David Ortiz
Artificial Intelligence (AI) stands at the forefront of contemporary advancements in technology, reshaping the landscape of various industries and the lives of individuals around the globe. As AI implementation permeates our society, we are compelled to answer essential questions some refuse to ask. It is pivotal for the future of our country to examine the tensions caused by the interplay of AI and the domains it has entered. The pervasive dissemination of AI applications beckons us to consider not only the gains in performance it can provide but also the inherent risks regarding our capacity to achieve human flourishing.
Early History
Artificial Intelligence has been a concept for thousands of years. Although the term was coined initially in 1955 by John McCarthy, artificial intelligence spans as far back as ancient philosophers. The inventors of the time (circa 400 B.C.) created mechanical structures called “automations,” which in Greek means “acting of one’s own will.”[1] While being an old concept, today’s technology holds deeper complexities in its structure. From the 1950s until now, most “AI” tools created were primarily structured systems:
If (user completes action A = complete action B)
If (User doesn’t complete action A = Complete action C)
By 1966, the first chatbot named ELIZA was created, utilizing ordered parameters with assigned weight to specific keywords. It was developed as a mock therapist, aiming to see the validity of human impersonation through convincing interaction. Due to the inflexibility of rule-based AI, the results would vary based on each user’s word choice. It was rudimentary at best, but the possibility of advancement grew in cultural interest with the sci-fi boom and innovation seeking to expand the scope of AI capabilities. ELIZA became the starting point of “deep learning,” where computers create artificial neural networks—this technique teaches computers to process data like a human brain, categorizing information, retaining connections, and enhancing future interactions by leveraging past experiences.[2]
During the peak of the artificial intelligence boom, funding increased for projects exploring new avenues of artificial intelligence. The Japanese government allocated $850 million in 1980 ($2 billion in current value) to create a computer that could translate, converse, and express reasoning on a human level. This led to artificial intelligence becoming the answer to non-numeric data processing, which was forecasted to become the dominant field by the 1990s and beyond.[3] This ushered in a new era of information processing. The current period is marked by fervent development and investment into a theoretical system that can match human intelligence on a broad scale: artificial general intelligence.[4]
Early Fears
After the release of 2001: A Space Odyssey, the cultural perspective on AI was redirected due to the film’s depiction of the robotic assistant HAL 9000—the first portrayal of artificial general intelligence. Robin Murphy, a professor at Texas A&M University, states, “That may be one of the movie’s greatest achievements: it placed AI into the mainstream consciousness even before the first AI robot, Shakey, was completed in 1969.”[5] HAL 9000 was not an automated machine in a factory but one that could reason, converse through natural language, understand the context of its environment, and make decisions, even if it was against the will of a human.
The fear of AI roused by this 50-year-old film, that an AI system could surpass human control, remains the same today. The rebellion of HAL in the movie pressed a crucial ethical question that has been left unanswered: “How much autonomy should we give to an AI, and what should we do when the AI starts recommending things we don’t necessarily like or agree with or may have negative consequences?”[6] Considering the reinvigorated AI development boom, we must tread cautiously.
Moving Forward
Our current period is highlighted by the depth and complexity of deep learning and the availability of information from Big Data (organizations with access to large amounts of data). The exponential increase in useability and functionality makes this period different from those of the past. Most notably, the 2020 beta testing of Open AI’s Generative Pre-Trained Transformer 3 (ChatGPT-3), which can respond to users in real-time, is nearly indistinguishable from text created by humans—even better in some cases. While initially focused on writing tasks, these functionalities have evolved significantly in the past three years. With the advent of GPT-4, users with no prior knowledge can code, create images, and perform various creative or technical writing tasks. Interestingly, GPT technologies have extended past language-based work, with external API integration across platforms. Essentially, this involves onboarding GPT technologies into a currently existing system. For instance, an AI model can assist with predictions of economic fluctuations in the stock market.
Artificial intelligence has made great strides in both ability and application. While there is a lengthy development history, the breakthroughs in AI technology are relatively new, with regulatory bodies struggling to legislate the technology. This is exemplified by prominent AI and tech experts unanimously calling for a six-month halt in development to slow the current tech giant “AI arms race.”[7] The results are too new and are coming too quickly. There is a desperate need to take a step back and construct a slower, more thoughtful approach to development, assess risks, and generate knowledgeable criticism against the drawbacks of freely accessible AI technologies.
Artificial Discipleship
The purpose of Artificial Intelligence’s design is to be an assistant in decision-making. Functions can be utilized for data processing, analysis, situational response, and even higher-level functioning in faith outreach. However, perhaps there are certain domains AI should not be allowed to enter. For example, churches have begun to experiment with AI integration. Some of these include generating devotionals, suggesting sermon topics, writing sermons, generating social media content, and even developing personalized content to connect with individuals on a deeper level.[8] Should we delegate the work God called Christians to do to a system that merely replicates the authenticity of a believer? (Mark 16:15).
While offering efficiency and convenience, AI-generated biblical teaching raises concerns about its fallibility. When John Boyles, an associate professor of Bible at Abilene University, tested AI on its exegetical performance, he states,
It vigorously pursues an interpretation when asked, privileging certain perspectives and sometimes outlawing or excluding others. It does so because the words are a statistical game, not scripture to be lived. It is only parroting what it has been trained on—which is a body of texts that it cannot identify because it seemingly no longer knows what they are (when asked about what sources it used in its interpretation).[9]
AI’s training data encompasses the broad scale of internet access, meaning it weighs in on millions of interpretations, teachings, naysayers, and Bible debunkers. Assertions AI makes on a topic can range from perfect accuracy (usually cases with slight interpretive variation. i.e., history, biology, law) to complete falsity because it has been trained on multiple versions of “truth.”
Moral Direction
The prophet Isaiah states, “Woe unto them that call evil good, and good evil; that put darkness for light, and light for darkness; that put bitter for sweet, and sweet for bitter.” (Isaiah 5:20). How might Christians justify the use of AI if it has a flawed perception of right and wrong? The datasets used to condition AI have been developed by millions, if not billions, of people—ultimately training these programs to have the likeness of humanity. For a system trained on flawed human beings, right and wrong begins to resemble “preference” or “opinions” that differ between individuals not grounded in objective truth.
Throughout history, humanity has attempted to establish subjective and objective incentives for living a moral life. Moral subjectivism, articulated by Jean-Paul Sartre in his works on existentialism, emphasizes individual freedom, personal responsibility, and the idea that “existence precedes essence.”[10] In other words, because only the self is truly known, it is up to the individual to determine the values that ought to be true; no other person can judge such preferences to be better or worse than another. Utilitarianism, which opposes this idea, is based on the metrics of benefit and consequence to the greatest number of people. The Proportionality Doctrine holds that actions are right if they tend to promote happiness and wrong if they produce the reverse, advocating for the assessment of actions to maximize a collective well-being over individual preference.[11] These moral frameworks have limitations. If AI is trained by people who hold to these philosophies, it will be susceptible to the same incomplete image of truth driven into every creature, fueled by sinful nature.
Without a higher moral authority, AI lacks an absolute reference point, rendering it incapable of establishing a valid distinction between true good and evil. The result is a morally adrift populace creating an intelligent program in their likeness, subject to shifting cultural norms, legal prescriptions, and ethical calculations.
Man & Machine
The leaders of our technological industry argue that transhumanism is the next step of evolution, combining man and machine with physical and mental alterations. Jeff Bezos, Elon Musk, and Google co-founder Larry Page are pushing for development for people to “go from human to something beyond human.”[12] With new proposals for human alteration, we risk asking questions that should not be asked. These queries challenge the traditional notion of the self and how much humanity is retained should society impose AI technologies on human bodies. Challenges to the natural order such as these blur the line between innovation and hubris.
Ethical quandaries arise in debates on transhumanism and equity. The prospect of assisting those with debilitating health conditions is a fantastic opportunity, but like many things, good can also unintentionally cause harm. Who will have access to these enhancements? Will they be a luxury merely for the affluent? Technological advancement may exacerbate social inequalities, fostering a divide between the enhanced and unenhanced. In a 2004 article, political theorist Francis Fukuyama called transhumanism “the world’s most dangerous idea…with a frightful mortal cost to human rights.”[13] When looking at the varying health of each country’s economic standing, these disparities may further encourage superior rights against those who do not have the means to attain an “upgrade.”
Humanity Altered
Amidst the qualms about human augmentation, CRISPR technology is equally controversial. CRISPR stands for “clustered regularly interspaced short palindromic repeats”—a natural defense mechanism found in bacteria (Robinson, 2020). The CRISPR system uses a protein called “Cas9,” which targets and cuts DNA. Within these bacteria, the DNA contains specific sequences that act as palindromes—sequences that read the same both forwards and backward. These palindromic sequences flank DNA fragments from viruses previously encountered by the bacterium. Whenever a virus infects the bacterium and survives, some viral DNA is stored between these palindromic sequences. The bacterium then produces small RNA from this stored viral DNA, which attaches to Cas proteins. If the same virus attacks again, the Cas protein guided by the RNA identifies and promptly destroys the invading viral DNA, preventing further infection. What scientists have discovered with CRISPR Cas9 is the ability to use a guide RNA to identify and slice into target DNA, thereby allowing for intentional genome modification to combat diseases and genetic disorders.[14]
With AI entering the mix, studies are beginning to arise on what challenges can be overcome with AI implementation. The newest AI model for CRISPR design, TIGER, is a deep-learning model that seeks to predict both on- and off-target activity. On-target is the focus of alteration and off-targets are unintentional mismatched interactions (partial match), insertions or deletions (indels adjacent to the target site), and non-specific binding (CRISPR binding to unrelated DNA). TIGER has shown unprecedented accuracy in forecasting potential off-targets, significantly surpassing previous CRISPR model accuracy. Researchers at New York University, Columbia University, and the New York Genome Center likened TIGER interactions to a light switch or dimmer knob—turning off specific DNA and RNA activity or tuning them down to reduce activity.[15]
AI has the potential to assist in many different fields, but most cutting-edge scientific research is focused on “fixing” or making people “better,” forcing modern philosophers and political theorists to ask questions “verging on theology.”[16] The power of new biotechnology makes this fact unavoidable, and it is not because of the current risks present with the use of CRISPR, such as off-targets and mutations (disruption of normal cell function). The danger of gene editing spans far beyond individual effects. Its potential to be used to weaponize biotechnology is a global threat. The Defense Advanced Research Projects Agency (DARPA), the Research and Development center of the U.S., is already working on projects to create “genetically enhanced soldiers” in response to China’s similar developments.[17]
Scientists face a serious ethical dilemma. With the possibility of success already established, it seems that the safest option for the U.S. is to push forward and dominate the scientific and technological front. U.S. intelligence officials label gene editing as a weapon capable of mass destruction. Amidst these fears, Swedish philosopher Nick Bostrom argues that a ubiquitous surveillance network under a single world order of preventative policing is the safest compromise to preserve civilization against powerful “black ball” technologies.[18] This omnipresent entity implies global governance, a prospect fraught with danger. Will privacy become a mere facade of national security, and will civil liberty be trampled to prevent an uprising? Will the power of AI technology grow so great that society must be restructured to defend against this new form of technological advancement?
Imitated Thought
Polanyi’s paradox, a theory developed by philosopher Michael Polanyi, states, “human knowledge of how the world functions and capability are, to a large extent, beyond our explicit understanding.”[19] In other words, we can know more than we comprehend. The paradox attempts to explain how we can intuitively understand many tasks yet cannot explain the reasoning or function behind them. Instead, we rely on tacit knowledge, an internalized, unconscious learning from experiences, similar to how we learned our first language.
Ironically, we know very little about how artificial intelligence thinks. Senior researcher Rich Caruana at Microsoft Research notes, “You may think that the principle of artificial intelligence is that we first understand humans, and then build artificial intelligence in the same way, but this is not the case.” The way developers got around this issue is by creating a new mode of thinking. If tacit knowledge and the rules of it cannot be defined (humor, language, emotional intelligence), programming its framework is not possible. The solution has been to create a data-driven neural network (digital brain) that learns tacit knowledge through context and experience.[20]
With this new type of intelligence, how can we distinguish between real and imitated thought? Philosophically, thought is the intentional cognitive process of judging, reasoning, problem-solving, and deliberation—a series of ordered concepts with a clear referent.[21] To be intentional requires the self-awareness to understand the possible outcomes of an action. Artificial intelligence is effective at cause-and-effect dilemmas and determining associated variables. It has the awareness to calculate variables and options but is dependent on the user. AI is limited by its lack of autonomy. It is the operator’s intent that is manifested in the AI.
Let us suppose that AI develops to a point where it has become its own self-conscious entity. Even a perfect replica of outcome would still be distinct from actual human thought. AI uses human intelligence as a component in its own process, synthesizing it into a program with the apparent ability to “understand,” “decide,” and “choose.” AI is not one of us, but it mimics us and knows more about us than we can fathom.
Prey on the Lonely
A continually rising trend in America is the growth of depression among adults between the ages of 18-29 (34.3%) and 30-44 (34.9%). Many have yet to recover from the isolation they experienced during COVID-19, and the rate of depression continues to grow.[22] A Pew Research Center survey conducted in 2018 revealed that approximately one in ten Americans feel isolated and lonely. This was further exacerbated due to Covid-lockdowns and continued poor habits developed in self-isolation.[23] AI technologies have found ways to prey on America’s lonely demographic, offering a replacement for social interaction. These “AI companions” include SimSimi, Kuki, Chai, Snapchat, and Replika (the most advanced of these). Replika tracks personal information—such as names, pets, locations, pictures, and stories—in order to converse about them over time. Furthermore, Replika supports voice calls that allow you to communicate verbally with your AI friend. Dr. Valentina Pitardi, Senior Lecturer in Marketing at the University of Surrey, set out to determine if these friendship apps improve the users’ well-being. She found that the addictive usage of such apps outweighs the benefits, causing continual detriment through overuse. She states,
During the Covid-19 pandemic, people worldwide experienced a never-before-felt level of loneliness. This level of isolation and loneliness continued to rise and is still felt to this day as people continue to experience social exclusion. As a result, many people turn to internet-related coping strategies, such as well-being apps developed to help users with their mental and physical challenges—including friendship and therapy-related well-being apps.[24]
Through a misalignment of goal and outcome, what was meant to bring reprieve to those suffering from loneliness has instead become a replacement that imitates social connection. These AI friends have become customizable—gender and appearance can be chosen, and preset relationship types are available: friend, mentor, or romantic partner. These social companions are designed to fulfill each user’s social need without considering that they could become a detrimental, “close-enough” replacement for those who struggle socially.
Deepfakes And Politics
With the growing capabilities of AI-generated text, audio, and video, there are fears that misinformation may end up turbo-charged. On November 7, 2023, Google and YouTube announced that they will soon require political ads to disclose any form of synthetic imagery and audio clearly. The ban is not directed towards AI-generated ads in general, but to those that skew the claim within the content. For example, Florida GOP Gov. Ron DeSantis has already utilized AI ad technology for his 2024 presidential campaign.[25]However, this top-down approach does little to mitigate the threat of synthetic content. With AI-powered technology, users can easily create “deepfakes.” Deepfake videos can make someone appear to say or do something that never happened. Currently, the Defense Advanced Research Projects Agency (DARPA) is at the forefront of deepfake technology, creating impeccable deepfakes to understand how to detect and prevent its use in attacking America’s democracy. Despite DARPA’s efforts to get ahead of the curve, the quality of deepfake content is already good enough to fool many people. Perhaps real events will be dismissed as fakes due to the public’s fleeting trust in video.[26]
With deepfakes, influential figures are targeted. Yet AI also has the ability to establish a herd mentality in bottom-up campaigns. A term coined as “astroturfing” seeks to produce grass-roots activism artificially. The distinction between grassroots and astroturfing is the amount of support from voters. Modern astroturfing is accomplished through fake social media accounts, blog posts, or letter-writing campaigns to imitate real support, with the intent to manipulate public opinion.[27] Astroturfing is becoming a more viable option for underhanded political tactics with AI’s increasing power and undetectability. For example, in 2017, when the Federal Communications Commission was passing the “net neutrality” bill, millions of messages opposing the bill were found to be fabricated—the nation’s largest broadband companies funded these messages.[28] In the last six years, AI has become more capable of creating unique and diverse responses. Future astroturfing campaigns will only become more successful in swaying public opinion and keeping the truth hidden.
Conclusion
The compounding growth of AI applications throughout the last decade has led to an increased delegation to AI systems. It cannot be denied that artificial intelligence can boost performance in a host of areas. Yet in the wake of progress, there is much that might be lost. As AI continues to permeate society, it is imperative that we engage in thoughtful dialogue and ethical reflection to ensure our technological pursuits align with principles of justice, compassion, and respect for human dignity. Let us draw inspiration from a higher moral authority and seek wisdom in our endeavors to harness the potential of AI, safeguarding the foundation of what it means to be an image bearer of Christ.
[1] “What is the history of Artificial Intelligence (AI)?,” Tableau, accessed March, 2024.
[2] Manisha Salecha, “Story of Eliza, the first chatbot developed in 1966,” October 5, 2016.
[3] Tohru Moto-oka, “Overview to the Fifth Generation Computer System Project,” Computer Architecture News 11, no. 3 (1983): 417–22.
[4] Tableau.
[5] Lauren Fuge, “Fifty years later, scientists reflect on the influence of 2001: A Space Odyssey,” April 30, 2020.
[6] Ibid.
[7] Kevin Roose, “A.I. Poses ‘risk of extinction’ industry leaders warn,” May 30, 2023.
[8] Sarah Long, “Top 10 ways for churches to use Artificial Intelligence,” May 23, 2023.
[9] John Boyles, “Misreading scripture with Artificial Eyes,” July 27, 2023.
[10] Jonathan Crowe, “Is an existentialist ethics possible?” Philosophy Now (2004).
[11] David Brink, “Mill’s Moral and Political Philosophy,” The Stanford Encyclopedia of Philosophy (fall 2022).
[12] Benjamin Stecher, “Transhumanism is the next step in human evolution,” March 16, 2017.
[13] Rene Ostberg, “Transhumanism,” October 12, 2023.
[14] Tara Rodden, “Gene editing,” January 30, 2020.
[15] “AI and CRISPR precisely control gene expression,” NYU, July 3, 2023.
[16] Yelena Biberman, “The ethics and security challenge of gene editing,” Georgetown Journal of International Affairs, July 14, 2023.
[17] Ibid.
[18] Nick Bostrom, “The Vulnerable World Hypothesis,” Global Policy 10. no. 4 (2019).
[19] “Polanyi’s paradox,” Scholarly Community Encyclopedia, November 2, 2022.
[20] “Top 5 philosophical issues of Artificial Intelligence (AI),” Medium, October 29, 2020.
[21] Kenneth Taylor, “The power of thought,” December 9, 2010.
[22] Dan Witters, “U.S. depression rates reach new highs,” Gallup, September 14, 2023.
[23] “Understanding the effects of social isolation on Mental Health,” Tulane University School of Public Health, December 8, 2020.
[24] Witters.
[25] “A.I.-generated voices, images in political ads to require disclosure on Google and YouTube,” CBS News, September 7, 2023.
[26] “Inside the Pentagon’s race against Deepfake Videos,” CNN, accessed March, 2024.
[27] “Astroturfing,” Political Dictionary, September 4, 2023.
[28] Matt Shuham, “American politics isn’t ready for these robots,” June 22, 2023.