Malthe Erslev – I forced a bot to read over 1000 articles from open-access journals and then asked it to write an article of its own. Here is the result.



  1. Thank you for the interesting piece, Malthe. Lucid observations. I would like to kick-start the discussion with a few questions:

    1) could you elaborate your contention as to a combined understanding of code & input being the best way to appreciate generative art (p. 4)? What might be the gains and losses?

    2) if bot-mimicry triggers (intellectual) curiosity & pleasures through deception and human attempts to decode such deception, could you speculate on how the phenomenon might develop if/when we reach the point of false-proof deception?

    3) what might be some alternative media or communicative contexts through which we can examine the seemingly paradoxical dynamics in cultural assumptions of A.I., as you described (pp. 5-6)? How might the dynamics change accordingly, if at all?


    1. Hi Carman. Thank you for your insightful questions. Here are some preliminary thoughts — I look forward to discussing further in Cambridge.

      1) To specify, I write ‘output’, not ‘input’. Thus, the argument is that when reading a generative work, neither the output nor the code is sufficient to establish a nuanced and in-depth reading of the work. This approach is of course grounded in software studies, and is thus biased towards a study of the convulgence of work and machine, rather than the context of implementation. As such, the approach is useful in uncovering the relation between conceptual inscriptions and output.
      For instance, it matters whether or not a bot writes according to a conceptually developed grammar and lexicon (e.g. the love letter generator developed for Manchester Mark I) or according to a machine learning process (e.g. recurrent neural networks) with an input corpus — and further, it matters what the input corpus consists of, and how it is processed to produce an output. Accordingly, it matter just as much how these conceptual inscriptions are realized in the output, in part because the very definition of generative code makes the output unpredictable, in part because the output is just a real a manifestation of the work as the code is (though this is not to say that they are equivalent or even similar). At least these aspects matter when reading a generative work according to its context of production — as an echo of the material with which it deals. As mentioned, the approach is not in itself attuned to reading societal implementations of generative software, though it is applicable in such a context if coupled with other methods.
      By the way, I should clarify that even though I write ‘code’ in that section, I do conceptually include the entirety of the generative apparatus, including database, input, hardware, etc., and not just the written code. Furthermore, I go on in my essay to conduct a reading inspired by software studies without any access to the code, while still arguing that it may enlighten important aspects of the work, but the argument for that is of course in the essay.

      2) If the point of false-proof deception is characterized by bots that pass the Turing test, then we are already there, at least in some genres and contexts. Nevertheless, and to take your question seriously, I believe such ‘intellectual curiosity and pleasure’ will always relate to a contemporary, and to certain shared conceptions of concepts such as language, originality, etc. Thus, I believe that it is more a question of form, style, and implicit expectations than one of content which drives this curioaity and pleasure. In other words, there are a lot of fear for stuff like ‘fake news’ or ‘filter bubbles’ at the moment, and bots get a lot of the blame — thus, the example provided in my essay deal with these issues through literary experimentation with style and rhetoric.
      I can only imagine that the questions of truth, human-vs-nonhuman-writers, and creativity will continue to be at the center of public discourse for some time to come. One interesting question is of course if (or rather, when) this type of writing becomes trivial, and which kinds of experiments will follow. In contrast to some, I do not believe that the experimentation will just stop, but I do believe it constantly changes.

      3) It feels like you have some contexts in mind — please share these, it would be great to discuss it further in other contexts. Coming from a design background, my first though was speculative and critical design, where issues such as this have been explored and problematized numerous times. But I am curious to explore further in other domains.


      1. Malthe, thank you for your detailed response and clarifications. Likewise, I would look forward to future experimentations after the oomph of bot-mimicry fades. I find it interesting how diverse approaches seem to (inevitably and) significantly connect with questions of culpability, accountability, and truth. For instance, your approach involving literary experimentation and rhetoric and Carleigh’s approach to glitch.

        My 3rd question comes from a place of curiosity. With my background in game studies and tangentially HRI, I wonder if and how these artifacts, technologies, contexts may intersect for a deeper understanding of what you articulate in your essay. I also shades of this inquiry in our dialog around my piece.

        The 1st question was perhaps misguided by my interpretations regarding output, “imprints of generator,” and relevant concepts. I appreciate your untangling them. See you at Cambridge in a few days.


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s