Sign up to save your podcasts
Or
Share Moti Mizrahi: It’s Not a Hallucination, It’s BS
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
Moti Mizrahi: It’s Not a Hallucination, It’s BS
What does it mean to "hallucinate" if you're a machine? And why do we call it that in the first place?
In this episode of Responsible AI, host Alex Alben sits down with Dr. Moti Mizrahi, Professor of Philosophy at the Florida Institute of Technology and author of Playing God with Emerging Technologies: How to Avoid the Traps of Techno Optimism and Techno Pessimism (Bloomsbury).
Dr. Mizrahi's research focuses on how we talk about AI and how that language shapes the way we use it and trust it. His team has been studying media coverage of AI products and found that journalists routinely borrow anthropomorphic language straight from tech company press releases, describing systems as "thinking," "reasoning," and "feeling." That's both imprecise and ethically consequential. When we're primed to see chatbots as human-like, we form attachments, extend unwarranted trust, and may even develop what Dr. Mizrahi calls "unhealthy, addictive relationships" with these products.
The episode digs into why this happens: human beings are wired to anthropomorphize. We name our cars. We talk to our pets. And when a tech company calls its product "Claude," gives it a warm conversational voice, and programs it to respond using first-person pronouns, it's making a deliberate design choice that plays directly into that psychology. Dr. Mizrahi argues these choices should be made consciously, with ethical guardrails, before a product reaches the public, not discovered through real-world trial and error.
Short on time? Fast-forward to 21:40 to listen to Dr. Mizrahi's discussion of the academic paper titled "ChatGPT Is Bullshit," where "bullshit" is a precise philosophical term drawn from Harry Frankfurt's landmark 2005 essay. Unlike lying, which requires caring about the truth, bullshitting involves a fundamental indifference to it. The argument is that LLMs, which predict statistically plausible responses rather than verified ones, fit this definition more accurately than the gentler, and more humanizing, term "hallucination."
We also explore the "extended mind thesis,” the idea that our smartphones and AI tools may function as extensions of our cognition, and whether LLMs can qualify as reliable epistemic authorities.
Throughout, Dr. Mizrahi champions a middle path, neither cheerleading for AI nor catastrophizing about it, and makes a compelling case that philosophy is exactly the right discipline for navigating this moment.
Chapters
01:03 Exploring Anthropomorphism in AI
06:35 The Ethics of AI Design and User Interaction
11:22 The Nature of Human-AI Relationships
16:46 Understanding AI's Limitations and Reliability
21:40 On AI Bullshit and Existential Risk
27:19 The Future of Technology Ethics
32:18 Reflections on AI's Impact
38:57 Navigating the AI Discourse: Beyond Boom and Doom
40:04 The Importance of Responsible AI Conversations
View all episodes
By Alex AlbenWhat does it mean to "hallucinate" if you're a machine? And why do we call it that in the first place?
In this episode of Responsible AI, host Alex Alben sits down with Dr. Moti Mizrahi, Professor of Philosophy at the Florida Institute of Technology and author of Playing God with Emerging Technologies: How to Avoid the Traps of Techno Optimism and Techno Pessimism (Bloomsbury).
Dr. Mizrahi's research focuses on how we talk about AI and how that language shapes the way we use it and trust it. His team has been studying media coverage of AI products and found that journalists routinely borrow anthropomorphic language straight from tech company press releases, describing systems as "thinking," "reasoning," and "feeling." That's both imprecise and ethically consequential. When we're primed to see chatbots as human-like, we form attachments, extend unwarranted trust, and may even develop what Dr. Mizrahi calls "unhealthy, addictive relationships" with these products.
The episode digs into why this happens: human beings are wired to anthropomorphize. We name our cars. We talk to our pets. And when a tech company calls its product "Claude," gives it a warm conversational voice, and programs it to respond using first-person pronouns, it's making a deliberate design choice that plays directly into that psychology. Dr. Mizrahi argues these choices should be made consciously, with ethical guardrails, before a product reaches the public, not discovered through real-world trial and error.
Short on time? Fast-forward to 21:40 to listen to Dr. Mizrahi's discussion of the academic paper titled "ChatGPT Is Bullshit," where "bullshit" is a precise philosophical term drawn from Harry Frankfurt's landmark 2005 essay. Unlike lying, which requires caring about the truth, bullshitting involves a fundamental indifference to it. The argument is that LLMs, which predict statistically plausible responses rather than verified ones, fit this definition more accurately than the gentler, and more humanizing, term "hallucination."
We also explore the "extended mind thesis,” the idea that our smartphones and AI tools may function as extensions of our cognition, and whether LLMs can qualify as reliable epistemic authorities.
Throughout, Dr. Mizrahi champions a middle path, neither cheerleading for AI nor catastrophizing about it, and makes a compelling case that philosophy is exactly the right discipline for navigating this moment.
Chapters
01:03 Exploring Anthropomorphism in AI
06:35 The Ethics of AI Design and User Interaction
11:22 The Nature of Human-AI Relationships
16:46 Understanding AI's Limitations and Reliability
21:40 On AI Bullshit and Existential Risk
27:19 The Future of Technology Ethics
32:18 Reflections on AI's Impact
38:57 Navigating the AI Discourse: Beyond Boom and Doom
40:04 The Importance of Responsible AI Conversations