Graham Harman + Manuel DeLanda ISSUE: Spring 19-01


On November 15th, members of the Tarp digital team conducted individual interviews with Graham Harman and Manuel DeLanda. Both interviews focused on the use of Artificial Intelligence as a means to produce architectural drawings. We chose this subject for its relevance to the broader topic of this semester’s Tarp interviews (the state of drawing in contemporary architecture practices), and for its relevance to the expertise of the two philosophers we interviewed. Manuel teaches a GAUD elective that focuses on the integration of neural nets and AutoCad. Specifically, our conversation focused on the Ontology of scripted drawings and was driven by the following inquiries: how does the use of AI complicate the question of authorship, should scripted and human produced drawings be evaluated by an identical code of values, and what is at stake for architects employing AI in their work?

To supplement our conversation, we presented Manuel and Graham with selected work from Karel Klein’s directed research studio. Students in Klein’s studio are using a Conditional Neural Network, designed by Mark Klein (Principal Research Scientist at MIT), to create architectural drawings and objects. There are three specific images that we chose for our conversation: two plan drawings that served as inputs for the neural network, and the resultant output as a third drawing. The first plan is from John Hejduk’s 1967 Diamond Museum C project. The second is a drawing by Amir Ashtiani, a third-year student in Klein’s studio.

Left to right: (1) Plan drawing by Amir Ashtiani (2) Diamond Museum C by John Hejduk (3) Combination of the two images by a conditional neural network
Output from Conditional Neural Net

Profiles and Theoretical Background


Graham Harman is best known for introducing contemporary architecture to Object Oriented Ontology. Object Oriented Ontology (OOO) is a niche within the broader Speculative Realism movement, of which Graham is a central member. Harman is appreciated for his lucid and unpretentious prose, which is aptly displayed in the following description of OOO’s defining principles:

" 1) Individual entities of various different scales (not just tiny quarks and electrons) are the ultimate stuff of the cosmos.

2) These entities are never exhausted by any of their relations or even by their sum of all possible relations. Objects withdraw from relation.”

Sourced from Graham Harman’s personal blog.

At face value, these principals don’t present themselves as being very radical. However, when considered in relation to the philosophers that contemporary architecture has maintained the most vehement interest in – Deleuze, Derrida and Latour can be called upon here- Harman’s contributions are actually quite unique. For Latour, it is the relations between objects- as opposed to the “haecceity”, or “thisness”, of that object- which defines the nature its existence. Whereas Deleuze allocates more importance to the haecceity of objects- see his notion of potentialities- he never extricates objects from their interactions the way that Harman does. Lastly, Harman contrasts with Derrida because when all is said and done, Harman’s objects are objects, not signifiers of something immaterial, as they are in Deconstructivist theory. In conclusion, it is Harman’s ability to detach an object’s ontology from it’s interactions that make his ideas so important.

Pratt GAUD work by third year student Elham Goodarzi for Professor Karel Klein


Whereas Harman describes himself as a Speculative Realist, DeLanda can be viewed as “Realist”. Unlike Speculative Realism, which is a burgeoning, focused movement in contemporary philosophy, Realism is a much broader and centuries old philosophical movement. Although the boundaries of realist philosophy are virtually impossible to define (nor should one want to), we can safely conclude that the following principle is a common thread appearing in all philosophical factions aligned with Realism: the world that we perceive is tangible and concrete. This stands in contrast to other traditions (Platonic Philosophy being the most famous), which maintain that what we perceive as material, is merely a corporeal replica of that object and not its true existence. Rather, in Platonic philosophical doctrine, an object’s real existence exists in one’s mind. One reason that DeLanda’s work is so radical, is that he is one of the most ardent dissidents of the Anti-Realist sway that dominated contemporary philosophy for much of the 20th Century.

DeLanda’s ability to combine his philosophical doctrines with topics in physical science is another reason that he is such a unique voice within Contemporary Continental Philosophy. More often than not, it is phenomena in geology, biology and chemistry- rather than examples from aesthetics or moralism- that he utilizes to demonstrate his philosophical theories. One of the best examples of this is his tendency to employ examples from nature to help readers navigate his theory of assemblages.

If OOO is what Harman is best known for, assemblage theory is the analogous concept for DeLanda. The term itself was introduced to architect’s vocabulary through the publication of Gilles Deleuze and Felix Guattari’s seminal 1980 text, A Thousand Plateaus: Capitalism and Schizophrenia. However, the origins of the concept behind the term- in Western Philosophy- can likely be traced to the work of Baruch Spinoza. The nature of Deleuze and Guattari’s texts are such that they never provided a fully digestible explication of the topic. Rather, Deleuze and Guattari constructed a theory of assemblages without applying it beyond the realm of contemplation. Thus, DeLanda is revered for taking assemblage theory and giving it the exhaustive, critical application that it deserves.

Authorship and Agency in The Machine Age

The introduction of Neural Networks into drawing leads to questions of Authorship and Agency, a topic at the forefront of the discourse since the rise of the Digital age. During our interview, each philosopher was asked to comment on the authorship of the outputted drawing. Harman, responded by saying that it was a “fusion” of Ashtaini and Hejduk that used the Neural Network as a “background condition” to be actualized. Delanda also attributed authorship to Mark Klein, the programer that created this Neural Net. However, both thinkers clearly agreed that the Neural Network requires a human critic to create something of worth: an individual who has the capacity to make an opinion of the output, and accept, reject or edit this output until the individual is satisfied.

Manuel DeLanda

Although both philosophers agree, each has a unique view of the drawing’s ontology, but the differences amount more to their individual perspectives rather than differences of opinion. Harman views any artist or designer as a passive actor in the process of actualizing their art, as if they are another medium for the object to pass through from virtual into actual, or real. To reinforce this notion, Harman referred to the example of John Coltrane, who famously confessed his jealousy towards first time listeners of his music. For Coltrane, his music has an existence of its own that he is burdened with actualizing. He feels as though the struggle he goes through in creating his music changes his level of appreciation and intimate knowledge of the final piece. As a maker he can never fully detach himself from the process of making and as a result can not simply enjoy the piece, but must always listen critically. This makes him jealous of those who listen to his music so nonchalantly, enjoying it without the intimate understanding he has. Ultimately Harman views AI as a new medium which, like all mediums, require skill and training to master it; otherwise, the piece will be of little worth.

Google Deep Dream Generator homepage

DeLanda compared the Neural Network to a human brain, referring to the origins of the Google Deep Dream App. Neural Networks are extremely complex due to the large amounts of “hidden layers” that are actively operating at once, similar to the way the human brain works. These “hidden layers” act as if they are decentralized filters each slightly adjusting and combining the images that are inputted resulting in the final product, like the neurons in our brain. However, DeLanda recognizes many of the same points Harman made and stated he’s “not sure to what extent bringing Neural Networks into the problem [of authorship and agency] will help”. As DeLanda put it, “we do not have a mathematical theory for aesthetics”. He went on to explain how artificial intelligence, like the Neural Network above, requires human interaction to create novel ideas. Although you may be able to train them to know what you like and what you do not like, “they do not have the kind of agency required to discover something you did not already know you liked”. DeLanda compared the issue of AI in contemporary design and architecture to the problem faced by the mid century modern artists and the struggle between Craft and Gesture.

Graham Harman

Gesture vs Craft

“...Duchamp brought the urinal into the gallery and that was pure gesture, but if you do it in the first 10 years of modernism, that’s awesome, right? The rest of his contemporaries, cubists, futurists, surrealists…all had craft, all had respect for the material aspects of art… After Warhol, however, gesture began taking over, and today craft and material experimentation are largely neglected.”

- Manuel DeLanda

DeLanda went on to explain how Duchamp’s work of pure gesture was unique because it was the first of its time, but as later movements come into fruition skill had to be re-implemented. One example is Dali’s work, which successfully combines gesture and craft. This represents the execution of these notions as separate but unified. Throughout art history, as DeLanda astutely points out, there is an easily recognizable contrast between the material artist and the symbolic artist. When these paths cross we see the convergence of gesture and craft. For instance, the work of Dali, which excites the subconscious and utilizes the skill of painting to make dreams as significant as reality.

The integration of a neural network into the architectural processes begs the following question: is the sacrifice of craft inevitable? A neural network-produced plan, based on human instruction, could be seen as pure gesture and total lack of craft; however, thanks to Harman and DeLanda an alternate view is provided.

Pratt GAUD work by third year student Amir Ashtiani using machine-produced imagery

The union of gesture and craft offers a new perspective for the future. DeLanda and Harman suggest that the introduction of neural networks into the architectural discipline still requires a critical mind because there is no mathematical algorithm for aesthetics. We can train AI to create what we already like, but it is our humanity that continues to dictate what we did not know we liked.

Harman and DeLanda agree that Artificial Intelligence needs the interaction of a highly trained individual to produce a work of any value. In addition, authorship and agency will only be gained through an understanding of Craft and Gesture regardless of AI. If contemporary designers fail to utilize craft, AI will be viewed as the empty gesture of this generation.

Pratt GAUD work by third year student Brandon Wetzel in collaboration with AI


Interview Dominic Fiallo, Ciana Frenze / Content, writing, and editing Dominic Fiallo, Ciana Frenze, Matthew Mitchell, Carlos Acosta Perez, Nikhil Sharma

Report Abuse

If you feel that this video content violates the Adobe Terms of Use, you may report this content by filling out this quick form.

To report a copyright violation, please follow the DMCA section in the Terms of Use.