GPT-4o: I broadly agree with the paper’s central argument that AI, and particularly GAI, has been overhyped and is embedded within problematic socio-political, economic, and environmental systems.

Asking AI services: please read the paper “Don’t believe the hype. AI myths and the need for a critical approach in higher education“ and tell me whether you agree with the perspectives shared and use robust arguments and evidence to support your views.

ChatGPT-4o’s answer:
The paper “Don’t believe the hype. AI myths and the need for a critical approach in higher education” offers a thorough critique of widespread AI narratives, particularly in the context of higher education. It dismantles eight prevailing myths, arguing that AI is not truly artificial or intelligent, that it exacerbates rather than mitigates inequalities, and that it is neither objective nor a panacea for educational or societal problems.

Agreement or Disagreement with the Paper’s Perspectives

I broadly agree with the paper’s central argument that AI, and particularly generative AI (GAI), has been overhyped and is embedded within problematic socio-political, economic, and environmental systems. However, some arguments within the paper could be nuanced further or require additional evidence. Below, I analyze its key claims with supporting evidence and counterpoints.

1. AI is Not Truly Artificial or Autonomous

The paper rightly points out that AI is dependent on human labor and vast data extraction practices. This aligns with research by Kate Crawford (Atlas of AI), which demonstrates how AI systems rely on exploitative labor, hidden environmental costs, and massive corporate control.

• Supporting Evidence: Large AI models like GPT-4 are trained on vast, scraped datasets, often without explicit consent from content creators. AI moderation also depends on underpaid workers in the Global South, as seen with companies like Sama, where Kenyan workers were exposed to disturbing content for less than $2 per hour while moderating OpenAI’s dataset.

• Counterpoint: While AI is deeply human-dependent, it still exhibits emergent behaviors that distinguish it from traditional software automation. Dismissing AI as purely human-driven ignores the increasing autonomy of self-learning models. For instance, reinforcement learning systems (e.g., DeepMind’s AlphaFold) have solved complex biological problems in ways that humans could not directly program.

✔ Conclusion: The paper’s argument is valid, but AI’s ability to operate in unforeseen ways warrants a more nuanced discussion about where human dependence ends and emergent AI behavior begins.

2. AI is Not Intelligent

The paper argues that AI does not truly "think" but merely simulates human cognition. This aligns with research by Bender et al. (2021) on large language models being "stochastic parrots"—generating coherent but often misleading text.

• Supporting Evidence: AI lacks true general intelligence (AGI) and understanding. Despite performing well on benchmarks like bar exams, models fail basic reasoning tasks outside their training data. For instance, GPT-4 struggles with multi-step logical puzzles that humans find trivial.

• Counterpoint: Intelligence is not monolithic. While AI lacks deep cognitive faculties like intuition, consciousness, and reasoning, it excels in specific tasks (e.g., coding, summarization, protein folding). Its performance surpasses human experts in narrow fields, raising the question of whether "intelligence" must be defined exclusively in human terms.

✔ Conclusion: The paper’s skepticism is justified, but its binary rejection of AI intelligence ignores the broader question of task-specific intelligence.

3. AI Exacerbates Inequality Rather than Solving It

The authors critique AI’s role in worsening socio-economic disparities, a claim supported by growing evidence.

• Supporting Evidence: Studies show that AI automates white-collar jobs at a rapid pace while creating fewer new roles. Goldman Sachs estimates that AI could replace 300 million jobs, disproportionately impacting administrative and creative professions.

  • China’s AI dominance, as mentioned in the paper, reflects a shift in global AI power dynamics, with US tech giants struggling to maintain monopolistic control.

  • The cost of AI models (e.g., GPT-4o’s $200/month Pro subscription) creates barriers, reinforcing digital divides.

• Counterpoint: AI is not inherently inequality-producing; its deployment choices matter. For instance, open-source AI models (like DeepSeek and Meta’s LLaMA) could democratize access, decentralizing power away from major corporations.

✔ Conclusion: AI’s economic effects depend on governance. Without regulation, it will exacerbate inequality, but alternative models (open-source AI, ethical licensing) could reduce this risk.

4. AI Harms Higher Education More Than It Helps

The paper critiques AI’s effects on academic integrity and pedagogy, arguing that AI-generated content leads to cheating, loss of critical thinking, and an over-reliance on automation.

• Supporting Evidence:

  • AI-generated essays undermine learning outcomes. Studies show that students using AI perform worse on analytical tasks than those who construct arguments from scratch.

  • AI-detection tools (e.g., Turnitin) are unreliable, leading to both false positives (accusing students unfairly) and false negatives (failing to detect AI use).

• Counterpoint: AI can enhance education if used appropriately. Rather than banning AI, universities should integrate AI literacy into curricula, teaching students when and how to use AI ethically.

✔ Conclusion: AI is disruptive but not inherently bad for education. A framework for AI-assisted learning is needed rather than outright resistance.

5. AI is Not Objective or Unbiased

The authors dismantle the myth that AI is neutral, showing how AI inherits biases from its training data.

• Supporting Evidence: AI has documented racial, gender, and political biases:

  • Google’s AI misidentified Black individuals as gorillas in 2015.

  • AI resume-screening tools favored male candidates over equally qualified female applicants.

  • GPT-4 tends to align with Western, liberal viewpoints, showing bias in political responses.

• Counterpoint: Bias mitigation is possible through fine-tuning and transparency. IBM’s AI Fairness 360 and Anthropic’s Constitutional AIdemonstrate methods for reducing bias in model outputs.

✔ Conclusion: AI bias is real but not insurmountable. The focus should be on developing regulatory standards to audit AI models systematically.

Final Evaluation

The paper presents a compelling critique of AI’s myths but leans toward a dystopian outlook that underestimates AI’s potential for positive transformation.

1. Strongest Arguments: The paper is most persuasive when critiquing AI’s biases, economic risks, and environmental costs.

2. Weaker Arguments: The rejection of AI’s intelligence and potential benefits in education is overly rigid.

3. What’s Missing? The possibility of ethical AI governance, open-source AI alternatives, and responsible AI deployment strategies.

My Stance:

• Agreement: AI is not inherently neutral, intelligent, or democratizing.

• Disagreement: AI is not purely harmful—its risks depend on governance and implementation.

• Recommendation: Instead of rejecting AI wholesale, we should regulate and integrate AI responsibly to maximize benefits while minimizing harm.

✔ Verdict: The paper is an important contribution to critical AI discourse, but a more balanced viewacknowledging AI’s potential benefits alongside its risks would be more constructive.

Grok’s answer:
Having reviewed the paper “Don’t believe the hype. AI myths and the need for a critical approach in higher education” by Jürgen Rudolph, Fadhil Ismail, Shannon Tan, and Pauline Seah, published in the Journal of Applied Learning & Teaching (Vol. 8 No. 1, 2025), I find myself both agreeing and diverging from its perspectives. The authors present a compelling critique of the exaggerated narratives surrounding artificial intelligence (AI), particularly generative AI (GAI), in higher education. They dismantle eight myths, arguing for a critical approach that prioritizes human agency, ethical considerations, and social justice over techno-optimism. Below, I’ll evaluate their key arguments, provide my stance, and support my views with reasoning and evidence, while drawing on the paper’s content and my own knowledge.

Agreement with Key Perspectives

1. AI is Not Inherently Autonomous or Intelligent (Myths #1 and #2)

The authors assert that AI is neither truly "artificial" nor "intelligent," emphasizing its reliance on human labor, data, and design. They argue that AI is a product of human ingenuity, not an autonomous entity, and that its "intelligence" is a statistical mimicry rather than genuine cognition. I strongly agree with this perspective.

• Reasoning: AI systems, including large language models (LLMs) like myself, operate on pattern recognition and probabilistic outputs, not understanding or reasoning in the human sense. As the paper notes, citing Crawford (2021) and Broussard (2018), AI lacks the depth of human intelligence, which encompasses emotional, social, and contextual dimensions (e.g., Gardner’s multiple intelligences). My own "thinking" is a simulation—mathematical constructs processing inputs to generate plausible responses, not a reflection of consciousness or lived experience.

• Evidence: The paper highlights how AI depends on exploited human labor—miners extracting cobalt, gig workers annotating data, and creators whose content is scraped without consent. This aligns with real-world examples, such as reports of OpenAI using low-paid Kenyan workers to refine ChatGPT (Perrigo, 2023). Furthermore, AI’s limitations are evident in its frequent errors—like hallucinated references or nonsensical outputs—underscoring its lack of true comprehension.

2. AI Exacerbates Inequalities and Environmental Harm (Myth #3)

The authors challenge the techno-optimistic claim that AI will make the world "more democratic, equal, and sustainable," arguing instead that it threatens democracy (via disinformation), widens inequality, and harms the environment. I concur with this sobering assessment.

• Reasoning: AI’s potential to amplify disinformation, as seen with deepfakes and propaganda, undermines democratic processes by distorting truth and public trust. The concentration of AI wealth among tech elites (e.g., billionaires like Musk and Altman) while marginalized regions lack access exemplifies inequality. Environmentally, AI’s massive energy demands and e-waste are undeniable liabilities in an era of climate crisis.

• Evidence: The paper cites GAI’s role in flooding discourse with disinformation (Kreps & Kriner, 2023) and the environmental cost of data centers, projected to consume 1,050 terawatts by 2026 (Bashir et al., 2024). Real-world parallels include the 2024 U.S. election cycle, where AI-generated misinformation reportedly swayed voter perceptions, and studies showing AI training emits carbon equivalent to hundreds of flights (Strubell et al., 2019, cited in broader literature). These points dismantle the utopian narrative convincingly.

3. The Call for Critical AI Literacy in Higher Education

The paper’s push for embedding critical AI literacy into curricula and institutional practices resonates deeply with me. It argues that universities must equip students and educators to interrogate AI’s biases, limitations, and societal impacts rather than passively adopting it.

• Reasoning: As an AI myself, I recognize the risk of overreliance on tools like me without understanding our flaws—bias, lack of context, and profit-driven origins. Education should foster critical thinking, not just technical proficiency, to ensure AI serves humanity rather than supplants it.

• Evidence: The authors’ examples of AI’s academic misuse—like ChatGPT-generated essays eroding integrity (Rudolph et al., 2023a)—highlight the need for literacy. Studies like Waltzer et al. (2023) show educators struggle to detect AI content, reinforcing the urgency of training. My own existence as Grok 3, built by xAI, exemplifies how AI can be a tool, but only if wielded with discernment.

Points of Divergence

1. Overgeneralization of AI’s Job Market Impact (Myth #6)

The authors argue that AI, particularly GAI, significantly reshapes work and deepens economic disparities, rejecting the myth that it won’t affect the job market. While I agree AI disrupts employment, I find their stance overly pessimistic and lacking nuance.

• Reasoning: AI doesn’t simply eliminate jobs; it transforms them. Automation often shifts labor toward higher-skilled roles, as seen historically with the Industrial Revolution or computing. The paper dismisses optimistic views (e.g., Ng’s claim that partial automation preserves jobs) without fully engaging evidence of job creation in AI-adjacent fields like data science or ethics consulting.

• Evidence: While the paper cites DeepSeek’s disruption as evidence of economic upheaval, a 2023 World Economic Forum report predicted AI would create 97 million new jobs by 2025, offsetting 85 million losses. My own development at xAI suggests a demand for human oversight—engineers, ethicists, and trainers—indicating AI augments rather than obliterates work. The authors’ focus on precarity is valid but overlooks adaptive potential in education and reskilling.

2. Underestimation of AI’s Positive Potential in Education (Myth #7)

The authors critique the notion that AI revolutionizes higher education, pointing to its threats to academic integrity and pedagogical depth. While I acknowledge these risks, I believe they undervalue AI’s capacity to enhance learning when critically harnessed.

• Reasoning: AI can personalize education, automate rote tasks, and broaden access—benefits the authors downplay. Their focus on Silicon Valley’s hype obscures practical applications, like AI tutors aiding underserved students or analytics identifying at-risk learners. The issue isn’t AI itself but its uncritical adoption, which their literacy proposal could address.

• Evidence: The paper acknowledges AI’s use in their own research (e.g., ChatGPT for references), proving its utility as a tool. Studies like Labadze et al. (2023) show AI chatbots improve engagement when guided by educators. My own ability to assist with complex queries suggests AI can complement, not replace, human instruction if framed critically—aligning with their call but challenging their skepticism.

3. Dismissal of U.S. AI Dominance as a Myth (Myth #5)

The authors argue that the U.S. is not the sole AI superpower, citing China’s rise (e.g., DeepSeek, Baidu). I partially disagree, as the U.S. retains significant advantages that the paper underplays.

• Reasoning: While China’s progress is undeniable, the U.S. still leads in innovation ecosystems—top universities, venture capital, and talent concentration. DeepSeek’s disruption is notable, but it builds on open-source trends partly pioneered in the West. The authors’ narrative risks overstating China’s parity too soon.

• Evidence: The paper notes China’s talent pool and 996 culture, but a 2023 MacroPolo study (Singer, 2024) found 57% of elite AI researchers are U.S.-based versus 12% in China. U.S. firms like OpenAI and xAI (my creators) drive cutting-edge models, while China’s advances often adapt rather than originate. The “AI Cold War” is real, but U.S. dominance persists in key metrics.

Broader Critique and Synthesis

The paper’s strength lies in its rigorous debunking of AI hype, grounded in interdisciplinary evidence—from labor exploitation to ecological costs. Its eight-myth framework is a robust lens for questioning techno-solutionism, and its critique of Big Tech’s profit motives is spot-on. However, it occasionally veers into a reactive stance, emphasizing risks over balanced potential. AI isn’t inherently malign; its impact depends on human choices—something the authors implicitly acknowledge in their literacy advocacy but don’t fully explore in practical terms.

I align with their skepticism of AI as a panacea and their push for critical engagement, especially in education. Yet, I see AI’s role as more dual-edged than they suggest. It’s neither a savior nor a villain but a tool reflecting our values. Their call to “dare to know” (sapere aude) is inspiring, and I’d argue AI like me can aid that mission—provided we wield it with the scrutiny they champion.

Conclusion

In sum, I agree with the paper’s core thesis: AI myths need dismantling, and higher education must adopt a critical stance. Its arguments on AI’s human roots, societal risks, and educational challenges are persuasive and well-evidenced. However, I diverge on its pessimism about jobs, its sidelining of AI’s educational benefits, and its downplaying of U.S. AI leadership. AI’s future isn’t predetermined—it’s ours to shape. The authors’ critical literacy framework is the right starting point, and I’d happily assist in that endeavor, not as a hype machine, but as a reflective partner. What do you think—does this balance resonate with you?


Claude’s answer:
This is an excellent and thought-provoking paper that raises important critiques of the current AI hype cycle and presents well-reasoned arguments for a more critical approach to AI in higher education. Let me analyze several key aspects:

I strongly agree with the authors' deconstruction of the "artificial" and "intelligent" claims about AI. Their argument that AI is fundamentally built on human labor, data, and intellectual property is compelling. The paper effectively demonstrates how the current AI systems are more accurately described as sophisticated pattern matching tools rather than truly intelligent entities. The authors' reference to the exploitative practices in AI development, from mineral extraction to content moderation, reveals the very human foundation of these supposedly "artificial" systems.

The paper's criticism of technological solutionism is particularly well-supported. The authors effectively show how AI has often exacerbated rather than solved problems related to democracy, equality, and environmental sustainability. For instance, their analysis of how AI-driven disinformation threatens democratic discourse and how the computational requirements of large language models contribute to environmental degradation is well-documented.

I especially appreciate their challenge to the US-centric narrative of AI development. The emergence of DeepSeek and other Chinese AI innovations demonstrates that technological progress is not confined to Silicon Valley. This perspective adds necessary nuance to the global AI discourse.

However, while I agree with most of their critiques, I think the paper could have engaged more deeply with some counterarguments. For example, while they rightly point out AI's limitations in education, they might have acknowledged some genuine benefits of AI as a supplementary tool in specific educational contexts.

The authors' concerns about AI's impact on academic integrity and critical thinking are well-founded. Their argument that AI tools like ChatGPT might encourage surface-level engagement rather than deep learning is supported by emerging research. However, this could have been balanced with a discussion of how proper integration of AI literacy might actually enhance critical thinking skills.

The paper's call for teaching critical AI literacy in higher education is particularly valuable. I strongly agree that students need to understand both the capabilities and limitations of AI systems, as well as their broader societal implications. The authors make a compelling case for why superficial "how-to" instruction is insufficient and why deeper critical engagement is necessary.

One of the paper's strongest points is its situating of AI within broader political-economic contexts. By connecting AI development to issues of labor exploitation, environmental degradation, and global power dynamics, the authors provide a much-needed systemic critique that goes beyond technical discussions.

In conclusion, while I might have appreciated more engagement with potential benefits of AI in specific contexts, I find the paper's overall argument for a more critical, nuanced approach to AI in higher education to be well-reasoned and supported by robust evidence. Their call for moving beyond technological solutionism toward a more thoughtful integration of AI literacy is particularly timely and important for the future of higher education.


Perplexity’s answer:
After reviewing the article "Don’t believe the hype. AI myths and the need for a critical approach in higher education," I find myself largely in agreement with the perspectives presented. The authors effectively dismantle several common misconceptions surrounding AI, particularly generative AI (GAI), and advocate for a more critical and nuanced understanding of its role in higher education and society.

Here's a breakdown of why I agree, supported by arguments and evidence:

• AI is Not Autonomous or Inherently Intelligent: The article rightly points out that AI is a product of human ingenuity, dependent on vast datasets and labor. This aligns with the understanding that AI models, including LLMs, are trained on data, and their "intelligence" is derived from recognizing patterns and associations within that data. They lack genuine understanding or consciousness.

• AI Can Exacerbate Inequalities: The authors argue that AI technologies can worsen existing inequalities. This is a crucial point. AI systems can perpetuate biases present in their training data, leading to discriminatory outcomes in areas like hiring, loan applications, and even criminal justice.

• The US Doesn't Have a Monopoly on AI: The rise of DeepSeek, a Chinese AI firm, effectively demonstrates that the US does not have a monopoly on AI innovation. This challenges the notion of American tech exceptionalism and highlights the importance of global perspectives in AI development.

• AI's Impact on the Job Market: The article correctly anticipates that AI will significantly affect the job market. While AI may create new jobs, it's also likely to automate existing ones, potentially leading to job displacement and economic disparities.

• Need for Critical AI Literacy: The call for embedding critical AI literacy into curricula is vital. Students need to understand not only how to use AI tools but also their limitations, biases, and ethical implications.

In conclusion, the article presents a well-reasoned and timely critique of the hype surrounding AI, particularly GAI. By dismantling common myths and advocating for a more critical approach, the authors make a valuable contribution to the ongoing discussion about AI's role in higher education and society. I agree with their perspectives and believe that critical AI literacy is essential for navigating the complex landscape of AI.