to imitate the appearance and voice of a person. Initially, this type of content has educational or entertainment purposes; But a malicious tinge in its use has placed SMEs in a risky field. Why?
DEEPFAKE DYNAMICS
Deepfakes are produced with adversarial generative networks, which are called GANs. "There is an intelligence that is responsible for creating false content contrasted with an intelligence that is responsible for detecting it," explains Javier Escurra, project engineer at Hytera Perú. In short, one of the GANs is the generator and the other is the discriminator that evaluates the results and distinguishes them from real data.
This process is repeated until convincing content is produced, so much so that the discriminator can no longer identify it as false. To facilitate interpretation, Escurra proposes an example: a teacher asks his student to paint a first version of the Mona Lisa. After delivery, the teacher detects disjointed strokes and asks for a second sample. Then a third, a fourth and a fifth installment arrive. In this way, the student is trained so that the false work is unrecognizable.
The specialist lists some methods of locating incorrect use of deepfake: gestures not synchronized with the voice, robotic sounds or slightly slow cadences. "AI is not natural, it usually sustains monotonous noises," he says.
DEEPFAKE AND SCAMS: THE GREATEST RISK
Deepfake fraud allows a potential scammer to impersonate someone else, usually someone the victim knows: a boss, the company's CEO, or another trusted figure. In that sense, Javier Kuan, director of digital technology strategy at Esencia, of the Efitec Corporation group, describes it as a significant threat to all types of governments and industries, but "particularly for micro and medium-sized companies."
"For them it is very sensitive not only because of the nature of informality that we have locally, but also because of the economic factor: SMEs do not have the technological budget of the financial sector in Peru. There is also a lack of specialized personnel and an II area (information technology) that is very reduced or absent," he details.
In this regard, in a large company, which has a hiring of technical personnel, "the processes are well established or much more structured", while "SMEs suffer from more formal processes." This occurs in the midst of the effort that small and medium-sized companies make to take care of their budget and work on the side of formality: they consider, in general, technological care as an expense and not so much as an investment.
Lagging behind technology in a company's progress plan, regardless of its size, is a concern that the National Center for Strategic Planning (Ceplan) also includes in its report Artificial intelligence: challenges and opportunities for Peru.
"The most advanced countries, but also other countries in the region, are investing in this technology, because there is evidence of great improvements in productivity and innovative solutions for industries, but also for the services and daily life of the population. Peru cannot be left behind [...] Otherwise, there is a risk of, once again, being left on the periphery of technology, innovation, productivity and well-being, he states.
AI AS AN ALLY
To balance the balance and turn AI into an ally, Kuan believes that awareness must be reinforced, because, according to his estimates, by 2030, 90% of audiovisual content would be deepfake. Given this scenario, it is important that the user recognizes what generative AI can do, both for good and bad.
Along these lines, he maintains that SMEs "should address this under a type of simple mechanism so that, through a key or password, they can verify that the audio, video or approval of a deposit are really authentic and issued by the person to whom it corresponds,"
However, he points out that, as AI registers important changes every three months, the work belongs to SMEs, but also to governments, chambers of commerce and leading companies in the industry. A 'cross' in the challenge.
PERU AGAINST THE BAD PRACTICES OF AI
The National Center for Strategic Planning prepared the report Artificial intelligence: challenges and opportunities for Peru. This document highlights that the key to Artificial Intelligence lies in its condition as a technology that has a transformative impact at an economic and social level.
"On the economic side, estimates of its contribution to macroeconomic measures are presented, such as the Gross Domestic Product (GDP) and the Gross Value Added (GVA)," it can be read. For this reason, Ceplan considers that correct use favors increased productivity and job automation.
In this sense, it recommends working on several fronts: human talent and training of specialists, investment in digital infrastructure, closing gaps with a territorial approach, opening entities' data for greater transparency, generation of technological business ecosystems, strengthening the legal framework of intellectual property or, among others, the disclosure of the potential for its incorrect use.
