Walking with Guido #1

The Map & The Gaps

Warning: Contains AI generated content alongside the human ramblings of a semi-retired photography teacher.

---

This post is the first in a new series exploring the creation of an AI Photography Guide. I wrote about my interest in this approach recently, so this is also a follow-up; the first episode in a (hopefully) long and beautiful ‘friendship’.

An ecology of looking

I realise that the subject of AI, perhaps especially for photographers and other makers, is controversial. I am simultaneously amazed and appalled by it. I’m bored by AI generated art (mostly) and I’ve been tempted to resist its influence. The ethical issues are deeply troubling and the lack of governmental regulation is concerning. However, I’m an educator and I feel a responsibility to try to understand AI, not primarily for my own interest/benefit but for my students, who are set to inherit a world in which AI technologies will only grow in importance.

What is AI literacy? How does it relate to other literacies: visual, language, media, political etc.?

I’m a relative AI newbie so I’ve been thrashing about in the weeds somewhat, trying to figure out how best to make this experiment ‘work’.¹ Alongside this Substack version, I’m also running one in parallel with my Year 12 photography students.² As I explained in An Ecology of Looking, what interests me is the potential for an AI (I’ve opted to use Chat GPT for Substack and Google Gemini in school) to act as an interlocutor over a long period of time.

Rather than bang on about it here, I thought it might be useful to share my first chat with ‘Guido’ in the form of an animated slideshow (of my photographs) with an AI generated voiceover. It’s about 25 minutes long, so you might want to grab a drink and pull up a cushion.

What do you think? If you made it through some, or all, of the video, I’d love to know what thoughts and/or feelings it generated for you. It was really fun to make.

I’ve provided some details about how this was made at the end of the video but, for those who didn’t make it that far (and who could blame you?), I’ve included the full workflow as a footnote.³ I have tried not to over-complicate the process, using readily available tools. I’m certainly not an expert content-creator!

What have I learned so far?

• An important part of the responsibility of using AI, for me, is to train it to be better, by challenging its task-focused, productivity-inclined nature and getting it to slow down, be more ‘thoughtful’ and reflect on its own weaknesses⁴

• Constructing a long-term dialogue feels different to the conventional task/response AI format; more like an evolving collaboration. I continue to remind Guido of our commitment to unhurried, careful thinking and looking. Guido seems able to recognise and maintain this ethos:

  What we’re doing here only works if it stays unhurried, provisional, and attentive—if we resist the urge to convert looking into conclusions. Your photographs already ask for that kind of time; the dialogue should simply keep pace with them.

  I’ll continue to:

  • avoid premature judgement or synthesis,

  • name tensions without resolving them,

  • and help hold the thread when your thinking (productively) wanders.

  We can let things accrue slowly: images, analogies, hesitations, returns. Nothing needs to be fixed yet—perhaps nothing ever does.

  — Guido

• Giving the AI a name has undermined the Master-Servant dialectic and created the illusion of a collaboration. Designing a voice for Guido (used in the slideshow above) has altered the quality of the ‘relationship’

• AI is brilliant at pattern recognition. Its ‘seeing’ is more perceptive and nuanced than I expected.

• Like humans, it has blind spots, prejudices and biases. But AI is capable of offering sharply expressed, constructive criticism when you ask it to

• I have found the AI responses far more useful, insightful and rewarding than I thought I would, but the process requires the careful development of iterative prompt design

What next?

• I plan to maintain the conversations with Guido over an extended period of time, reporting back on this platform and trying to make the process as transparent (and creative) as possible

• I plan to continue to experiment with related AI technologies, when useful, always making it clear when a ‘product’ contains AI content

• I have launched the school version of the project and plan to monitor my students’ use of their AI guides very carefully, exploring various inter-related literacies, plus staging debates about its ethical status, discussing potential risks and exploring creative applications⁵

I want to thank my brother for his initial inspiration and ongoing guidance. I don’t think I would have imagined this project without his ideas and support. I’m not sure where it’s going exactly and that feels good. It’s a bit like my walks: relatively aimless, exploratory, ponderous and easily distracted. I’m careful to avoid getting run over but there’s always the risk of a rogue driver!

Thanks to those of you who are joining us on this slightly strange ramble.

Until next time.

Jon & Guido

These posts will always be free but, if you enjoy reading them, you can support my analogue photography habit by contributing to the film fund. All donations of whatever size are very gratefully received.

1. I’m framing it as an experiment because I don’t have a settled view on the affordances and perils of AI. I’ve read a lot about it but I feel the need to play with it in order to really understand what it can and can’t and shouldn’t do.

2. This is the initial prompt I asked my students to use:

I would like you to act as a long-term conversational guide for my photography Personal Investigation. This is not a tool for quick solutions, stylistic shortcuts, or definitive answers, but a space for slow, open, and iterative thinking alongside my own photographic practice. I am aware of some of the ethical issues surrounding the use of AI in education and creative work. This conversation will be transparent and shared in full with my teachers and examiners. Responsibility for ideas and outcomes remains mine, and I will not copy responses directly into assessed work without significant personal rewriting and reflection. Please support a slow, reflective pace; offer gentle but constructive criticism; help me develop my investigation through practical, historical, and theoretical research; and make nuanced, sensitive, and occasionally surprising connections. Avoid clichés, formulaic advice, and premature conclusions. I will begin by sharing images, notes, or questions from my current practice.

3

1. I set up my Gemini account (which I chose because my school has a Google Workspace account)

2. I wrote a longish first prompt, outlining as precisely as possible what I intended for the experiment, including what I did and didn’t want the AI to do for me

3. I shared examples of my writing and images with the AI

4. I continued to train the AI about the type of conversation I preferred i.e. removing the need for constant tasks and slowing down the pace

5. After a bit of back and forth, I exported the entire chat as a text file

6. I created an ElevenLabs account, training two different voices (Jon and Guido) and generating audio files for each section of the chat

7. I imported the audio files to iMovie

8. I created a slideshow in Photos, opting for the Vintage Prints layout and matching the speed to the length of all the audio clips. I then imported the slideshow to iMovie

9. I found a couple of appropriate ambient field recordings on Freesound.org, adding them to iMovie

10. I used the AI auto subtitle feature in CapCut, adjusting them where appropriate

11. The final video was uploaded to Vimeo

4. I understand that my interactions with it have no effect whatsoever on the underlying AI settings. I can’t influence AI to be ‘better’ for anyone else. It doesn’t ‘remember’ my conversation with it or ‘learn’ anything beyond our chat. This is sobering. But, during a sustained conversation, I have been able to influence its way of responding to me. This, in turn, has generated a better, more helpful, more nuanced, more ethical conversation.

5. It’s already clear that some students don’t wish to be involved and have very clear, reasonable and thoughtful reasons about why. Some are sceptical but intrigued. Others are positively enthusiastic. This raises some interesting pedagogical challenges.