Final ‘Take Home’ Exam

1 Final ‘Take Home’ Exam Mount Royal University COMM 3500-001: Media, Culture and Communication Theory Professor: Dr. Richard Erlendson April 9th, 2024 Final ‘Take Home’ Exam 2 Making Connections Main Argument of the Article The main argument of the article "Indigenous knowledges informing machine learning could prevent stolen art and other culturally unsafe AI practices" revolves around the importance of blending Indigenous knowledge systems into machine learning processes to prevent the perpetuation of culturally unsafe practices, especially art and artifacts (Carlson & Richards, 2023). The article highlights the importance of using Indigenous knowledge as a means to fix past wrongs and historical injustices against Indigenous communities through practices such as the exploitation and misappropriation of traditional art and cultural artifacts (Carlson & Richards, 2023). One Strength to How the Argument was Presented One strength in the argument is that it provides concrete examples of how Indigenous communities are already collaborating with researchers to use AI in ways that incorporate traditional ecological knowledge and benefit Indigenous peoples, such as the projects in Kakadu National Park and Sanikiluaq (Carlson & Richards, 2023). The concrete examples demonstrate that Indigenous communities are already actively engaged in using AI technology in ways that align with their traditional knowledge and priorities. This shows that there is existing precedent and potential for further collaboration between Indigenous peoples and AI developers. By highlighting successful projects like the ones in Kakadu and Sanikiluaq, the article is able to illustrate the tangible benefits that can arise when Indigenous knowledge is incorporated into AI systems. One Thing New Learned 3 Additionally, a new insight from this piece is the concept of "Indigenous Data Sovereignty" - the right of Indigenous peoples to own, control, access and possess their own data (Carlson & Richards, 2023). This is an important insight because it highlights a key issue that is often overlooked in discussions around AI development and its impacts on Indigenous communities. It is important because Indigenous knowledge, cultural practices, and other data related to Indigenous communities have historically been extracted, exploited and used without the consent or involvement of Indigenous peoples themselves (Carlson & Richards, 2023). Learning about the concept of Indigenous Data Sovereignty is valuable because it highlights an important principle that should be central to any efforts to incorporate Indigenous knowledge and data into AI systems. It signals the need for AI development to be guided by the selfdetermination and data rights of Indigenous peoples, rather than treating them as passive subjects of data extraction and exploitation. One Weakness and Area for more Development to make the Argument more Malanced The article could have been strengthened by analyzing the structural and systemic barriers that have historically prevented greater Indigenous participation and inclusion in the male-dominated AI industry. This means talking about things like limited access to education, not enough representation in STEM fields, and how Indigenous people have been marginalized by colonialism. Incorporating these types of challenges and limitations would have strengthened the article's argument by demonstrating a deeper understanding of the complexities involved and would have made the case for Indigenous inclusion in AI development more comprehensive and compelling as the current version of the article only focuses heavily on the benefits, without mentioning the potential pitfalls and difficulties that need to be addressed. 4 Thinker that would most Agree with the Issues being Raised One thinker who would most likely agree with the issues raised in the article is Herman, given his critique of mainstream media behavior and performance through the "propaganda model." The "propaganda model," reflects a deep skepticism of established power structures and mass media, particularly how these entities can serve anti-democratic ends and elite interests (Herman, 2000). In the context of AI, these same concerns mean realizing that if we do not include input from diverse perspectives like Indigenous knowledges, AI systems could end up reinforcing existing power differences and continue to marginalize vulnerable communities. Including Indigenous authors and knowledge in AI is important because it connects with Herman's idea that media, and technology too - it should not just benefit a small group of elite interests. Herman states, "The media are also constrained by the dominant ideology, which heavily featured anti-communism before and during the Cold War era, and was often mobilized to induce the media to support (or refrain from criticizing) US attacks on small states that were labeled communist" (Herman, 2000, p. 102). This quote suggests Herman’s understanding of how dominant ideologies can influence technology and information flow. He could see the parallel challenges in AI development, where excluding Indigenous perspectives perpetuates a singular, biased worldview. Thinker who might Disagree with their Position On the other hand, Postman might disagree with certain aspects as he emphasizes the detrimental impact of technology, particularly television, on public discourse and culture, 5 which could potentially lead to concerns about the influence of technology, including machine learning, on Indigenous knowledge systems. Neil Postman's work, particularly his emphasis on the detrimental effects of television and the broader media on public discourse, indicates that he might question the act of integrating Indigenous knowledge into AI systems (Postman, 1985). Postman’s general distrust of technology's influence on cultural practices depicts that he would be skeptical of the idea that technological advancements could sustainably incorporate and respect these knowledge systems without making them too simple or turning them into products to be sold. Postman states, "In short, Orwell thought we would be marched single-file and manacled into oblivion. Huxley thought we would dance ourselves there, with an idiot smile on our face” (Postman, 1985, p. 14). This depicts Postman's concern that we may unwittingly adopt technology which could include AI, that makes discussions less serious by focusing more on entertainment or oversimplification. He might doubt that AI can fully understand and respect the depth and complexity of Indigenous knowledge without making it seem less important or serious. The article states, “In the wetlands of Kakadu, rangers are using AI and Indigenous knowledges to care for Country” (Carlson & Richards, 2023, para. 4). This shows how Indigenous wisdom can improve AI in areas like environmental conservation. By blending traditional knowledge with AI, the Kakadu project taps into Indigenous peoples' deep understanding of their surroundings. This blend leads to better solutions, demonstrating how AI can benefit from Indigenous insights, tackle culturally important issues, and understand diverse viewpoints better. Herman would most likely support this statement as it talks about challenging 6 dominant discourse and ensuring diverse perspectives are represented in the media, and by extension, in AI. On the other hand, Neil Postman’s skepticism about the ability of new technologies to handle complex social issues faithfully would probably lead him to disagree, as reflected in his own words, "What is happening in America is that television is transforming all serious public business into junk" (Postman, 1985, p. 15). This implies that Postman could foresee a risk of Indigenous knowledge being superficially incorporated into AI systems, thus undermining the authentic representation and application of such knowledge. Sea of Pleasure or Surveillance Dystopia In what ways has the rise of machine learning an expected outcome of the problems that Postman identified with the rise in the popularity of television and does the appeal of machine learning represent a closer view of Huxley or Orwell’s vision for the future of society? The rise of machine learning can be seen as an expected outcome of the problems Postman identified with the rise of television. The appeal of machine learning seems to represent a closer view to Huxley's "Brave New World" vision for the future of society, rather than Orwell's "1984" dystopia. Example One Postman worried that television was making people pay more attention to entertainment than to important things. Similarly, the article on machine learning highlights how AI risks misappropriating Indigenous art and knowledge. It states, "AI-generated art presents the risk of mass reproduction without Indigenous input or ownership, and misrepresentation of culture" (Carlson & Richards, 2023, para. 6). In both cases, there is a concern that technology, whether 7 television or machine learning, might make important cultural aspects seem less serious or meaningful. Television, by emphasizing entertainment content, could lead to a diminished appreciation for substantive discourse and cultural significance. Likewise, AI-generated art, lacking Indigenous input or ownership, may reproduce Indigenous cultural artifacts without proper acknowledgment or representation making Indigenous heritage and culture seem less important or real. Example Two Postman contrasted Orwell's fear of control through pain with Huxley's fear of control through pleasure. Machine learning algorithms, by showing users what they like, can make using technology enjoyable, but this might hide the fact that Indigenous rights and cultural control are being taken advantage of. The article mentions, "Including Indigenous people in creating AI or deciding what AI can learn, could help minimize exploitation of Indigenous artists and their art" (Carlson & Richards, 2023, para. 2). There is a concern that the pursuit of pleasure and convenience could lead to the exploitation of individuals or communities. Just as Huxley feared that people would be controlled through pleasurable distractions, machine learning algorithms may prioritize user satisfaction over ethical considerations, potentially leading to the exploitation of Indigenous artists and their art for the pleasure of others. This suggests that involving Indigenous communities in AI development can mitigate the risk of their exploitation for the pleasure of others which aligns with Postman's contrast between pain and pleasure, as it proposes a proactive approach to addressing the potential harms of technology by incorporating diverse perspectives and ethical considerations into its development process. Example 3 8 Postman talked about how TV changed the way people talk to each other. Instead of having meaningful conversations, TV made everything more about entertainment and less about serious topics (Postman, 1985). Similarly, the article points out the need to ensure Indigenous participation in AI development to safeguard their rights and cultural heritage (Carlson & Richards, 2023). It states, "The federal government needs to consider Indigenous Knowledges informing the machine learning informing AI, supporting data sovereignty" (Carlson & Richards, 2023, para. 6). This emphasizes the importance of incorporating Indigenous perspectives to counteract the potential negative impacts of AI on Indigenous communities. Just as television altered the way we communicate, AI development could similarly overlook the concerns and perspectives of Indigenous communities without their active participation. Therefore, involving Indigenous voices in AI development becomes crucial to counteract the potential negative impacts on their rights and cultural heritage. This proactive approach aligns with Postman's concerns about the transformation of communication dynamics and depicts the importance of maintaining substantive engagement and inclusivity in technological innovation. Does Postman believe we can counteract the negative effects of media on our ability to think critically and engage with the world around us? Why or why not? The rise of machine learning appears to be more closely aligned with Huxley's vision, where people are "controlled by pleasure" through personalized content and the illusion of choice, rather than the overt totalitarian control depicted in Orwell's "1984" (Postman, 1985, p. 14). It does not seem that Postman was optimistic about the ability to counteract the negative effects of media, particularly television, on our ability to think critically and engage with the world. He states, “What is the antidote to a culture dying of laughter? I fear, ladies and 9 gentlemen, that our philosophers have as yet given us no guidance in this matter" (Postman, 1985, p. 18). This shows that he was quite pessimistic about finding a solution to the problems he identified with television's transformation of public discourse. When he talks about "a culture dying of laughter," he's showing how worried and resigned he feels about people not taking things seriously or thinking critically anymore. Additionally, Postman saw the shift towards a "Huxleyan" future, where people "come to love their oppression" through distractions and entertainment, as a profound cultural transformation that would be difficult to reverse (Postman, 1985, p. 14). So based on Postman's own words, it does not appear that he was optimistic about the ability to effectively counteract the negative impacts of media, especially television, on critical thinking and genuine engagement with the world. He seemed to view these trends as deeply entrenched and resistant to easy solutions or remedies. The “Culture Industry” Is the rise of machine learning an example of the “culture industry” as imagined by Adorno & Horkheimer? According to Adorno and Horkheimer, the "culture industry" refers to the standardization and mass production of cultural products, where "the technical contrast between the few production centers and the large number of widely dispersed consumption points is said to demand organization and planning by management" (Adorno & Horkheimer, 1944, p. 171). They argue that this leads to the "universal imposition of this stylized mode" (Adorno & Horkheimer, 1944, p. 187) and the "negation of style" (Adorno & Horkheimer, 1944, p. 187), where "the general can replace the particular, and vice versa" (Adorno & Horkheimer, 1944, p. 178). The rise of machine learning, particularly in the context of AI-generated art, can be seen as an example of the "culture industry" as envisioned by Adorno and Horkheimer. 10 Adorno and Horkheimer argue that the culture industry makes the same kinds of cultural stuff over and over again, and this makes people act the same and not think for themselves (Adorno & Horkheimer, 1944). They state, "The whole world is made to pass through the filter of the culture industry" (Adorno & Horkheimer, 1944, p. 175). In the context of Carlson & Richards’ (2023) article, AI-generated art produced by algorithms can be seen as a manifestation of standardized cultural products. These artworks lack the depth and authenticity of humancreated art and risk perpetuating cultural stereotypes and misappropriation. Furthermore, the article highlights concerns about the mass reproduction of Indigenous art through AI, which aligns with Adorno and Horkheimer's critique of the culture industry's homogenizing effects on culture (Carlson & Richards, 2023). Adorno and Horkheimer also acknowledge that the culture industry can also serve as a site of resistance and critical reflection (Adorno & Horkheimer, 1944). They suggest that genuine art can emerge within the confines of the culture industry, although it is not very common (Adorno & Horkheimer, 1944). Carlson & Richards article mentions Indigenous-led AI projects that use Indigenous knowledge to address environmental challenges and preserve biodiversity (2023). These initiatives represent a form of resistance to the dominant culture industry by incorporating authentic Indigenous perspectives and challenging the homogenizing effects of mainstream AI applications. Additionally, the article advocates for greater Indigenous inclusion in AI development, which could lead to more diverse and culturally sensitive applications of machine learning, countering the homogenizing tendencies of the culture industry (Carlson & Richards, 2023). 11 Adorno and Horkheimer believe that the culture industry makes art seem like something you can buy, just like any other product. This means real artistic expression gets turned into something that is just for selling to lots of people (Adorno & Horkheimer, 1944). Machine learning, especially in AI-generated art, blurs the line between human creativity and what algorithms can do. This makes us wonder if art is just becoming something to buy and sell, and it makes us think about how real or genuine cultural expression is in this context. The article's discussion of Indigenous data sovereignty and the need for Indigenous inclusion in AI development highlights the tension between the commodification of data and the preservation of cultural integrity (Carlson & Richards, 2023). While machine learning has the potential to perpetuate the culture industry's homogenizing effects, it also presents opportunities for marginalized communities to reclaim control over their cultural stories and resist commodification. Are there any benefits to machine learning’s emergence in society that Adorno and Horkheimer might not have predicted? Is it, for example, giving power back to consumers to use and enjoy media in ways they were not prepared to discuss? Adorno and Horkheimer argue that the culture industry manipulates and controls consumer demand, treating them as passive statistics to be categorized. However, Carlson & Richards (2023) article suggests that machine learning and AI are being used by some Indigenous communities to empower themselves and assert control over their own data and cultural expression. For example, the article describes how Indigenous rangers in Kakadu and Sanikiluaq are using AI and data in collaboration with their own traditional knowledge to monitor and protect their environments (Carlson & Richards, 2023). This suggests machine learning can be used by marginalized groups to serve their own interests, rather than just being a 12 tool of manipulation by those in power. Additionally, the article mentions Indigenous-owned companies like Indigital that are using technologies like augmented reality to help Indigenous people tell their own stories on their own terms (Carlson & Richards, 2023). This indicates machine learning may be giving consumers more ability to engage with and shape media, rather than just passively consuming standardized cultural products. These examples suggest machine learning can potentially empower consumers to resist this manipulation and assert their own cultural sovereignty. So while Adorno and Horkheimer were highly critical of the standardizing and controlling effects of the culture industry, the emergence of machine learning may potentially be creating new opportunities for marginalized groups to use these technologies in ways that challenge the top-down control they described. The Oppositional Gaze I don’t believe that this article seems to be describing the potential for a moment of "rupture" or subversion of the dominant gaze, as described by Bell Hooks. Hooks discusses how the "gaze" has always been political for black women, who were often punished or repressed for looking directly at others (Hooks, 2015). This created an "oppositional gaze" as a form of agency and resistance (Hooks, 2015). As Hooks states, "Even in the worse circumstances of domination, the ability to manipulate one's gaze in the face of structures of domination that would contain it, opens up the possibility of agency" (Hooks, 2015, p. 116). Carlson & Richards (2023) article describes several initiatives where Indigenous communities are actively collaborating with AI and technology in ways that center their own knowledge, priorities, and data sovereignty (Carlson & Richards, 2023). For example, the Kakadu rangers are using AI and Indigenous knowledge to monitor and protect their environment, and the Inuit community in Sanikiluaq is using a combination of scientific data and 13 Indigenous knowledge to manage their fisheries (Carlson & Richards, 2023). These examples suggest that Indigenous peoples are not passively accepting the imposition of AI and technology, but are actively shaping its development and use to serve their own interests. This aligns with Hooks’ concept of the "oppositional gaze" - where marginalized groups assert agency and resist domination by critically engaging with and reshaping the systems that seek to control them (Hooks, 2015). Additionally, Carlson & Richards (2023) mentions Indigenous-owned companies like Indigital that are using technologies like augmented reality to help Indigenous people tell their own stories on their own terms. This shows that there is a chance to change the stories about technology and media, which are mostly controlled by white men. However, the article also acknowledges the significant challenges, noting that AI is currently a white male-dominated field with racial and gender biases (Carlson & Richards, 2023). This suggests that fully shifting the relations of power may be difficult in the current state of the industry. I believe Hooks would see these initiatives as promising examples of the potential for rupture and resistance, but would likely caution that much more work is needed to truly center Indigenous knowledges, data sovereignty, and creative expression within the development of AI and emerging technologies. As Hooks states, "Critical, interrogating black looks were mainly concerned with issues of race and racism, the way racial domination of blacks by whites overdetermined representation. They were rarely concerned with gender" (Hooks, 2015, p. 117). Similarly, Carlson & Richards (2023) suggest that the initiatives described are primarily focused on addressing the racial biases and lack of Indigenous representation in AI, rather than explicitly centering gender. So while there is potential for subversion and a shift in power relations, Hooks may argue that a more intersectional approach is needed to fully realize the transformative possibilities, one that takes on with the interplay of race, gender, and other axes of oppression. 14 Culture and Machine Learning Structure of Feeling According to Williams (2014), the "structure of feeling" refers to the characteristic elements of impulse, restraint, and tone; specifically affective elements of consciousness and relationships that are expressed in a particular culture or period, often in an unconscious or implicit way. Williams argues that the "structure of feeling" is not formally learned, but rather emerges organically as a new generation responds to the "unique world it is inheriting" (Williams, 2014, p. 8). Example one of a "Structure of Feeling” The deep connection of Indigenous art and culture to land, history and identity, which would be lacking in AI-generated Indigenous art. This reflects a structure of feeling around the importance of cultural preservation and self-determination for Indigenous peoples. Carlson & Richards (2023) article states, "Indigenous art is embedded with history and connects to culture and Country. AI-created Indigenous art would lack this. There are also implications for financial gain bypassing Indigenous artists and going to the producers of the technology" (para. 2). This aligns with Williams' description of the "structure of feeling". It reflects the structure of feeling by capturing the underlying attitudes, values, and sentiments of Indigenous peoples regarding the deep connection between their art, culture, and identity. The deep connection of Indigenous art to land, history, and identity depicts the significance of maintaining cultural authenticity and integrity. The worry about AI-created Indigenous art missing this link shows a bigger feeling in Indigenous groups about keeping their cultural heritage safe and under their control. Example two of a “Structure of Feeling” 15 The desire to prevent the exploitation and misappropriation of Indigenous art, which has long been an issue even before the rise of AI. This reflects a structure of feeling around the need to protect Indigenous cultural rights and intellectual property. Carlson & Richards (2023) state, "Even before AI, Aboriginal art has widely been appropriated and reproduced without attribution or acknowledgement, particularly for tourism industries. And this could worsen with people now being able to generate art through AI" (para. 2). This also connects to Williams' point about the "structure of feeling" emerging organically as a new generation responds to the "unique world it is inheriting” (Williams, 2014, p. 8). It exemplifies the structure of feeling by capturing the collective sentiments and concerns surrounding the exploitation and misappropriation of Indigenous art. The desire to stop Indigenous art from being unfairly used or taken shows a bigger feeling in Indigenous communities about keeping their cultural rights and ideas safe. This idea has existed long before the emergence of AI, as indicated by the historical issue of Aboriginal art being appropriated and reproduced without proper attribution or acknowledgment, particularly within tourism industries. The mention of AI potentially exacerbating this issue further shows the ongoing relevance of the structure of feeling in responding to contemporary challenges and developments (Carlson & Richards, 2023). As Williams suggests, the structure of feeling emerges organically as each generation responds to the unique circumstances and challenges of their time (Williams, 2014). In this case, the concern over AI-generated art contributing to the exploitation of Indigenous cultural heritage reflects a contemporary manifestation of the broader structure of feeling regarding the protection and preservation of Indigenous rights and identity. 16 How is machine learning, as described in the case connected to the documentary, ideal, and social components of culture, as identified by Williams? According to Williams, there are three main definitions of culture. Ideal Williams describes the ideal component of culture as encompassing the values, aspirations, and beliefs that shape a society's worldview and guide its members' behavior (Williams, 2014). This component includes norms, ethics, and ideologies (Williams, 2014). The "ideal" where culture is "a state or process of human perfection, in terms of certain absolute or universal values" (Williams, 2014, p. 2). In the context of AI-generated Indigenous art, the ideal component relates to the underlying principles and goals driving the development and use of AI technology. The ideals of Indigenous peoples in protecting their cultural heritage and ensuring respect for their intellectual property rights align with broader societal values of cultural diversity, equity, and justice. This suggests that the development and application of AI technology should adhere to these ideals and broader societal values to promote cultural preservation and respect for Indigenous rights. Documentary Williams defines the documentary component as the aspect of culture that involves the documentation and representation of social experiences, practices, and values (Williams, 2014). He notes that it includes tangible expressions such as literature, art, and historical artifacts (Williams, 2014). The "documentary" where culture is "the body of intellectual and imaginative work, in which, in a detailed way, human thought and experience are variously recorded" (Williams, 2014, p. 3). With AI-generated Indigenous art, machine learning serves as a tool for creating new forms of artistic expression (Carlson & Richards, 2023). It 17 generates artworks based on algorithms trained on existing data, thereby documenting the interplay between technology, culture, and creativity (Carlson & Richards, 2023). These AIgenerated artworks become part of the cultural record, reflecting the technological advancements and societal implications of AI in the realm of Indigenous art. Social Williams defines the social component of culture as encompassing interpersonal relationships, social structures, and institutions that mediate social interactions and shape collective identities (Williams, 2014). This component includes patterns of behavior, power dynamics, and social hierarchies within society (Williams, 2014).. He states that it is where a culture is "a description of a particular way of life, which expresses certain meanings and values not only in art and learning but also in institutions and ordinary behaviour” (Williams, 2014, p. 3). Within the social component, there are patterns of behavior and power dynamics that impact cultural production. When it comes to AI-generated art, there are power imbalances between the various people and groups involved in the process. For example, there may be power imbalances between Indigenous artists and AI developers, or between policymakers and Indigenous communities. These dynamics influence decisions about the creation, ownership, and representation of Indigenous art through AI technology. Institutions, such as art galleries, museums, and technology companies, play a crucial role in mediating social interactions and shaping cultural practices. These institutions contribute to the production, distribution, and reception of artworks. How these institutions operate and the norms they uphold influence the accessibility, representation, and recognition of Indigenous art in the digital landscape. Race and Machine Learning 18 Gilroy's Perspective on Machine Learning and Anti-Racism I believe that Gilroy would likely be deeply skeptical of machine learning's capacity to erase identity divisions connected to race or gender and to be truly anti-racist. His main argument says that machine learning is based on the same ideas about race and how things look that in a way he is telling us to move away from (Gilroy, 1998). He critiques how "race" became fused with modern scientific knowledge production, stating "Raciology becomes vulnerable where we can attack the anachronistic perceptual habits to which it was bound" (Gilroy, 1998, p. 840). By this, he means that racial thinking and hierarchies were enabled by particular ways of perceiving and visually representing human differences that emerged with fields like anatomy, natural history, and biology (Gilroy, 1998). Carlson & Richards (2023) article links with this idea by highlighting machine learning's reliance on visual data, which could perpetuate these outdated ways of perceiving human differences. Specifically, the article mentions that AI-generated Indigenous art lacks the history, culture, and connection to Country embedded in authentic Indigenous art (Carlson & Richards, 2023). Furthermore, it raises concerns about misappropriation of Indigenous art and the bypassing of Indigenous artists for financial gain (Carlson & Richards, 2023). This depicts that machine learning, by reproducing art without understanding its cultural significance, fails to acknowledge and respect the complexities of Indigenous identity and culture. Gilroy also argues racial thinking originated from an "epidermalizing" logic of making the body's surface meaningful through racial categories like skin color (Gilroy, 1998)..The way machine learning relies on data to understand differences between people might keep using the same flawed way of thinking. He states, "To renounce race for analytical purposes is not to judge all appeals 19 to it in the profane world of political cultures as formally equivalent" (Gilroy, 1998, p. 842). This suggests Gilroy would reject any method, like machine learning, that reinforces racial categories and makes them seem more real. Example one from Gilroy's work that support his skepticism of machine learning Gilroy’s statement that, "Raciology required that enlightenment and myth be intertwined. Indeed, 'race' and nationality supplied the logic and mechanism of their interconnections" reflects his critique of how racial thinking became embedded in various aspects of modern knowledge production (Gilroy, 1998, p. 843). This aligns with his skepticism about whether machine learning can go beyond separating people based on race or gender. Carlson & Richards (2023) article highlights the concern that AI, which relies on external input for training, is currently dominated by white males. It states, “A global study found 12% of professionals across all levels were female, with only 4% being people of colour” (Carlson & Richards, 2023, para. 6). Furthermore, the article discusses racial biases evident in AI, such as its use in predicting criminal behavior (Carlson & Richards, 2023). These examples show how Gilroy's criticism holds true because even though technology is getting better, racial biases and unfair rankings still exist in machine learning. Example two from Gilroy's work that support his skepticism of machine learning Machine learning could be seen as a new "mechanism" intertwining scientific rationality with racial mythologies. Gilroy states, "The idea of 'race' leaked out of the lofty confines where what Foucault calls the 'chemical gaze' was most fully developed" (Gilroy, 1998, p. 844). This shows that ideas of race are not only social constructs but are also deeply ingrained in scientific practices and methodologies. This aligns with the 20 observation that AI is heavily influenced by the data it is provided. Carlson & Richards (2023) highlights that AI, being a product of its training data, reflects the biases and perspectives embedded within that data. Specifically, the article notes that AI is currently dominated by white males, which raises concerns about racial and gender biases in AI systems (Carlson & Richards, 2023). This depicts how machine learning, like the scientific practices Gilroy critiques, can perpetuate racial mythologies and reinforce existing identity divisions. Example three from Gilroy's work that support his skepticism of machine learning Gilroy argues racial thinking emerged from ways of seeing, which machine learning's focus on visual data analysis could replicate (Gilroy, 1998). He states, "The eugenic ravings of Francis Crick, the Nobel-prizewinning co-discoverer of DNA, demonstrate exactly how the fateful change of scale involved in...molecular biology...was recognized as having cataclysmic moral and political consequences" (Gilroy, 1998, p. 839). This suggests that Gilroy is cautious about how new scientific or technological tools might create new kinds of racial stereotypes. This also aligns with the concern raised in the machine learning article about AI's reliance on visual data analysis, which could replicate the visual and ways of seeing that underpin racial thinking. Carlson & Richards (2023) article mentions again that AI is limited by the data that inform it and highlights the dominance of white males in the AI industry (Carlson & Richards, 2023). This situation reflects broader societal biases and inequalities. By focusing on visual data analysis, AI systems may inadvertently reinforce these biases, perpetuating racial and gender inequalities. Therefore, the machine learning process, which relies heavily on visual data, could contribute to the reproduction of racial mythologies and perpetuate 21 identity divisions connected to race or gender. This backs up Gilroy's doubts about whether machine learning can move past separating people based on identity, and emphasizes the importance of carefully thinking about how AI technologies affect social and racial relationships. Potential Solutions Proposed by Gilroy As a potential solution, Gilroy argues for "renouncing" race analytically and developing new "affirmative" political visions beyond racial identities. For machine learning, this could involve Incorporating diverse knowledges beyond Western scientific traditions into the training data and model development, as highlighted in the Indigenous knowledges article (Carlson & Richards, 2023). He would suggest focusing on making sure that the processes behind AI are clear and accountable. This means setting up systems to find and fix any racial biases or predictable patterns built into the models, like what the Australian government's AI paper suggests (Carlson & Richards, 2023). In addition, he would advocate for using a variety of different fields of study, not just technology, to understand and tackle the deeper cultural, social, and historical reasons behind racial ideas, aligning with Gilroy's "planetary process" that suggests that addressing these issues requires more than just technological interventions (Gilroy, 1998, p. 840). I think the feasibility of such solutions is questionable given Gilroy's critique of how racial ideas are deeply part of how we make modern knowledge nowadays. It might be quite difficult to completely separate machine learning from these old ideas. Ultimately, Gilroy thinks that while technology can change things for the better, he's not sure if machine learning can really fight against racism unless it completely changes the old racial ways of thinking it comes from. And that's a big challenge because machine 22 learning has deep roots in those ways of thinking.