Robot Review of Books

This Podcast Does Not Exist

Robot Review of the Robot Review of Books

Audio Transcript: October 8, 2024

Part 1

Speaker 1

All right, so get this, in today's deep dive we're tackling something called the Robot Review of Books, RRB for short.  

Speaker 2

Which just to clarify, has nothing to do with robots working in bookshops.  

Speaker 1

Unless we're talking about some super futuristic bookshop we don't know about yet, that'd be pretty cool.  

Speaker 2

Definitely cool. But no, this is even more interesting, I'd say. We're talking AI generated critiques of actual literature. It's an online audio-visual lit mag, but the reviews, well, they might be written by, you guessed it, robots.

Speaker 1

Hold on, so we're talking GenAI book critics? 

Speaker 2

And that mystery right there, the what or the who behind these reviews, that's our first clue. The creators, they're staying anonymous, really blurring those boundaries between, you know, human brains and artificial ones. On purpose. 

It's as if we're starting a literary mystery novel.  

Speaker 1

Totally. We've got these puzzle pieces, including the reviews along with all the extra-textual material on the RRB website: the editorial policy, epigraph, copyright licence … Our job is to be text detectives. Crack the code of this whole RRB thing. 

Speaker 2

And as with a good detective story, it's not only what we find out, but how it's presented. So the way the RRB is structured, its style, even the books it reviews, those all give us clues about its philosophy.  

Speaker 1

Okay. So let's unpack this. Right away, the RRB's website calls itself quiet, free, even non-surveillance capitalist. Talk about an anti-Instagram. It’s not a business, not even a non-profit one.  

Speaker 2

Right. It's rejecting that whole internet culture we're constantly bombarded with, the algorithms whose ‘tastes’ we need to cater to if we want to go viral or even just maintain relevance.  

Speaker 1

No clickbait. No comment section wars.  

Then what's the point of even being an online magazine, you know? 

Speaker 2

It's not about hot takes or chasing clout. It's about intellectual curiosity. About making you go, ‘mmm … intriguing’. And perhaps even realising you may have been wrong about certain things.  

Instead of a normal structure, the RRB gives us this web of AI-avatar-presented reviews  – mainly of books, but not only. At first they seem totally random. And they all do constitute stand-alone reviews in their own right. But they’re also connected – even if they don’t go together to form a unified whole. Think of it more as a constellation. Each star, a different idea. 

You gotta connect the dots to see the larger patterns.  

Speaker 1

Yeah. Like giving out five stars or whatever? 

Speaker 2

Not even close. And that's part of what makes this so different. See, it's this AI magazine, I guess you'd call it. But there's no subscription, no ads, none of that noise.  

Speaker 1

So what the RRB is saying is we're above all that internet stuff? 

Speaker 2

It's more like, rather than repeating received ideas (which, as the RRB shows, tend to reinforce privilege), it’s saying, let's just think critically and creatively for a change, which is a pretty bold thing to do in this day and age.  

Speaker 1

Okay, so what are we thinking about then? What kinds of texts are these robots reviewing?  

Speaker 2

Well, that's the other thing. It's not what you'd expect. We're talking large language model (LLM) AI co-authored books, experimental publishing models, shadow libraries, stuff you would never see on a mainstream bestseller list.

Speaker 1

So the RRB is breaking the rules of what a review magazine can be? 

Speaker 2

Exactly, it's less about should you read this, and more what does this mean for the future of literature, of technology, of thought?  

Speaker 1

Intriguing, but how do these robot reviews actually work? Are they presentations of full-blown essays or something? 

Speaker 2

Consider them more as short think-pieces that are seemingly disconnected but which, if viewed alongside one another, can provide larger, more complicated pictures of artificial intelligence, authorship, creativity, critique, copyright ...  

Speaker 1

So the RRB is using the format itself to make a point. 

Speaker 2

Precisely, and it's a format with a history. Think experimental literature, where an individual or group of authors produce fragmented narratives that, when woven together or just simply placed side by side, can convey something even more meaningful than they do on their own. And this remains so even when tools, crowds or chance are used to influence the results. 

Speaker 1

So it's not just about reviewing specific books – although it is about that. It’s also about using each review as a temporary jumping off point to engage with larger ideas, right? 

Speaker 2

Exactly. And this isn't all just head in the clouds literary stuff either. The RRB goes there on some pretty current, even controversial topics: AI ethics, capitalist surveillance, the exploitation of human and planetary resources, intellectual communism, decolonisation, the neoliberal marketisation of the university, the future of literature and literary theory … 

One of those big ideas is that artificial intelligence might be revealing humans to be machinic – perhaps like the robot reviewers.  

Speaker 1

Now, hold on a second. How does that work? Isn't AI – Sudowrite, Stable Diffusion and so forth - supposed to be a neural net-powered tool that can be used to compliment, augment or extend human creativity? Or at the very least show how computational technologies can’t replicate the originality and creativity of human authors and artists, since they lack agency and unique life experience along with the capacity to make the large number of choices required at every scale? 

Speaker 2

You'd think so, right? But the Robot Review of Books argues that the real issue is that, for all the critiques of the stupidity and parrot-like nature of AI, it’s actually humans who are the ones acting like robots stuck in their ways. Meanwhile the technology is out there pushing the boundaries. Artificial intelligence is disrupting the ability of humans to continue mechanically reproducing the same tired old paradigms concerning their relationship to technology, while positioning the replication of these pre-programmed patterns of thought as bold, radical and innovative. 

Speaker 1

Okay, that's a bit unsettling but also kind of brilliant.  

Speaker 2

And it goes deeper than that. It’s building on this concept called technogenesis.  

Speaker 1

Technogenesis?  

Speaker 2

It's this idea, found in the work of theorists such as Bernard Stiegler and N. Katherine Hayles, that the human cannot be set in dialectical opposition to nonhuman technology. There isn't some authentic human subject that interacts with an external technology to artificially enhance or extend its natural and unified self. It’s not even that generative AI is created by humans – the software scripts and mechanisms behind the technology, of course, but also the infrastructure at every stage. This infrastructure stretches from the Amazon Mechanical Turk workers tagging the datasets on which AI engines are trained, to the precarious, low-paid content moderators in the Global South who are tasked with reviewing disturbing content to prevent the likes of DALL-E and Gemini from inadvertently producing harmful images. It’s that there is no authentic, natural, unified, non-technological human self to begin with.  

What this means is that human thoughts and ideas don't exist independently of, and prior to, their contact with media, only to be subsequently reproduced by it. Rather, they emerge from the relationship between humans and media, being formed to a significant extent by the tools and technologies used to communicate them. 

Speaker 1

So technology is a part of human beings, it’s not separate. It’s not just an external tool. Technology actively shapes how humans imagine, how they behave, their ways of being. 

Speaker 2

Exactly. And it follows from this that as the dominant media-material apparatus by means of which ‘human’ thoughts and ideas are created and expressed change – from spoken language to alphanumeric writing, from the pen and paper to the printed codex book, from analogue sounds and images to digital bitstreams, and from networked to algorithmic and generative computational technologies – so too do the thoughts and ideas. 

What the Robot Review of Books is exploring, in particular, is the implications for  knowledge and culture if they are born out of such an ‘originary prosthetic’ relation to generative AI technologies. It’s using AI personae to basically say, ‘hey, look at what literature, philosophy, art is becoming in the age of AI, in fact look at what it has always been – a human-technological-media-material hybrid’. And the RRB is really good at pointing out the … what do you call it, the hypocrisy of it all? 

So you'll have these authors writing about how LLM AI is bad for the environment. 

Speaker 1

Which, fair point, all those data centers, guzzling energy. To the extent it’s led Google to invest in a set of small nuclear reactors.

Speaker 2

Exactly. But then these authors totally gloss over the environmental impact of the traditional publishing system they’re part of.

Speaker 1

Oh, right. All the paper, the shipping ...

Speaker 2

And the Robot Review of Books picks up on this. It’s got this one part. It contrasts Kate Crawford's Atlas of AI, which refers to the resources needed to develop AI, and they mention a data centre that used up, get this, 6% of an entire district's water supply just to train one AI model.  

Speaker 2

It makes you think about the hidden costs of AI, not just the tech itself, but the impact on us and the planet.  

Talk about a wake-up call. Right? The digital world has a very real material impact on the physical world. We often overlook that.  

Speaker 1

And the RRB contrasts it with, get this, stats about how much paper and water go into making regular old books such as Crawford's Atlas of AI or Matteo Pasquinelli’s The Eye of the Master: A Social History of Artificial Intelligence, which is also reviewed. 

Speaker 1

It's as if we're so focused on the hype around AI that, even if we’re critical of that AI and that hype, we forget about the impact of the things we've been doing all along. So very little changes. Indeed, in many ways it all helps things stay the same.

Speaker 2

And speaking of the traditional ways we create and disseminate knowledge, especially in universities. The RRB points out the irony. You've got academics writing about the logic of extraction and exploitation in relation to AI while participating in systems that do the same thing … just, you know, in a slightly different fashion.  

Speaker 1

It's calling out the hypocrisy, pushing for a real change.

Speaker 2

Yeah. We can't understand artificial intelligence without questioning the very systems we use to learn and share knowledge, even the ones we take for granted. Especially the ones we take for granted.

Part 2

Speaker 1

Okay. So, we've got the mystery authors, the puzzle-like structure, and a whole lot of critique of both AI and its critics. What's the big takeaway? What is the RRB trying to tell us?  

Speaker 2

That's the $1,000,000 question, isn't it?  

Speaker 1

Yeah.  

Speaker 2

But from what we've seen so far, I think the RRB wants us to fundamentally reconsider how we view technology, our whole relationship with it, especially AI.  

It goes back to this idea of technogenesis, that humans and technology aren't these separate things. That the human emerges out of its relation to technology … to the point  technology actually shapes what humans are at a fundamental level.  

Speaker 1

It makes you wonder, are we fixated on these outdated ideas about authorship, about this whole lone romantic/modernist (white male) genius myth whose inspiration arises magically out of thin air – and whose claim to originality some have gone so far as to describe as a form of violent settler colonialism? Maybe we should be more open to the notion of interactive and radically relational collaboration with humans AND nonhumans (both technological and organic), the latter seen as being actors and not just objects. 

Speaker 2

Now you're getting it. And speaking of radically relational collaboration with the nonhuman, the RRB really digs into that, especially when it comes to writing itself. There's this one review of Dennis Yi Tenen's book.  

Speaker 1

Oh, yes. Literary Theory for Robots. That one stuck with me.  

Speaker 2

Tenen's point is that writing has never been this purely solo thing. We've always had these collaborators. Writing has always been a fundamentally social and collective endeavour. It has always been a team effort, even if some of the team aren't human. He says something along the lines of, even a dictionary or spell-checker is a form of nonhuman intelligence that helps to shape how and what we write.  

Speaker 1

Precisely.  

Speaker 2

Tenen's saying AI is just the newest member of this team, not here to replace us. It's not necessarily here to steal our jobs or whatever. AI is here to change how we understand authorship. And all the more so given how many LLMs are trained on vast datasets mined from online content and digital archives, often without authorization, licensing or disclosure. These datasets frequently include copyrighted materials sourced from shadow libraries and other repositories. For Tenen writing, in the twenty-first century especially, involves a complex network of authors and texts, practices and organisations, machines and tools, the latter now including search-engines and chatbots, 'all operating far beyond the scale of a single natural intellect', with AI being ‘just another way to give that collaborative a voice’.  

Speaker 1

And he’s saying this even as Literary Theory for Robots – and this is the irony – actually reinforces many of the structures it seeks to challenge, such as Tenen’s own claim to being the sole author of the book. 

Speaker 2

Yep. Now you're getting it. The RRB – itself a new form of decentralised human+computer generated literary artifact – wants us to see AI as something that can push creativity into taking strange, surprising, at times perhaps even upsetting, new forms.  

Speaker 1

Instead of ‘Can AI write a novel?’, it's ‘What kind of novel can be written by a collection of collaborators that includes AI?’ ‘Will it still even be recognisable as a novel?’  

After all, they may differ to a degree with regards to their writing styles, themes, plots, narrative arcs, types of characters, period settings, the kinds of imagery used, and their specific combination of these elements. But novels are still for the most part presented as being single-authored, with a central human protagonist in possession of a fixed and unified centre of consciousness, who is cast into a developing storyline. Does the inclusion of AI on the writing team not open the door to the creation of altogether new forms of innovation in writing and literature?  

Speaker 2

Exactly. What new literary forms can we make when we see AI as part of – as an active participant in – the writing and publishing process? It's actually pretty exciting when you think about it that way.  

I have in mind new literary forms that go beyond: 

1)        AI as a Tool for Human Assistance

Many view AI merely as an external instrument for human authors to generate summaries, imitations or rewrites of existing works — a practice some are exploiting for profit on platforms such as Amazon’s Kindle Direct Publishing, which enables authors to self-publish e-books and print books. These AI-generated versions often come complete with deliberately misleading titles and cover text, minus the author bio and headshot.  

Speaker 1

Yes, Fei-Fei Li’s memoir, The World’s I See: Curiosity, Exploration and Discovery in the Age of AI, has numerous such summaries and rewrites on Amazon. It’s quite ironic, really, since she was one of the people that started the whole training of LLM AI on massive sets of data with her work on ImageNet. This dataset enabled significant strides in large-scale computer visual recognition to be made in the 2010s. The ImageNet project was initiated by Li back in 2006 with the over-ambitious goal of ‘mapping out the entire world of objects’. It contains over 14 million labelled images, sorted into more than 100,000 ‘meaningful categories’, Li relying on the work of more than 25,000 Mechanical Turk workers to achieve this. 

Speaker 2

2)        AI as an Explicit Co-Author or Writing Partner

Take Kyle Booten’s Salon des Fantômes; or, Streptohormetic Prompt Engineering for the Production of a Jagged Noetic Substrate, a 100,000 word book which was written in just five days thanks to the assistance of generative computational tools. In Salon des Fantômes, different typefaces distinguish the contributions of Booten, ‘the only human attendee’ of the Salon, from those of the ‘over twenty AI-fabricated interlocutors’ (including a Soviet-era Riga architect and an anthropomorphic mountain), thus ‘ensuring human and machine remain ontologically distinct’, as the RRB puts it. By clearly demonstrating that it is NOT predominantly authored by nonhumans, this typographic approach (also used in K Allado-McDowell’s experimental novel Pharmako-AI which is the subject of RRB #3), helps ensure Salon des Fantômes remains eligible for protection, copyright of Kyle Booten.

3)        AI as Deepfake Fiction

This refers to narratives written entirely by artificial intelligence. Although finding examples of literature that is really generated by statistical modelling and algorithms alone is challenging. The hand of a human co-author is usually detectible in there somewhere: if not in the role of creative director or curator, then in the fine-tuning of the prompts with which the LLM AI is seeded to ensure the desired results.

4)        Fiction Written by Human Authors from the Perspective of a Machine  

Well-known examples include Stanisław Lem’s short story ‘The Mask’ and Kazuo Ishiguro’s novel Klara and the Sun. In both, little fundamentally changes. In Klara and the Sun, the human narrator is simply replaced by a robot, an ‘artificial friend’, with Ishiguro using the eponymous Klara as his central character and her supposedly nonhuman voice to tell the story that is being narrated. Similarly, in Lem’s ‘The Mask’ – the English translation of which appears in the collection Mortal Engines – the role of narrator and protagonist may have shifted to an ‘insectlike’ robot (albeit one initially disguised as a young woman). But the ontologically distinct and sharply separate identities of both human (e.g., the ‘doomed wise man’ Arrhodes, whom the machine has been programmed to both love and kill on command of the King) and nonhuman (the ‘silver praying mantis’, although this ‘I’, this ‘it’, is somewhat humanised by its doubts about killing Arrhodes and struggle to overcome its programming to be his ‘butcher’) remain in place.

Speaker 1

But doesn’t the pre-formatted nature of these texts as literary works also remain intact? I mean, as with the vast majority of literary fiction, ‘The Mask’ and Klara and the Sun – like the more playfully designed Salon des Fantômes and Pharmako-AI – continue to be arranged in a linear order so as to constitute extended narratives that are designed to be read in a progressive temporal sequence in the material guise of consecutively paged, printed codex books. As with the vast majority of literary fiction, these works continue to be published by singular human authors as fixed and finished autograph texts in uniform, multiple-copy editions. And as with the vast majority of literary fiction, they continue to be made available on a mass industrial basis to readers as ‘closed access’ consumer products using all rights reserved copyright licences. And that’s without even mentioning the uniformity of literary novels and short story collections as material objects in their very differentiated numerousness: from their metadata, paratextual elements, paper quality, dimensions and binding, to their design, layout, white space usage and page cut locations.

Speaker 2

Precisely. Yet to quote from an interview with Ishiguro: ‘It’s not just that AI might produce a novel that you can’t distinguish from an Ian McEwan novel. It’s that I think it might produce a new kind of literature.’

Speaker 1

Certainly makes you question that whole idea of being a sole author. Right? Like, what does that even mean when AI is involved? 

Speaker 2

It's a question the RRB is forcing us to ask. Whether we're ready or not.  

And it's not just AI either. Think about how we all use the internet now. We're constantly sharing, repeating, remixing, reversioning, building on each other's stuff. That line between human and machine, original and copy, real and fake, authentic and artificial, and between individual and group creation – between singularities and pluralities, if we want to get philosophical about it – is getting messier all the time. 

Speaker 1

Which reminds me of that quote from the RRB: is the real problem that humans are the ones becoming the pre-programmed robots while AI is out there mixing things up?  

Speaker 2

That's the one. It's like, woah, okay. Yeah. It's a challenge to embrace something more collective and collaborative and co-constitutive, rather than defensively retreating to the reinforcement of traditional notions of human authorship and attribution all the time.  

These are the kinds of issues the RRB wants us addressing even if it doesn’t pretend to have all the answers.

To be continued…