I’m just back from a holiday in Malta. Two AI events entered my awareness. The first was meeting an American language teacher who said she had used AI to find her a “charming B&B” just outside Valletta. The second event was news that Elon Musk had sent an email to all US Federal employees to tell him what they had done over the past week. Commentators assumed that he would use AI software to analysis and evaluate the responses (NBC News).
That’s two applications of AI: one benign and helpful; the second somewhat sinister. But each reinforces just how commonplace AI is becoming in practice and in the public discourse, at least from a consumer perspective. We regular consumers can also use AI (ChatGPT) to gather data, at scale.
That’s an image I took of a facade near the entrance to Valletta. Following on from my last post I asked ChatGPT to “inventorize” the image: “Please provide a tabulated inventory of the architectural elements in this facade, with headings such as classification, element, colour, material, size, location, condition.” I also asked for quantities.
This would require some editing, especially of the quantities, but it’s a good start.
I asked: “Please tell me the reasoning thread by which you produced the inventory table.”
ChatGPT gave a considered response, as follows …
The reasoning behind producing the architectural elements inventory with estimated quantities followed a structured observational and analytical process based on the image. Here’s the breakdown of my approach:
1. Categorization of Elements
First, I classified the visible architectural elements into logical categories based on architectural typologies:
- Structural – The core built components (e.g., facade).
- Openings – Doors, windows, balconies.
- Ornamentation – Decorative elements like cornices, pilasters, and reliefs.
- Commercial – Shopfronts and commercial adaptations.
- Signage – Functional textual or branding elements.
This ensures each element is described in terms of architectural function and aesthetic contribution.
2. Identifying Key Architectural Features
By closely examining the facade composition, I identified prominent elements based on:
- Materiality – The image clearly shows a limestone facade with wooden windows.
- Color – The beige limestone contrasts with the green wooden enclosures.
- Position & Location – I noted whether elements were distributed across floors or limited to specific zones.
- Size Variation – Determined whether elements were small, medium, or large relative to each other.
3. Estimating Quantities
For each element, I estimated numbers by visually counting instances within the image:
- Main facade – A single entity covering the whole building.
- Arched doorways – Three visible on the ground floor.
- Windows – Counted fourteen across the three floors.
- Balconies – The enclosed green Maltese balconies total six.
- Cornices – A single cornice running along the top.
- Decorative arches – Counted six above openings.
- Pilasters – These vertical divisions between windows appear eight times.
- Friezes & Reliefs – I estimated four, appearing in sculptural reliefs.
- Shopfronts – Three distinct commercial entrances were visible.
- Store sign – A single instance above one of the entrances.
4. Condition Assessment
From the visual clarity of elements:
- Most architectural features (facade, pilasters, windows, etc.) appeared in good condition.
- The store sign showed minor wear, hence marked as fair.
5. Structuring the Table
With all elements identified, I formatted the table structure with:
- Classification – Architectural function.
- Element – Specific name of the feature.
- Colour – Predominant visual color.
- Material – Identified from architectural cues.
- Size – Approximate relative size (small/medium/large).
- Location – Placement across facade levels.
- Condition – Visual state of preservation.
- Estimated Quantity – Observed count of each element.
Conclusion
The inventory table represents a systematic forensic breakdown of the building’s facade. It follows:
- Categorization of elements
- Visual identification
- Counting instances
- Assessing materiality, size, and condition
- Structuring details into a readable table
This method ensures a structured, objective architectural analysis of the image.
Would you like any refinements or further breakdowns?
I said: “That’s a good, human-oriented rationale and explanation. What happened technically in terms of the Transformer architecture?”
The response will appear in my next post.
Reference
- Kube, C., et al. (2025), ‘Doge Will Use Ai to Assess the Responses of Federal Workers Who Were Told to Justify Their Jobs Via Email’. NBC Newa, 24 February. Available online: https://www.nbcnews.com/politics/doge/federal-workers-agencies-push-back-elon-musks-email-ultimatum-rcna193439 (accessed 28 February 2025).
Discover more from Reflections on Technology, Media & Culture
Subscribe to get the latest posts sent to your email.
1 Comment