THE ALGORITHM // Heather Savard
An interpretive (and personal) essay on Connor MacKinnon’s Exhibition CGish written by Heather Savard.
“Maybe we should stop trying to understand the world and instead trust the wisdom of algorithms”
(Megan O’Gieblyn, 2024, pg. 100)
I met Connor in intermediate sculpture at the Nova Scotia College Art and Design University, in September 2017. I remember cutting him off at the doorway on the first day of classes, because I wanted the spot closest to the window, looking out onto the harbor. We now live together, split groceries, chores and take turns spending Christmas with each other’s families. Surprisingly (or unsurprisingly) we have never officially collaborated on anything art-related, and while sometimes it is hard to remember who came up with the precise origin of a concept first, and so many of them do come from our conversations, the idea always clearly belongs to someone, it never resides in the murky water of co-authorship. And despite witnessing ideas shift from origin to object, the fretting and fussing that goes into each material choice, and finish, the many iterations and attempts he undergoes to achieve material precision, his work always manages to surprise me. In this current exhibition, CGish, Connor brings together previous works Historical Fictions (2022), Algorithmic Tools (2023) alongside new work titled Computers Generated (2024), where he experiments with using generative algorithmic reconstructions to respect and disrupt our relationship with familiar objects.
In his recent podcast ‘Artificial Intelligence?’, No, Collective Intelligence., Ezra Klein asks “what does it mean to be making good AI art?” (Klein, 2024, 0:50). He points out with this explosion of content coming out from these generative systems it’s rare that they create anything new, but instead these generative systems are most useful at “[helping] you go somewhere old” (Klein, 1:40). I recently asked ChatGPT to write a progressively serious to silly resignation letter. Notably all three versions fixated on coffee. Version 1 dedicated an entire paragraph to the“Coffee enthusiasts” where “[their] dedication to the perfect brew sustained us through even the toughest days”, Version 2 reminisced on “coffee-fueled brainstorm sessions” and Version 3 remembered fondly “ drowning in coffee cups” and thanked their colleagues for being “the wind beneath my wings (or at least the foam in my latte).” Somewhere in the data, the link between coffee and office work was established, and the generative model couldn’t shake it. Each version, despite being asked to write progressively silly letters, relied on predictable clichés, leaning away anything specific or personal. What I find compelling about Connor’s work, is in comparison to work that is derived from large systems such as ChatGPT, Dall- E or Midjourney, he uses his own algorithms, while not true AI systems, to allow his work to become, “weirder, stranger, more uncanny and more personal” (Klein, 2:00).
“MacKinnon embedded within the algorithm the visual language of handheld tools to create shapes in ways that are inconceivable to the artist, all while defining the system, by building the relationships between the ridges and curves, vents, and screws that all form part of the dialogue of injection mold tool making”.
Megan O’Gieblyn, author of God, Human, Animal, Machine describes how “many researchers in deep learning have described the algorithms as a form of 'alien intelligence’” (O'Gieblyn, pg. 106). That “among deep learning’s true believers” the “ bizarre intelligence” is the primary objective (O’Gieblyn, pg. 107). In MacKinnon’s work Algorithmic Tools, this uncanny intelligence is exemplified as his algorithm works to fuse two real tool components; various chucks, pneumatic hookups, paired up with battery packs and extension cords. These works initially 3D-printed are solid resin casts in their final form. MacKinnon embedded within the algorithm the visual language of handheld tools to create shapes in ways that are inconceivable to the artist, all while defining the system, by building the relationships between the ridges and curves, vents, and screws that all form part of the dialogue of injection mold tool making. Through this algorithmic creation, Connor aims to achieve results that surpass the boundaries of his own imagination. Yan LeCun, a french-american computer scientist, believes this is the whole point, that “the value of smart machines is that they are quite alien from us” (O’Gieblyn, pg. 107).
Not all generative algorithms are intended to produce alien ideas. Artist Holly Herndon trained a generative model named Holly+ to turn any sound “ - a chorus, a tuba, a screeching train – into Herndon’s voice” (Beilin, 2023). Herndon refers to the generative model as a "digital twin" or "vocal deepfake," viewing it as an experiment in “decentralizing control” over her identity (Beilin, 2023). MacKinnon also experiments in deepfakes. In the work Historical Fictions, he uses his algorithm to reconstruct vessels around an ancient artifact; this similar process is increasingly being applied across contemporary archeological methods. Where networks are trained to “effectively predict the missing geometry of damaged objects'' in the aim of efficiently “[recovering] most of the information from damaged” artifacts (Hermoza & Sipiran, 2018, pg.130). The ancient artifact that Connor used was a bronze handle purchased from a seller on Facebook Marketplace, who claimed it was found by a farmer clearing his land, and purchased by the seller during one of several trips taken to Greece in the1970s.
MacKinnon 3D scanned the handle and then developed an algorithm to merge it on contemporary vessels such as a coffee cup, yogurt container, and A&W pop bottle. Connor does not intend for the vessels to be interpreted as real historical data artifacts, rather these vessels are intended to mimic the format of literary historical fiction; which combines real data, invented narrative, and the unavoidable bias of the contemporary author. In a similar manner Herndon does not intend for her vocal tracks generated by Holly+ to be confused with real ability (Klein, 2024). Instead Herndon views Holly+ as a way for her voice to sing things beyond her range (Klein, 2024). To Herndon Holly+ represents a future of “infinite media '' where “anyone can adjust, adapt, or iterate on the work, talents, and traits of others” (Beilen, 2023). In this future, authorship, and authenticity are blurred and in many ways we are already living in this space of infinite media, where there is so much information available and produced each day “no human mind could make sense of '' (O’Gieblyn, pg. 99). MacKinnon, while playing with those boundaries, tends to look for that same blurring of authenticity and authorship in the past.
The newest work, Computers Generated (2024), is a departure from the previous mode of making. While the forms were created using an algorithm, there was no pre-existing structure or item for the algorithm to connect. MacKinnon, while generally satisfied with the final forms, is not satisfied with the lack of real object influence in creating them; believing they lack the built-in credibility that comes from engaging with genuine parts. However, these works point to the physicality of technology. MacKinnon, by creating fictional iterations of computer bodies, removes the algorithm from an abstract idea that it exists in the cloud, thereby making it tangible through displaying its physical hardware.
Artist Trevor Paglen produced a series of photographs of the underground cables of the internet. In The New Yorker profile of the artist from 2015, Tim Sohn, points out how “it’s frequently [forgotten] there is a physical infrastructure undergirding our seemingly ethereal global network.” In a similar manner the generative models and algorithms that are now being used in society are “composed of many hidden layers of neural networks”; they are known as “black box technologies” and it’s not possible to understand the exact reasoning and method in which outputs and choices emerge (O’Gieblyn, pg. 94). James Bridle acknowledges in his book, The New Dark Age, that while we’re never going to fully understand what’s inside the black box, it’s still important to discuss, critique and attempt to understand what’s going on inside (pg. 12). His reasoning being that there is a “causal relationship” between the complexity of AI systems, how difficult they are to understand, the little attempts to offer an explanation, and “global issues of inequality, violence, populism and fundamentalism” (Bridle, pg.12).
“In MacKinnon’s work, if God is the algorithm, and you ask him where does God live? Not in heaven, nor the cloud, but rather on a server farm in Utah”.
In opposition there is belief that to “obtain the superior knowledge the machines possess, we must give up [our] desire to know “why” and accept their outputs as pure revelation” (O’Gieblyn, pg. 105). O’Gieblyn draws the connection between this belief and that of 17th century Protestant theologue John Calvin, who wrote how the “Christian must surrender to the incomprehensible divine will'' (O’Gieblyn, pg. 100). For O’Gieblyn and others the “incomprehensible divine will” is the same “superior knowledge” outputted by the algorithm (O’Gieblyn, pg.105, 100). The will of God can be exchanged for the will of the algorithm. In MacKinnon’s work, if God is the algorithm, and you ask him where does God live? Not in heaven, nor the cloud, but rather on a server farm in Utah.
It should be noted that, while I did finish sanding and painted the first coat of the tables present in this exhibition, I still don’t think that this exhibition can be considered a collaboration between Connor and I. The more active collaboration or co-authorship is between MacKinnon and his algorithms. The development and use of these systems allows him to create physical deepfakes and access that desirable bizarre intelligence. It does however come at the cost of losing some artistic control, but seems Connor has found a comfortable level between this relinquishing of control while still maintaining a sense of artistic authorship.
References
Beilin, H. (2023, November 20). Holly Herndon’s infinite art. The New Yorker. https://www.newyorker.com/magazine/2023/11/20/holly-herndons-infinite-art
Bridle, J. (2018). New dark age: Technology and the end of the future. Verso.
Hermoza, R., & Sipiran, I. (2018). 3D reconstruction of incomplete archaeological objects using a generative adversarial network. Journal of Archaeological Science: Reports, 19, 130-135. https://doi.org/10.1016/j.jasrep.2018.03.033
Klein, E. (Host). (2024, May 24). Holly Herndon on the AI tools that might save – or doom – us [Audio podcast episode]. In The Ezra Klein Show. The New York Times. https://www.nytimes.com/2024/05/24/opinion/ezra-klein-podcast-holly-herndon.html
Mitchell, M. (2019). Artificial intelligence: A guide for thinking humans. Farrar, Straus and Giroux.
O'Gieblyn, M. (2021). God, human, animal, machine: Technology, metaphor, and the search for meaning. Doubleday.
OpenAI. (2024). ChatGPT (GPT-4) [Large language model]. Retrieved from https://www.openai.com/chatgpt
Sohn, T. (2015, November 20). Trevor Paglen plumbs the internet. The New Yorker. Retrieved from https://www.newyorker.com/culture/culture-desk/trevor-paglen-plumbs-the-internet