On May 22, a fake Twitter account called “Bloomberg Feed” posted a photo that suggested a large explosion in the Department of Defence’s Pentagon complex in Virginia, the United States.
While the account had less than 1,000 followers and the photo was quickly debunked by the local police department as an AI-generated image, the misinformation had already been spread by larger accounts and reportedly caused a small dip of 0.26% in the stock market before it quickly bounced back.
Previously in March, AI-generated images of Pope Francis in a white boutique coat were spread across social media and were indistinguishable from normal photographs.
This pushed tech publications to create detailed analyses that breaks down evidence of AI manipulation, but not before it successfully enforced fringe beliefs about the papacy that had already circulated among users on the Internet.
These two events demonstrate the latest capabilities of emerging AI models that specialise in generating text or images.
The International Telecommunications Union, for one, has proclaimed the “death of authenticity” over how easily synthetic media produced by AI can spread online.
While they remain an impressive feat of engineering, generative AI models now clearly present risks whereby malicious actors can spread misinformation in a highly efficient manner.
This is a problem unique to AI as it is generally developed to reduce the costs associated with labour-intensive processes, such as copy-writing and graphic design.
It can be made realistic enough to fool regular users and perhaps sway an election, which in Indonesia will take place in about seven months.
For Indonesia, where bookstores are closing operations because of lack of interest, this will inadvertently exacerbate the national issue of literacy.
Already, we are placed among countries with the lowest literacy skills worldwide.
A 2015 Organization for Economic Cooperation and Development (OECD) report found that only 5.4% of adults in Jakarta had attained an acceptable level of literacy, and almost 70% were only able to process information with basic vocabulary knowledge.
A 2016 survey by the Central Connecticut State University that ranks the world’s “most literate nations” placed Indonesia in the second-lowest country in terms of reading interests.
This is confirmed by the Education Ministry’s reading index (Alibaca) which shows that, on a 100-point scale, the nation’s access to books ranks at only 23.1 while the reading culture sits at 28.5.
At a more granular level, gender issues have also contributed to the problem, with the OECD report showing that women scored significantly lower than men in both literacy and numeracy.
Additionally, high levels of literacy are shown to have continuity across generations, which also suggests a strong link between social units such as families and the maintenance of reading ability.
It must be emphasised that literacy rates are not the same as literacy skills, as the ability to read words does not automatically correlate to actually understand what is written.
Instead, being able to read without clearly understanding them may actually worsen the situation, as it presents an environment where AI-generated information can be quickly disseminated with few verifiability controls.
The irony, despite the lack of interest in reading, is that Indonesians are highly active Internet users.
World Bank data shows that 62% of the population had access to the Internet as of 2021.
Analysis by data firm Kepios shows a 5.2% increase of new Internet users between 2022 and 2023, or an increase of around 10 million people.
Furthermore, Indonesians are generally more exposed to AI misinformation, as the country ranks among the largest social media markets in the world.
Social media firm Digital Business Lab reports that there were 191.5 million social media users as of January 2022, equivalent to 68.9% of the total population.
Media heavy platforms such as WhatsApp, Instagram and TikTok rank as the most popular platforms among Indonesians.
We need to understand that the proliferation of generative AI models may actually amplify human errors and prejudice in decision-making.
The necessity of having to verify each new piece of information can take longer than the speed in which misinformation can entrench itself in the public consciousness.
For a country seeking to industrialise and prove to the world that it is ready to become equal among peers, this is an unacceptable risk to its economic and sociopolitical landscape.
Indeed, governments worldwide have acknowledged the “ethical-legal gap” between the spread of new technological innovations and the creation of new policies that regulate them.
There are already technical solutions available to detect AI-generated content, but they are typically playing catch-up with successive new developments of better AI models.
Therefore, one solution to combat misinformation caused by AI manipulation relies on a two-pronged approach rooted in human agency: The government must begin to invest in better access to sources of credible information, and civil society must build a culture of reading among active Internet users in Indonesia.
The government can take note of the European Union’s AI Act, a proposed law intended to assign AI applications to different categories of risk.
By adapting the law to Indonesia’s more information-rich environment, the government could create an authorised list of credible sources that interested parties could access with clear differentiation of AI use.
On a more grassroots level, local governments can elect to reform existing facilities, such as creating 24-hour town libraries with little bureaucratic barriers that would entice people to start reading for free.
A long-term implementation of such could see the development of reading habits in civil society as a self-motivated individual endeavour.
AI may very well be an unavoidable fact of life in the near future, requiring users to become savvier in consuming every piece of content.
Indonesia must prepare by addressing the most basic issue of literacy to mitigate any potential risks before it opens up the country to more AI use. — The Jakarta Post/ANN