Photography, responsible AI, and generative media

Our sixth KurdSoftware session covered photography, responsible AI, Kurdish language data, filmmaking, farming, and personalised learning.

Photography

Omid briefly shared his experience with his camera and microphone setup and talked about moving from Sony to Panasonic. He recommended the Lumix S9 as a great camera. Soheil also knew the model and noted that it is compact, great for video, and supports Open Gate recording.

Photography has always been on my to-do list. Soheil suggested starting with mobile photography instead of buying a dedicated camera, and Omid agreed. Omid also mentioned that the Lumix S9 is a beginner-friendly option and briefly explained the importance of understanding the exposure triangle: ISO, aperture or F-stop, and shutter speed.

Responsible AI

Soheil is organising an event called Sustainable Development in the Age of Intelligent Agents, and we wanted to better understand the idea behind it. Intelligent agents are moving beyond simple tools and starting to take part in analysis, planning, and decision-making. The event asks how we can use that power responsibly and in line with sustainable development, with a focus on innovation, trust, governance, value creation, and better decision-making.

The central point is that the future will be shaped not only by technology, but also by responsibility, dialogue, and stronger frameworks.

Teaching LLMs to Understand Kurdish

This is a project that Arash and I are working on to explore how we can make LLMs understand Kurdish Ardalani better. We discussed one of the biggest challenges: building high-quality Kurdish datasets.

For Kurdish AI, the quality of the data may matter more than the size of the model. For Ardalani, I chose a very small base model because it is easier to fine-tune and can run on devices with limited resources, such as the ESP32. Arash suggested another approach: starting with a model that has already been optimised or trained for Kurdish, so we can build on its existing knowledge instead of starting from scratch.

AI and filmmaking

We discussed how cheaper and faster video generation could transform filmmaking. Films may eventually become generative software that adapts the story or ending while someone watches.

The same models are learning motion, causality, and physical behaviour, which could also support robotics and world models. If production becomes mostly a compute cost, ideas, taste, storytelling, and distribution become the real advantages.

Farming and learning

We briefly explored vertical farming and ways AI could help through sensing, prediction, and automation.

I also shared an idea for personalised learning with LLMs. While reading Designing Data-Intensive Applications for the third time, AI has helped me turn difficult and abstract concepts into explanations that make more sense to me.

Across all these topics, the shared lesson was that AI lowers the cost of creating, but quality, responsibility, and human judgement still matter.