I finally got access to GPT4 and with it dall-e integration. From playing around with it and using prompts, it seems far better for this use case as a tool than Midjourney as seems epsecially fitting given the project type. 

A big advantage with GPT4 is the ability to input images. Therefore, I used sketches and quick massings as well as images of the Vajont Dam to influence the outputs.
I started off just with text based prompts:
After the past 2 outputs not really being what I expected or wanted. I then started to input images. These included both images of the vajont dam to prompt it to an idea of scale (later outputs seemed to get confused when I tried to remove the water elements...) as well as my initial sketches (from above).
I then experimented with creativity, however, this took it far too conceptual especially for the RIBA requirements. However, it does go back to my ideas in my dissertation, the tangibility of imagined space. How does AI "imagine" space? How tangible is it within a metaverse?
I then took some of these and started feeding in AI images, text prompts, sketches and real life vajont dam images. I then started to get some very interesting results.
A scheme that is reminiscent of the Highlines in New York. Reminds me of shards of rock, similar to the rock chips off the dolomites that make up the avalanches of land shifting. Potentially a very interesting future take on the path idea as part of the group work. Also was very interesting that now it started populating the scene, adding a real sense of scale perhaps not seen before.

After trying to remove the water from the scene to make it more similar to the current Vajont valley, I started to get far more traditional "utopian" architecture. Not the style I think fits this narrative for my architecture nor really fits the mountains I'd like to emulate and respect with more angular architecture.
Refining prompts, describing my sketches and inputting them started to get some results I loved. These really do go along with the "attack the dam" concept. This is really good as a massing concept and I think they start to really give a dominant yet inspiring interaction that can help to demystify the dam, creating hope, breaking down the barrier. The wrapping down of the dam is something I had considered early on with my "portal concept" shown in my mirror cover pages in my portfolio. These perfectly return to my early concept of wanting to break through the dam as a "portal", a great interpetation of web3 and the metaverse. A merge of physical and digital I really resonate with and am inspired by ever since I became involved with cryptography and web3 and web 3.0 in 2020/21.