Mobile AI does fieldwork

My book came out this week: AI and Language in the Urban Context published open access (i.e. free to download) by Routledge. In that book I advance the case that places and spaces are enjoyed, appreciated, resisted, interpreted and even created through language — including conversations.

LLM applications are evolving constantly. It’s feasible to think now of having a “conversation” with an AI while in a place, about the place, much as one would with a travelling companion.

At the moment there are a few major differences. At least in my experience, conversational partners in a place like to make declarations (evaluations such as “I like that”), and raise questions: “what do you think that is?” “Do you like the view?” Rarely is it just about facts or information. It’s more likely characterised as a light-hearted “pooling of ignorance” than exchanging information — with expertise as an optional add-on.

Furthermore, we don’t always talk about the place we are in, but talk through it, drawing on place-based metaphors that fit the moment. Or our conversation is simply animated by the place, e.g. transporting us into a relaxed state where we open up to new topics and possibilities. See posts about “soft fascination.” Perhaps it’s the opposite, as a place exerts anxieties that inform our conversation. See post tagged “vertigo.”

A country walk

There’s a new housing development about one kilometre from where I live. It was a sunny April day and I took a solitary walk along a country lane to get there. I decided to test the idea of mobile LLM (ChatGPT on my smartphone) as travelling companion.

In summary, the exchange was less conversational than consultative: that of client in the company of an expert — though without the smalltalk. The exchange does raise questions about the role of a (human) consultant in such situations.

I started with a panoramic photo. My own professional interests kicked in and I prompted: “What can you tell me about this setting from an urban design point of view?”

I thought that this prompt might suggest an assessment of landscape character that I could then compare with other settings. As the attached transcript of our conversation shows, the exchange moved quickly to sustainable drainage systems (SuDS) and ended up with a seminar and fieldwork plan, complete with functioning hyperlinks and real references. See 36-page PDF of the transcript.

Reflections

My version of ChatGPT makes extensive use of retrieval-augmented generation (RAG), i.e. it draws on semantically encoded information from the web. At one stage I declared “your analysis, insights and proposals are so thorough and illuminating I wonder what is the role of the teacher?” The reply was

That’s a quietly radical and important question — especially in the context of CPD, design education, and LLM use in the field. Your reflection points toward a shift: not just in how we teach, but in how we understand pedagogical authority, authorship, and the conditions under which insight is produced.

It then explained that under the influence of LLMs a teacher might become less a source of knowledge and more a facilitator of “attunement,” a curator of encounters, an ethical interlocutor, a translator of method, and a “keeper of the situated.”

In short, the teacher becomes less a transmitter of content and more a steward of meaning, helping learners inhabit uncertainty, read context, and listen across registers — with the LLM as one voice among many, rather than a final word.

These are familiar ideas in educational theory, but here couched in the context of LLMs. Are these insights copy-pasted from some key sources? It explained that these insights about the contemporary professional were synthesised through its training, but it proffered related web links and references. I checked, and these references are accurate.

With the proliferation of LLM applications, it’s not that we won’t need teachers or experts any longer, but in light of pecuniary pressures we may need to diversify our skill set, offer something different, or there may need to be fewer of us! See post Welcome to the apocalypse.

Bibliography

• Selwyn, Neil (2021): “Should robots replace teachers?” – exploring the shifting ethical role of educators in AI-enhanced settings, Polity Press.