AI from a Christian Worldview

Would you believe the first AI program was developed and presented at a conference at Dartmouth College in 1956? Initially, AI focused on following specific instructions to solve problems, much like teaching a robot to follow a recipe. AI’s computing power progressed, demonstrating its ability to handle complex tasks and process natural language. AI's progressive capabilities became evident when IBM 's Deep Blue defeated world chess champion Garry Kasparov in 1997 and later in 2011 when IBM's Watson won Jeopardy! against top human contestants.

Later, AI made further strides with consumer-oriented applications such as Apple's Siri and Amazon's Alexa, showcasing more advanced and user-friendly uses of the technology.

The next monumental leap came with the release of ChatGPT by OpenAI . This innovation forged new paths in generating human-like text and conversation, setting a new standard for conversational AI.

How Does ChatGPT Work?

Several features of ChatGPT differentiated it from previous AI models when it was first released in June 2020. The primary contributors are the number of parameters and the transformer architecture of its underlying computing capacity. To illustrate this in oversimplified terms, think of ChatGPT as a mathematical formula, like the Pythagorean theorem, A2+B2=C2. This theorem is a formula that contains three variables (i.e., parameters): a, b, and c. ChatGPT's formula, the transformer architecture, is powered by a staggering 175 billion variables. This is a significant advancement from the 100-300 million variables used in most previous AI models. OpenAI has not disclosed the number of variables in its latest version of ChatGPT, but some speculate that the number is in the trillions, similar to several of its newest competitors.

ChatGPT's formula processes what you ask it by analyzing your question one word at a time. It determines your question's context and meaning through complex calculations involving the model's billions or trillions of variables. As it generates a response, it does so one word at a time, recalculating its response with each new word. While it can produce coherent and contextually appropriate responses, it does so without cognitive understanding or true comprehension. Its capabilities are rooted in statistical modeling and pattern recognition, not human-like cognition. This gives ChatGPT's replies the appearance of someone rapidly typing a response, a process that is both impressive and unique to ChatGPT, hence the term Generative AI.

The Future of AI:

Bloomberg reported that OpenAI executives unveiled a new five-level system to track its progress toward achieving Artificial General Intelligence (AGI), a more advanced form of AI. (1) The first level involves chatbots like GPT-4, capable of human-like conversations and multilingual text generation. (2) The second focuses on enhancing AI with advanced reasoning for human-level problem-solving. (3) The third level aims to create autonomous agents that perform tasks independently. (4) The fourth targets innovator AI, which generates new ideas and drives scientific advancements. (5) The final level envisions AI systems managing tasks across entire organizations, optimizing efficiency and versatility. As of July 2024, OpenAI claims they are very close to reaching level two of AGI.

Cause for Concerns

As if we don’t already have questions and concerns about AI capabilities today if any company successfully advances AI technology through all five levels, the ethical and societal implications will be immense. AI raises numerous questions and concerns, but fundamentally, I believe it comes down to three issues:

1.????? Black Box Problem: A common conversation regarding AI is how it formulates its responses. Our inability to know how AI responds is called “the black box problem.” While we understand the questions or prompts we submit to AI, we don’t understand its internal decision-making process. When an AI system produces an incorrect or unexpected response (often called a hallucination), we cannot diagnose or understand why, much less know how to correct it. This is not all that different from human responses. Can we attribute any of our human responses or beliefs to the specific pieces of knowledge or experiences that shaped it? Several organizations, including the U.S. Department of Defense (DARPA) and Elon Musk's company, xAI , are working on addressing this issue by developing “Explainable AI.” However, their efforts are still evolving.

2.????? Bias Infusion Problem: A common and often discussed challenge with AI is the inherent bias and imperfection of the training data used to train the AI. OpenAI, for example, builds its training data from three sources: “(1) information that is publicly available on the internet, (2) information that we license from third parties, and (3) information that our users or human trainers provide.” Subsequently, they apply filters to their training data to remove unwanted information, including hate speech and adult content. The training data and its applied filters contain a diverse range of conflicting perspectives on theological, philosophical, and scientific topics. Open AI acknowledges this reality on its site, where it addresses the topics of truth and bias. So naturally, if its training data is biased and imperfect, its responses may often be as well.

3.?????? Force Multiplier Problem: A force multiplier refers to something that dramatically increases something’s effectiveness. Bill Gates understood technology as a force multiplier when he wrote, “The first rule of any technology used in a business is that automation applied to an efficient operation will magnify the efficiency. The second is that automation applied to an inefficient operation will magnify the inefficiency” (Gates, 1996). I believe the same principle applies to AI. The first rule of any AI is that algorithms applied to a well-curated dataset will amplify its accuracy and fairness. The second is that algorithms applied to a biased or flawed dataset will magnify those biases and errors, potentially leading to more significant and widespread issues. If AI matures through the five levels, we can only wonder how catastrophic it will evolve.

A Christian Perspective on AI

Forbes reports that AI is expected to see an annual growth rate of 37.3% from 2023 to 2030. So, as Christians, what are we to believe about AI and its role in our lives and the lives of our children? We start with our Christian worldview. A worldview consists of beliefs in at least five major areas:? God (Theological), Reality (Metaphysical), Knowledge (Epistemological), Humankind (Anthropological), and Ethics (Moral). The Christian worldview is rooted in the nature and character of God. Reality begins with him and is revealed to us through general revelation (nature) and special revelation (Scriptures). Humankind was created in the image of God, but that image was maligned in the Garden when humankind sinned against God. In His grace, he restored us to Himself through a personal relationship with His Son, Jesus Christ. The basis of how we live, our ethics, is given to us in the inspired and authoritative Word of God (2 Timothy 3:16).

Once we establish a Christian worldview, we can then take this five-faceted template and build a Biblical perspective on Artificial Intelligence:

God: God is ever-present and all-knowing. The development, content, and consumption of AI are fully known to Him. Just as He confounded the construction of the Tower of Babel recorded in Genesis 11, AI's computational capacity will always remain within the bounds of God's sovereign and providential rule over creation. While we should advocate for its ethical use, we need not fear it.

Reality: All reality begins with God and is revealed to us through His self-revelation. Every idea, thought, or attitude expressed by artificial intelligence or experienced in virtual reality is accountable to God in ultimate reality. As we increasingly encounter false realities perpetuated by AI, such as deepfakes, disinformation campaigns, and identity theft, we are presented with greater opportunities to showcase genuine faith and exemplify Christ-like character through virtuous living.

Knowledge: The only reliable and true knowledge is God's, as revealed in the Bible. His knowledge is perfect and complete. AI will never attain this perfect knowledge despite its current and potential advancements. True knowledge begins with the fear of God (Proverbs 1:7), a reverence and understanding that AI can never achieve by its very nature. While AI models will continue to evolve their computational capacity, they will continue to operate within the limits and imperfections of their training data.

Humankind: As God created man in His image, man is attempting to create AI in its image. The challenges data scientists face with perfecting AI attest to the truth that humans are imperfect and sinful. They are flawed and finite beings. AI’s computational output reflects its training data and the people who curated it. AI is a reality with unknown potential in a society without laws governing it; therefore, Christians and their families must exercise great caution in using AI.

Ethics: For Christians, the ethical and moral dilemma is resolved by studying and practicing God's Word. The Bible is our instructions for daily living so that we may experience the fullness of God's will for our lives and our freedom from sin. We must prioritize nurturing our natural critical thinking and intellect. Relying too early on AI and similar technologies can lead to shortcuts or laziness in our personal development. God often presents spontaneous and unexpected moments requiring us to draw from our experiences and intellect without relying on devices or tools. Developing these critical skills is essential to preparing for life's unforeseen opportunities and fulfilling God’s purpose.