Can Artificial Intelligence Make Art without Artists? Ask the Viewer
Round 1
Reviewer 1 Report
This short article sets to examine the question whether machines can make art. The author examines a series of examples from the field of machine-made art to draw some insights on authorship, creativity and agency. The article is very readable and is written with an elegant turn of phrase. However it is weakened by a slight problem: it doesn’t fully know what it tries to achieve. In spite of what the abstract states, no ‘philosophical impasse’ is competently addressed questioned, or resolved. This is an issue of coherence that needs addressing by refocusing on what the key questions driving this article may be (which is not entirely clear to me).
What this article does, on the other hand, is to put together a series of quotations concerning the general topic of authorship in art (with or without machines). However, the sequence of citations never reaches a fully cohesive argumentation. As much as I appreciate reference to Foucault and Barthes in addressing issues of authorship, they should be complemented by more extensive references to the literature dealing with authorship in machine-produced art. As this is not exactly a new field the article should expand its references (some exhibition catalogues worth checking: Machine Art 1934, Cybernetic Serendipity 1968, Ghost in the Machines 2012). A poststructuralist approach might correctly be used to lead to an emphasis on the nonhuman (of machines, of agency, of technology etc), but by using a Foucault (and colleagues’) perspective what it is missed is the emergence of algorithmic nonhuman agencies (and yes there are examples of this: high frequency trading flash crash). Section 5 is puzzling for a number of reasons. To start with, it seems to take the article in too many divergent directions. Then, it makes claims that are not entirely convincing: for instance it conflates art production with technological innovation – I am not sure that the examples of personal assistants such as Siri are fully relevant here. Likewise, Turkle’s quotation (p.5) does not seem to add much to the argument. I would suggest either finding a more suitable one from one of her books. More in generally Id advice to use quotations judiciously by making sure that they are not just placeholders but fully integrated in the body of the text and with a clear rationale for being chosen. Tom Ray’s work (not in bibliography) is particularly intriguing and I would have wanted to hear more, but the final reference to the cultural impulse to imagine a ‘Creator ‘is misplaced and frankly naïve. The very last para is irrelevant and should be cut altogether.
I would suggest to look at other artists' work to better substantiate some of the affirmations (to name just a few: Franco Vaccari's technological unconscious, Roxy Paine’s sculpting machines). The structure of the article needs revision: it needs a proper introduction where the key themes and definitions of the key terms employed are clearly stated, and the key question clearly expressed. most important, it needs to set out clear boundaries of what it wants to investigate and achieve, and how, and clarify these points in its introduction. The author writes very well, and has the capacity to draw the reader in. This is good. What it's needed is more thoughtfulness about what is being said and why.
At the moment, there is an ongoing lack of clarity concerning machines, robots and AI. They are not the same things but the author seems to think so.
Several items cited in the text are not in the bibliography, this needs to be updated. And in alphabetical order by author surname please.
Author Response
* This short article sets to examine the question whether machines can make art. The author examines a series of examples from the field of machine-made art to draw some insights on authorship, creativity and agency. The article is very readable and is written with an elegant turn of phrase. However it is weakened by a slight problem: it doesn’t fully know what it tries to achieve. In spite of what the abstract states, no ‘philosophical impasse’ is competently addressed questioned, or resolved. This is an issue of coherence that needs addressing by refocusing on what the key questions driving this article may be (which is not entirely clear to me). What this article does, on the other hand, is to put together a series of quotations concerning the general topic of authorship in art (with or without machines). However, the sequence of citations never reaches a fully cohesive argumentation.
The authors concur that our original draft did not draw the connections between quotes clearly enough. In response to this valid criticism, we have added transition sentences and been more selective about quotes to strengthen the flow of the argument.
* As much as I appreciate reference to Foucault and Barthes in addressing issues of authorship, they should be complemented by more extensive references to the literature dealing with authorship in machine-produced art. As this is not exactly a new field the article should expand its references (some exhibition catalogues worth checking: Machine Art 1934, Cybernetic Serendipity 1968, Ghost in the Machines 2012). A poststructuralist approach might correctly be used to lead to an emphasis on the nonhuman (of machines, of agency, of technology etc), but by using a Foucault (and colleagues’) perspective what it is missed is the emergence of algorithmic nonhuman agencies (and yes there are examples of this: high frequency trading flash crash).
The authors appreciate the suggested historical references; however, our aim in this essay is not a review of past approaches to this topic from within the discipline. Rather, we have tried to select specific perspectives from outside of a narrowly defined history of machine-produced art in order to suggest new ways of thinking about whether machines can be artists. While algo trading is certainly an interesting model for a nonhuman software agent, we don’t immediately see how this example would prompt viewers to imagine its designer to the same extent as the examples we chose.
* Section 5 is puzzling for a number of reasons. To start with, it seems to take the article in too many divergent directions. Then, it makes claims that are not entirely convincing: for instance it conflates art production with technological innovation – I am not sure that the examples of personal assistants such as Siri are fully relevant here. Likewise, Turkle’s quotation (p.5) does not seem to add much to the argument. I would suggest either finding a more suitable one from one of her books. More in generally Id advice to use quotations judiciously by making sure that they are not just placeholders but fully integrated in the body of the text and with a clear rationale for being chosen. Tom Ray’s work (not in bibliography) is particularly intriguing and I would have wanted to hear more, but the final reference to the cultural impulse to imagine a ‘Creator ‘is misplaced and frankly naïve.
The authors have reduced several digressive references in this section by omitting them or relegating them to footnotes, and attempted to make explicit the relevance of the examples that remain. We have added some more detail about Ray’s work.
* The very last para is irrelevant and should be cut altogether.
The authors believe Helmreich’s critique is important because it points to the social dimensions of the meta-artist function. We have added a transition that we hope clarifies our reason for including this example.
* I would suggest to look at other artists' work to better substantiate some of the affirmations (to name just a few: Franco Vaccari's technological unconscious, Roxy Paine’s sculpting machines). The structure of the article needs revision: it needs a proper introduction where the key themes and definitions of the key terms employed are clearly stated, and the key question clearly expressed. most important, it needs to set out clear boundaries of what it wants to investigate and achieve, and how, and clarify these points in its introduction. The author writes very well, and has the capacity to draw the reader in. This is good. What it's needed is more thoughtfulness about what is being said and why.
While they do suggest interesting tangents, the examples of Vaccari and Paine seem less relevant than the artists discussed in this essay, in part because to our knowledge they have not created artificial agents with the same degree of autonomy. The authors are nevertheless grateful for these examples.
* At the moment, there is an ongoing lack of clarity concerning machines, robots and AI. They are not the same things but the author seems to think so.
This is a fair criticism, and the authors have tried to use more specific language throughout. However, in select cases we chose less technical terminology to make the text more readable--a style that seems implicit in the anthology’s title, “Machine As Artist.”
* Several items cited in the text are not in the bibliography, this needs to be updated. And in alphabetical order by author surname please.
The authors have made this revision.
Reviewer 2 Report
The author presents a variety of viewpoints on how to frame the question "Can machines make art?". The presentation is of interest and touches upon several issues, such as the relationship between the artist and the viewer. From the perspective of a Computer Scientist the analysis presented could be made more compelling by explicitly framing the problem along quantifiable axis such as the artist-machine relation, the viewer-machine relation, the artifact-viewer relation. The artist-machine relation, for example, could be quantified as the degree of independence between the two, measured by the level of surprise of the author with respect to the machine output. The viewer-machine relation has ties with the notions of "interpretability" or "transparency" of modern algorithmic approaches (where a large corpus of computer science literature is building rapidly). Finally the artifact quality-viewer relation could be measured by evaluating (using psychometric techniques) the extent to which the generated artifact "can put us in touch with the complexity, contradiction and limitations of the human life cycle [Turkle 06]".
The work is however of interest to raise several issues and warrants publication.
Typos: line 193: missing foot note.
Author Response
* The author presents a variety of viewpoints on how to frame the question "Can machines make art?". The presentation is of interest and touches upon several issues, such as the relationship between the artist and the viewer. From the perspective of a Computer Scientist the analysis presented could be made more compelling by explicitly framing the problem along quantifiable axis such as the artist-machine relation, the viewer-machine relation, the artifact-viewer relation. The artist-machine relation, for example, could be quantified as the degree of independence between the two, measured by the level of surprise of the author with respect to the machine output. The viewer-machine relation has ties with the notions of "interpretability" or "transparency" of modern algorithmic approaches (where a large corpus of computer science literature is building rapidly). Finally the artifact quality-viewer relation could be measured by evaluating (using psychometric techniques) the extent to which the generated artifact "can put us in touch with the complexity, contradiction and limitations of the human life cycle [Turkle 06]".
The authors are grateful for the intriguing suggestion of modeling the perceived autonomy of artmaking artificial intelligences on these three axes. Unfortunately completing this analysis would require more research and ink than we have room for in this essay, which focuses on a more specific thesis. Nonetheless the idea suggests future directions for study.
* The work is however of interest to raise several issues and warrants publication.
* Typos: line 193: missing foot note.
We’ve corrected this.
Round 2
Reviewer 1 Report
I appreciate the rigour and showed by the author in addressing the reviewer's comment and I am satisfied with this marked improvement.
