The Battle for Democracy: Turkish Elections, Deepfakes, and the Threat to Press Freedom.

"The Demagogue"

 José Clemente Orozco, 1946.

Public Domain.

via Wikimedia Commons.

In an article today by Ben Hubbard and Gulsin Harman from The New York Times, the recent Turkish presidential elections are reported to be heading into a runoff. After failing to secure a majority, the incumbent president Recep Tayyip Erdoğan will face opposition candidate Kemal Kılıçdaroğlu in a two-week battle that could redefine Turkey's political landscape. Erdoğan, having served as Turkey's dominant political figure for 20 years, has seen a decline in his standing due to economic issues and his consolidation of power. Meanwhile, Kılıçdaroğlu represents a coalition of six opposition parties, campaigning as the antithesis of Erdoğan.

Meanwhile, David Klepper and Ali Swenson of The Associated Press this morning discuss the potential political risks posed by advanced artificial intelligence (AI) tools leading up to the 2024 elections. The availability of inexpensive and increasingly sophisticated AI technology has heightened concerns about the creation of convincing deepfakes, including fake images, videos, and audio. These deepfakes, in conjunction with social media algorithms, have the potential to spread misinformation quickly and target specific audiences, possibly leading to new levels of election manipulation. Experts like A.J. Nash from the cybersecurity firm ZeroFox warn that society is unprepared for the impact of these AI-driven media manipulations, which could be used to confuse voters, slander candidates, or even incite violence. As we approach the 2024 elections, there are increasing calls for legislative measures to combat the potential misuse of AI in political campaigns. 

That may very well be a move too far. Daren Butler of Reuters, in an October 13, 2022 article, noted Erdoğan's use of misinformation and its impact on Turkey's democracy. The article refers to a controversial law enacted by the Turkish parliament, proposed by Erdoğan, which can jail journalists and social media users for up to three years if found guilty of spreading "disinformation". Lawmakers from Erdoğan's ruling AK Party and its nationalist allies MHP, who collectively had a majority in the parliament, approved the bill despite strong opposition from both national and international critics. These critics argue that the law's vague definition of "false or misleading information" could be exploited to suppress dissent.

Engin Altay, a member of the main opposition Republican People's Party, expressed concerns that the law will further limit press freedom in a country already lagging behind in this area. The AK Party, without shame, claims that the legislation is necessary to combat misinformation and false accusations on social media, asserting it will not be used to suppress opposition voices. The fairly clear intent of the legislation is to control narratives and limit critical voices, thereby influencing public opinion. Reuters' investigation highlighted the state of media freedom in Turkey, describing mainstream media as adhering to government-approved headlines while independent and opposition media face penalties. The Venice Commission, an advisory body to the Council of Europe, expressed concerns that this law could induce a "chilling effect and increased self-censorship" ahead of the elections.

The proliferation of misinformation, further magnified by artificial intelligence, presents a considerable challenge to democratic values, underlining the indispensable role of press freedom. Drawing parallels with Kléōn's emotional manipulation in the Mytilenean Debate, contemporary misinformation leverages public sentiment, especially anger, to influence portions of the electorate. However, Aristotélēs maintained that the robustness of democracy resides in the collective wisdom of its populace. This collective wisdom, cultivated from a variety of perspectives and experiences, tends to yield clear choices, exemplifying the tenacity of democratic systems. There's an expectation that the probabilities underpinning the Condorcet Jury Theorem have remained (mostly) unaffected by misinformation. However, it's almost certain that augmenting governmental control over information could lead to a detrimental shift. Other than ongoing voter education, few options exist to ensure that clear choices will still surface in the aggregate. The freedom of the press remains a key factor in assuring that the electorate is well-informed and capable of exercising its collective wisdom, thereby reinforcing the effective functioning of democratic systems, despite the emergence of AI-powered misinformation or even more worryingly, justifications for state control of the press based on such grounds.

AI Transforming Self-Perception.

Sigmund Freud, c. 1921.
By Max Halberstadt.
Public Domain.
via Wikimedia Commons.

The Economist's essay this week discusses the rapid advancements in artificial intelligence (AI), particularly large language models (LLMs) like GPT-4. The rise of LLMs has sparked an entrepreneurial explosion, altering how people use computers and creating new business models. However, this development also presents risks, such as spreading disinformation and creating deep fakes.

The essay draws comparisons between LLMs and pivotal moments in history, such as the introduction of the web browser, the printing press, and psychoanalysis, each of which significantly altered how people access knowledge and perceive themselves. As LLMs evolve, they could fundamentally shift the way we engage with information and each other. Some researchers even envision AI entities developing unique personalities, becoming externalized versions of users' inner voices or emulating the personalities of deceased individuals.

The true nature of AI models remains a contentious issue among researchers. Some argue that these models have no real understanding and merely parrot patterns from training data ("pseudocognition"), while others believe the models possess abilities that cannot be distinguished from genuine understanding. This debate echoes Freud's concept of the uncanny, and may influence how people perceive themselves, potentially reinforcing the idea that humans are not masters of their own existence.

There are further drawbacks to the rise of LLMs. They are capable of generating plausible but false information, a phenomenon known as "hallucination" or "confabulation," raising concerns about the potential for spreading disinformation, deep fakes, and fabricated content. This challenges the integrity of public debate and highlights the need to address the negative implications of AI-generated content while leveraging its potential benefits.

To address the implications of LLMs, the article emphasizes the importance of considering AI ethics, including unconscious biases in training, the responsibilities of AI creators, and the regulation of AI upbringing. It calls for a thorough examination of human desires and motivations in relation to LLM development and the potential societal impact. As AI continues to evolve, society must prepare for both the positive and negative consequences.