## Changes to make to the Midjourney paper May 2024 - emphasise that our version is a middle way. - title: generative ai and the medium of gardening - Shorten  the introduction –maybe start at page 3. - ~~Add the JAC 2023 paper to note 2~~~~ - ~~Deal with the agency objection: even if we accept this idiosyncratic way of talking about agency it downloapys the ole of the user - Expand the wine example at the end of section 2 - site the JAC paper and the elizabeth/sabrina paper at the beginning of the tool section. - Weekend after next submit. - ~~Add the year to lowe on page 11~~ - typo (capital letter) top of page 12 - ~~Start of 4.2. We introduce the category of [[machina naturans]] to characterise those machines that are capable of generating and shaping things in the same way as nature. Generative AI system are paradigm cases. The practice of making images with Midjourney is in this sense strikingly similar to gardening.  **It is** a system under incomplete control of the user.~~ - ~~To of page 18 the quote is weird. “[t]he medium present[ing] particular challenges and possibilities for artistic creativity, and the artwork makes manifest the artist’s response to these challenges and possibilities” (2016, p. 40; see also Nguyen 2020, pp. 145–152).~~ - ~~Page 18 "Specifically, given the autonomous and generative character of [[natura naturans]], gardens exhibit what we might call a [[dynamic recalcitrance]] characterised by unpredictability and autonomy of biological processes."  Say, "THE MEDIUM OF GARDENING EXHIBITS..."~~ - ~~page 18: ...depending on the art form, each can present different degrees of recalcitrance." "Moreover each art form can present..."~~ - Something more at the top of 5.2: BEFORE CITING THOMSON JONES IN THE FIRST PARAGRAPH add '[[Artistic media]] do on the one hand shape the creative activity of the artist, and on the other, the appreciative response of the audience'. 'This point has been made by Lopes 2014 [[BEYOND ART]], Gaut 2010, Nannicelli 2017. Thomson Jones effectively/usefully summarises it as follows.... (Quote) - "This is the same for paint on canvas, or film or words or marble, or gardening, or any other materials that are used to create art. This completes our positive proposal." - Section 6: *At the beginning of this paper we mentioned that Midjourney is not the only generative AI around. There are text to text systems, most famously ChatGPT, but also Llama 2, and Claude 2. There are also various other text to image models as well as Midjourney, such as [[Stable Diffusion]] and*... put the text to image generators first. - should be Italics at the top of 23. - Conclusion: "Characterising Midjourney as a type of agent captures its autonomous and unpredictable nature, BUT DOWNPLAYS ITS ESSENTIAL DEPENDENCY ON A HUMAN CREATIVE PROJECT" OR Downplays [[the role]] of [[the prompter]]. - TOOL DOWNPLAYS the autonomy of the system - This allows for the recognition of the system's autonomous nature, similar to the agency view, BUT ALSO ALLOWS MIDJOUNREY TO BE A MEANS OF CREATION, AS PER THE TOOL VIEW, AND YET ACKNOWLEDGE IS UNPREDICTABLE NATURE WITHOUT CONSIDERING IT A MALFUNCTION. without requiring the attribution of [[mental states]]. ## feedback Enrico got from Dartmouth people - there is a sense of agency which is not related to consciousness agency, just being autonomous. same as the referee suggestion - But this would mean things are collaborative - But wouldn't collaboration mean homogenety or agents? - Filmer and actors is collaboration. Actors can be recalcitrant, but the director can negotiate... we can't negotiate with machine - They are not the same kind of agency as we are - Ambassadors negotiate - it would require collective intentionality –is this possible with different intentionality? - Kulvicki: are their styles in prompting? - Cut back the examples as well as the introduction? - Shorten the quotes? 21DALL-E 3. Can what we say about Midjourney be applied to these ‘text to text’ models as well?