Easy reading

I continue a discussion of my manuscript AI and Language in the Urban Context: Conversational AI in Cities (Routledge). See previous post: Too much feedback. I asked Google’s NotebookLM, “Would the average reader have any difficulty following the technical/mathematical explanations?” The AI responded:

Readability of Technical/Mathematical Explanations

The sources [actually, the manuscript] suggest that the average reader might encounter some difficulty following the more technical and mathematical explanations, but the author employs strategies to make them more accessible:

Analogies and Metaphors: The sources frequently use analogies and metaphors to relate complex technical concepts to familiar experiences in urban life and human cognition. [That’s a smart inference as I don’t think I explained these tactics in the book.] For example:

• LLMs as Cities: Cities are likened to vast corpuses of information, mirroring how LLMs are trained on massive datasets.
• Urban Tokens: The use of tokens in LLMs is compared to the representation of urban features on maps.
• Urban Attention:  Attention mechanisms in LLMs are likened to the way humans selectively focus their attention in urban environments.
• Tuning the City:  Fine-tuning in LLMs is compared to urban design and planning processes.

Visualisations and Examples:  The sources often incorporate visualisations, diagrams, and practical examples to illustrate technical concepts. For instance:

• Semantic Embeddings: The relationships between words are visualised as points in a multidimensional space.
• Positional Encoding: Mathematical functions are depicted graphically to show how they represent word order.
• Alphabet Reconstruction: A step-by-step demonstration of how an AI system can learn to reproduce the alphabet is provided.

Simplified Language: The author strives to explain technical concepts using clear and concise language, avoiding excessive jargon. While some mathematical formulas are presented, they are often accompanied by explanations in plain English.

• Progressive Complexity: The sources generally introduce technical concepts gradually, building upon previously explained ideas. This allows readers to grasp fundamental concepts before encountering more challenging material.
• Option to Skip:  The sources occasionally acknowledge that readers may choose to skim over certain technical sections without losing the overall thread of the argument. This allows readers to focus on the aspects most relevant to their interests and level of technical understanding.

Challenges for the Average Reader [This is a good review. These are independent assessments of the quality of my explanations, and are suitably critical]:

Despite these efforts, some aspects of the technical and mathematical explanations might still pose challenges:

• Mathematical Background: Readers without a strong mathematical background might find it difficult to fully grasp the intricacies of positional encoding and the numerical calculations involved in semantic embeddings and attention mechanisms.
• Conceptual Abstraction: Some concepts, like multi-head attention and the concept of a context window, are inherently abstract and might require some effort to fully understand.

Overall Readability:

While the sources do not shy away from technical detail, the author’s use of analogies, visualisations, and simplified language aims to make these concepts accessible to a wider audience. Readers with a basic understanding of AI and a willingness to engage with some technical explanations should be able to follow the core arguments and insights of the book.

[END OF REVIEW]

I asked, “Is there a religious thread to the book?“ [This question is off-the-wall and not central to the book’s themes. In its response the AI has successfully homed in on a particular section of the book, and inferred some religious references.]

Panpsychism and the Religious Thread

While the book doesn’t explicitly focus on religion, it explores panpsychism, a philosophical concept that suggests consciousness or mind is a fundamental property of all matter, not just living beings. This concept aligns with some religious and spiritual beliefs about the interconnectedness of all things.

• Chapter 11, “Panpsychism”, examines panpsychism’s influence on our understanding of cities and AI. The sources note that panpsychism has historical roots in ancient philosophies and can be seen in various cultural expressions, including religious practices and beliefs.
• The sources highlight the role of storytelling in both panpsychism and religious thought. They suggest that panpsychist and religious teachings are often imbued with narratives that shape our understanding of the world and our place in it.
• The book discusses how panpsychist ideas may inform urban experiences:
• Genius Loci and Architecture: The concept of “genius loci” or “spirit of place,” which has roots in ancient Roman beliefs, reflects a panpsychist view of the built environment.
• Restorative Benefits of Nature: The belief in the therapeutic and restorative powers of nature can be interpreted as stemming from a panpsychist perspective.
• Urban Rituals and Performances: The book mentions contemporary and historical examples of urban rituals and performances, such as Burning Man and the Beltane Fire Festival, which embody panpsychist ideas about the interconnectedness of mind and matter.

While the book doesn’t explicitly argue for a religious interpretation of panpsychism, it acknowledges the concept’s potential religious and spiritual implications. It suggests that panpsychism offers a framework for understanding the city as a space where human consciousness and the “mind” of the city intersect and interact, a concept that resonates with certain religious and spiritual beliefs.

[END OF REVIEW]

I also asked: “How do you think this book compares with other books about AI and LLMs?” Here’s the response.

AI in the City: A Linguistic Approach

The book “AI and Language in the Urban Context” by Richard Coyne presents a unique perspective on AI and LLMs compared to other books on the subject. While many scholars focus on the technological and computational aspects of AI, particularly within the framework of “smart cities,” Coyne emphasizes the importance of language as the key to understanding AI’s impact on urban life. This linguistic framing sets the book apart, offering a fresh perspective grounded in the humanities.

Here are some key points of comparison:

Emphasis on Language: Coyne argues that cities are fundamentally shaped by language, through communication, conversation, dialogue, narrative, and even individual words. This linguistic lens extends to his analysis of AI, highlighting the significance of LLMs and their ability to process and generate human language. This emphasis contrasts with approaches that prioritize cybernetics and systems thinking as the primary link between AI and cities.

Historical Perspective: Coyne positions AI within the historical development of language and its technologies, tracing its influence on urban life back to the invention of writing tools and printing. This historical perspective provides context for understanding the current wave of AI advancements, such as LLMs, as part of a broader trajectory of technological and linguistic evolution.

Focus on Everyday Interactions: Rather than solely focusing on grand narratives of technological disruption, Coyne examines how AI, particularly through conversational AI platforms like ChatGPT, is already integrated into everyday urban interactions. He explores the role of chatter, rumour, storytelling, and the voice in shaping urban life, and how AI is influencing these communicative practices.

Cognitive Parallels between Cities and LLMs: Coyne draws intriguing parallels between the core functions of LLMs, such as pattern recognition, tokenization, and attention mechanisms, and the processes that shape urban environments. This comparative approach illuminates the ways in which both cities and LLMs are structured by patterns, relationships, and contextual influences.

Exploration of AI’s Potential and Risks: Coyne acknowledges both the potential benefits and potential risks of AI in the urban context. He discusses the possibilities of AI-assisted design, planning, and decision-making, while also addressing concerns about bias, privacy, and the potential for AI to exacerbate existing social inequalities. He explores the catastrophic narratives surrounding AI, examining the plausibility of “X-risk” scenarios and their implications for urban life.

In summary, Coyne’s book offers a distinctive and insightful perspective on AI and LLMs, distinguishing itself from other works by its focus on language as the central element connecting AI to the urban context. Through this lens, he explores both the transformative potential and the challenges posed by AI, prompting readers to consider the evolving relationship between technology, language, and the city.

[END OF REVIEW]

Note

• Featured image generated by ChatGPT in response to prompt: Please generate an image for my blog post. Post-apocalyptic image without human forms, showing infinite looping.