We welcome you to our AI sessions. We strongly recommend having a basic understanding of AI before attending as the experts will assume foundational AI knowledge. Below is a glossary of AI terms:

AI Basics

• Algorithm: a step-by-step procedure or formula for solving a problem or performing a task (Cormen et al., 2022).
• Learning: a computer program is said to learn with respect to a set of tasks if its performance at these tasks, measured by a quantitative performance metric, improves with experience (Mitchell, 1997).
• Artificial Intelligence (AI): a field of research in computer science that develops and studies methods and software that enable machines to perceive their environment and use learning and intelligence to take actions that maximize their chances of achieving defined goals (Russell & Norvig, 2021).
• Machine Learning (ML): a set of methods within the discipline of AI that can automatically detect patterns in data, and then use the uncovered patterns to predict future data, or to perform other kinds of decision making under uncertainty (Murphy, 2012).
• Supervised Learning (SL): a type of ML where a model is trained from labelled examples, so the model can learn the relationship between inputs and outputs (Hastie et al., 2009).
• Unsupervised Learning (UL): a type of ML where a model finds patterns in data without being given correct answers. The system must identify structures within the data on its own (Hastie et al., 2009).
• Deep Learning (DL): a set of methods within the discipline of ML that focuses on utilizing multilayered neural networks (Goodfellow et al., 2016).
• Natural Language Processing (NLP): a field of AI focused on enabling computers to understand, interpret, and generate human language (Jurafsky & Martin, 2008).
• Computer Vision (CV): a field of AI focused on enabling computers to understand, interpret, and generate visual information (that is, images and video). Example CV tasks include identifying objects, recognizing faces, reading text, and understanding scenes (Szeliski, 2010).
• Reinforcement Learning (RL): a discipline within ML that focuses on decision making by autonomous agents (Murel, 2024).
• Autonomous Agent: in the context of AI (RL in particular), any system that can make decisions and act in response to its environment without direct instruction by a human user (Murel, 2024).
• Overfitting: a common problem in ML where a model adapts to training data too much, so that noise and random fluctuations are modelled alongside underlying relationships. This leads to poor outcomes on previously unseen data (Goodfellow et al., 2016).
• Bias-Variance Tradeoff: an important concept in ML that describes the balance between the ability of a model to perform well on training data (decreased bias) and its ability to generalize to unseen data (decreased variance). Negotiating this balance is a crucial aspect of designing ML models (Hastie et al., 2009).

Deep Learning

• Neural Network: A system loosely inspired by biological neural networks that consists of interconnected nodes (called neurons) that process and synthesize information. Each neuron has a collection of weights that determines how much influence that neuron has on other neurons. Neural networks can pick up on patterns in data by adjusting these weights (Goodfellow et al., 2016).
• Backpropagation: The primary learning algorithm that adjusts the weights of a neural network to improve their performance (Rumelhart et al., 1986).
• Zero-Shot Learning: The ability of an AI system to perform a task it has never been specifically trained on. For example, a model trained on general language may be able to write a haiku, even though it has not seen an example of a haiku before (Palatucci et al., 2009).
• Few-Shot Learning: The ability of an AI system to quickly adapt to new tasks after seeing a handful of examples. For example, a model trained to recognize dog breeds could acquire the ability to recognize a new dog breed from just a few images (Finn et al., 2017).
• Transfer Learning: A technique where information gained by a model from solving one problem is applied to a different but related problem. For example, a model trained to classify dog breeds could be transferred to classify cat breeds (Pan & Yang, 2009).
• Fine-Tuning: A type of transfer learning where a previously trained AI model is slightly adjusted to specialize in a particular task (Howard & Ruder, 2018).
• Attention Mechanism: a technique which takes items in a sequence and determines how strongly inter-related these items are to one another (Vaswani et al., 2017).
• Transformer: A type of neural network architecture that uses attention mechanisms to process large sequences of information simultaneously and understand relationships between distant elements (Vaswani et al., 2017).
• Neural Machine Translation (NMT): An AI model that can translate text from one language to another using neural networks (Bahdanau et al., 2014).

Generative AI

• Generative AI (GenAI): AI models (unsupervised learners) trained to create new content that resembles existing training data (Bond-Taylor et al., 2021).
• Large Language Model (LLM): An AI model trained on large quantities of text data to understand and generate human-like language. Examples include ChatGPT (OpenAI), Claude (Anthropic), Llama (Meta), and Qwen (Alibaba) (Radford et al., 2019).
• Context: The information (text, images, video, audio, etc.) that a generative model can directly leverage to generate a response. For example, during the course of a conversation, an LLM can leverage the context of earlier dialog (Radford et al., 2019).
• In-Context Learning: The ability of LLMs to learn and adapt to new tasks by being given examples within the conversation (and no additional training). The model uses the context of the conversation to respond appropriately (Brown et al., 2020).
• General Pre-trained Transformer (GPT): A specific type of LLM that uses the transformer architecture and is trained to perform a variety of language tasks (Radford et al., 2018).
• Diffusion Model: A type of GenAI model that generates new content by starting with random noise and gradually refining it into a sensible output. These models are commonly used for text-to-image generation tasks (Sohl-Dickstein et al., 2015).
• Reinforcement Learning from Human Feedback (RLHF): A RL-based training method where AI models learn to adapt to human preferences by receiving feedback from their interactions (Ouyang et al., 2022).
• Retrieval Augmented Generation (RAG): An approach that combines text generation with information retrieval. Based on a prompt, the system first searches for relevant information from a database (like a Google search). Based on the prompt and the retrieved search information, the system generates a textual response (Lewis et al., 2020).

Cyber Infrastructure and Protocols

• Data Center: A large facility housing thousands of computer servers that provide the computational power required to maintain modern cyber infrastructure, including training and running AI models. Many centers contain specialized hardware optimized for AI workloads. Data centers require significant amounts of energy to power the intensive computations being performed (Patterson et al., 2021).
• Application Programming Interface (API): A set of protocols that allows different software applications to communicate with each other (Fielding, 2000).
• Model Context Protocol (MCP): A protocol developed by Anthropic that allows AI models to share information between one other and external systems (Anthropic, 2024).
• Neural Scaling Laws: Trends that describe how AI model performance improves as you increase model size, amount of training data, or the computational resources used for training. These laws help predict how AI systems will perform under various constraints (Kaplan et al., 2020).
• Federated Learning (FL): A distributed learning approach where AI models are trained across many servers without centralizing the dataset (McMahan et al., 2017).

Long-Term Concerns

• Transhumanism: A movement that advocates using new and future technologies to improve human longevity, cognition, and well-being (Bostrom, 2005).
• Artificial General Intelligence (AGI): A hypothetical AI system that possesses human-level intelligence across all domains (Russell & Norvig, 2021).
• Artificial Super Intelligence (ASI): A hypothetical AI system that surpasses human-level intelligence across all domains, potentially by orders of magnitude (Bostrom, 2005).
• Singularity: A hypothetical point in time AI development (or technology development more generally) accelerates beyond human comprehension or control (Kurzweil, 2006).
• P(Doom): The estimated probability of AI posing an existential threat to humanity. This estimate varies widely among experts: Max Tegmark (co-founder, Future of Life Institute) places the probability at >90% (Tegmark, 2025), while Richard Sutton (2025 Turing Award winner) places the probability at 0% (Sutton, 2025).
• Sentience: In the context of AI, sentience implies a genuine, internal sense of self and awareness, not just the ability to process information and mimic human responses. In general, sentience is not a goal of AI researchers (Russell & Norvig, 2021). However, it is frequently debated whether an AI system could attain some form of sentience.
• Agency: the capacity of an AI system to act independently, make decisions, and perform tasks based on its environment and objectives (Sutton & Barto, 2018).

Two Christian Calls for AI and Ethics

• The Lausanne Call for AI and Digital Ethics: A recent initiative within the Lausanne Movement—a global network of evangelical leaders and scholars committed to world evangelization—that seeks to articulate a distinctly Christian response to artificial intelligence and emerging technologies. Building on earlier Lausanne commitments (The Lausanne Covenant , Manila Manifesto , and Cape Town Commitment ), the Lausanne Call affirms that technological innovation must serve human dignity, truth, justice, and the proclamation of the gospel. It emphasizes the responsible use of digital technologies for the common good, warns against dehumanizing or idolatrous uses of AI, and invites Christians to participate thoughtfully and ethically in shaping the digital future.

• The Rome Call for AI and Ethics: is a global initiative launched to promote the ethical development and use of artificial intelligence (AI), grounded in human dignity and the common good. It was first signed on February 28, 2020, at the Pontifical Academy for Life in Rome, and has since gained international traction with support from religious leaders, tech companies, and governments.

References

Anthropic. (2024). Model Context Protocol.

Bahdanau, D., Cho, K., & Bengio, Y. (2014). Neural machine translation by jointly learning to align and translate. arXiv preprint, arXiv:1409.0473. https://www.shpylgoreih.fr/documents/Bahdanau_Neural_Machine_Translation_by_Jointly_Learning_to_Align_and_Translate_2015.pdf

Bond-Taylor, S., Leach, A., Long, Y., & Willcocks, C. (2021). Deep generative modelling: A comparative review of vaes, gans, normalizing flows, energy-based and autoregressive models. IEEE transactions on pattern analysis and machine intelligence, 44(11), 7327-7347. https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=9555209

Bostrom, N. (2005). A history of transhumanist thought. Journal of evolution and technology, 14(1). https://ora.ox.ac.uk/objects/uuid:55ab57ec-70d0-4b93-b058 0d7f57167cc2/files/me7362f022f645636ed10948d03a7bfab

Bostrom, N. (2014). Superintelligence: Paths, Dangers, Strategies. Oxford University Press.

Brown, T., Mann, B., Ryder, N., Subbiah, M., Kaplan, J., Dhariwal, P., & Neelakantan, A. (2020). Language models are few-shot learners. Advances in neural information processing systems, 33, 1877-1901. https://proceedings.neurips.cc/paper_files/paper/2020/file/1457c0d6bfcb4967418bfb8ac142f64a-Paper.pdf

Cormen, T. H., Leiserson, C. E., Rivest, R. L., & Stein, C. (2022). Introduction to Algorithms, Fourth Edition. MIT Press.

Fielding, R. T. (2000). Architectural styles and the design of network-based software architectures. University of California, Irvine. https://www.proquest.com/openview/fc2d064044b971dda476dfb429a2b344/1?pq-origsite=gscholar&cbl=18750&diss=y

Finn, C., Abbeel, P., & Levine, S. (2017). Model-agnostic meta-learning for fast adaptation of deep networks. International conference on machine learning, (PMLR). https://proceedings.mlr.press/v70/finn17a/finn17a.pdf

Goodfellow, I., Bengio, Y., & Courville, A. (2016). Deep Learning. MIT Press.

Hastie, T., Tibshirani, R., & Friedman, J. H. (2009). The Elements of Statistical Learning: Data Mining, Inference, and Prediction, Second Edition. Springer.

Howard, J., & Ruder, S. (2018). Universal language model fine-tuning for text classification. arXiv preprint, (arXiv:1801.06146). https://arxiv.org/pdf/1801.06146

Jurafsky, D., & Martin, J. H. (2008). Speech and Language Processing, 2nd Edition. Pearson Prentice Hall.

Kaplan, J., McCandlish, S., Henighan, T., Brown, T., Chess, B., Child, R., Gray, S., Radford, A., Wu, J., & Amodei, D. (2020). Scaling laws for neural language models. arXiv preprint, arXiv:2001.08361.

Kurzweil, R. (2006). The Singularity Is Near: When Humans Transcend Biology. Penguin Publishing Group.

Lewis, P., Perez, E., Piktus, A., Petroni, F., Karpukhin, V., Goyal, N., & Küttler, H. (2020). Retrieval-augmented generation for knowledge-intensive nlp tasks. Advances in neural information processing systems, 33, 9459-9474.

McMahan, B., Moore, E., Ramage, D., Hampson, S., & Aguera y Arcas, B. (2017). Communication-efficient learning of deep networks from decentralized data. Artificial intelligence and statistics, (PMLR), 1273-1282. https://proceedings.mlr.press/v54/mcmahan17a/mcmahan17a.pdf

Mitchell, T. M. (1997). Machine Learning. McGraw-Hill Education.

Murel, J. (2024, March 25). What is reinforcement learning? IBM. Retrieved August 11, 2025, from https://www.ibm.com/think/topics/reinforcement-learning

Murphy, K. P. (2012). Machine Learning: A Probabilistic Perspective. MIT Press.

Ouyang, L., Wu, J., Jiang, X., Almeida, D., Wainwright, C., Mishkin, P., & Zhang, C. (2022). Training language models to follow instructions with human feedback. Advances in neural information processing systems, 35, 27730-27744. https://proceedings.neurips.cc/paper_files/paper/2022/file/b1efde53be364a73914f58805a001731-Paper-Conference.pdf

Palatucci, M., Pomerleau, D., Hinton, G., & Mitchell, T. (2009). Zero-shot learning with semantic output codes. dvances in neural information processing systems, 22. https://proceedings.neurips.cc/paper_files/paper/2009/file/1543843a4723ed2ab08e18053ae6dc5b-Paper.pdf

Pan, S. J., & Yang, Q. (2009). A survey on transfer learning. IEEE Transactions on knowledge and data engineering, 22(10), 1345-1359. https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=5288526

Patterson, D., Gonzalez, J., Le, Q., Liang, C., Munguia, L.-M., Rothchild, D., So, D., Texier, M., & Dean, J. (2021). Carbon emissions and large neural network training. arXiv preprint, arXiv:2104.10350. https://www.kathimerini.gr/wp-content/uploads/2024/07/2104-10350_1.pdf

Radford, A., Narasimhan, K., Salimans, T., & Sutskever, I. (2018). Improving language understanding by generative pre-training. https://www.mikecaptain.com/resources/pdf/GPT-1.pdf

Radford, A., Wu, J., Child, R., Luan, D., Amodei, D., & Sutskever, I. (2019). Language models are unsupervised multitask learners. OpenAI Blog, 1(8), 9. https://storage.prod.researchhub.com/uploads/papers/2020/06/01/language-models.pdf

Rumelhart, D., Hinton, G., & Williams, R. (1986). Learning representations by back-propagating errors. Nature, 323(6088), 533-536. https://www.nature.com/articles/323533a0.pdf

Russell, S. J., & Norvig, P. (2021). Artificial Intelligence: A Modern Approach. Pearson.

Sohl-Dickstein, J., Weiss, E., Maheswaranathan, N., & Ganguli, S. (2015). Deep unsupervised learning using nonequilibrium thermodynamics. International conference on machine learning, (PMLR), 2256-2265.

Sutton, R. (2025, June 6). NUS120 Distinguished Speaker Series. YouTube. https://www.youtube.com/watch?v=f9KDMFZqu_Y

Sutton, R. S., & Barto, A. G. (2018). Reinforcement Learning: An Introduction. MIT Press.

Szeliski, R. (2010). Computer Vision: Algorithms and Applications. Springer.

Tegmark, M. (2025). My P(doom) Estimate. X. Retrieved August 11, 2025, from https://x.com/tegmark/status/1917580821101437280

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. Advances in neural information processing systems, 30. https://proceedings.neurips.cc/paper/2017/file/3f5ee243547dee91fbd053c1c4a845aa-Paper.pdf