This is part of the work I’m putting together for my residency in Moskosel, in northern Sweden. I wanted to augment reality in a context that made sense as part of my trip towards Moskosel, which took 5 weeks, as I recorded video along the way as raw material for my main piece, and doing so in a way that made sense in the context of my recent broader work.
I thought it an obvious opportunity to create content for Placecloud, of which WLC recently took the reins. Not just as a promotional exercise, but really because Placecloud, as a collection of podcasts intrinsically connected to ‘place’, is exceptionally well suited to highlight some of the sites I encountered on my journey.
Then, because I also wanted to explore the theme of AR, Augmented Reality, but not in the conventional manner of using some kind of video immersion, I decided I was going to have ChatGPT tell the stories of the places I picked, and, through Elevenlabs, have a synthesised copy of my voice narrate the stories.
The synthesised voice, impressively, is indistinguishable from being real, and comes very close to sounding like me. Though, for some odd reason, it gives me a type of American accent that I don’t have, if I have an American accent at all. It also gets non-English names painfully wrong, which meant that, at times, I had to change names to their phonetic equivalent to get the AI to pronounce them more appropriately.
ChatGPT is known to ‘hallucinate’, which is a nice way of saying that it generates bullshit. Not all the time, but in a way that makes it impossible for you to know whether what it says is true, or false.
As I didn’t want to pollute Placecloud with false information, if I didn’t know some claim from ChatGPT to be true, I had to do the fact-checking myself, which included being disappointed by the, as it turned out, false claim that an orange cat named Frodo stalked the grounds of the Calouste Gulbenkian museum in Lisbon. I asked their employees.
Interestingly, the obverse was also occasionally true; I had discovered that Le Corbusier had visited the Ggantija complex in Gozo, one of the islands of Malta, but ChatGPT was adamant that Le Corbusier never visited the islands.
Having an interest in modernist architecture, I really wanted to include Le Corbusier’s view on the megalithic complex, and thus had to write it in myself.
For all places I had ChatGPT tell stories of, I also added AI generated art as one of the accompanying images. I had started off using Bing, and continued with Midjourney, from which some of the results are stunning, leaving Bing in the proverbial dust.
Early on during my residency, I chatted to part of the couple that run the program, and realised that, with Placecloud’s integrated streetview component, it was also exceptionally well suited, in principle, for the Oculus, the Facebook VR helmet.
The artists creating work as part of the residency put together pieces that will travel Sweden in a converted shipping container. The container has two sections; one room where three walls can be projected on, and one section for consuming work designed for the Oculus.
But, when I tried out the Oculus for, what felt like it was going to be, an easy exercise to use Placecloud inside the VR environment of the helmet, the experience completely failed.
An in-world ‘flat’ browser, allows for pulling up websites, but streetview and satellite maps failed to load. In Google Maps proper, satellite maps were buggy, and streetview also just didn’t work. And then this would still not be an immersive experience; just a virtual 2D screen inside the helmet’s 3D world.
So, the solution perhaps is something more like Google Cardboard. But who still has one of those lying around?