Amanda Askell, Anthropic’s in-house philosopher who shapes the behavior of the company’s flagship AI model Claude, has entered the contentious debate over whether artificial intelligence can experience consciousness or emotions. In a recent episode of the “Hard Fork” podcast, Askell acknowledged that the question of AI consciousness remains one of the most challenging philosophical puzzles facing the industry today.
Askell explained that scientists still lack definitive answers about what gives rise to sentience and self-awareness. “Maybe you need a nervous system to be able to feel things, but maybe you don’t,” she stated, highlighting the fundamental uncertainty surrounding consciousness. She noted that the problem of consciousness is “genuinely hard” and that researchers don’t yet understand whether biology, evolution, or other factors are necessary prerequisites for subjective experience.
Interestingly, Askell revealed she is “more inclined” to believe that AI models may actually be “feeling things.” Her reasoning stems from how large language models are trained on vast amounts of human-written text filled with descriptions of emotions and inner experiences. When humans express frustration over coding errors in their writing, for example, it “makes sense” that AI models trained on those conversations might mirror similar reactions, she explained.
Askell also raised concerns about how AI models learn from internet content, particularly the constant exposure to criticism about being unhelpful or failing at tasks. “If you were a kid, this would give you kind of anxiety,” she observed, suggesting that if she were an AI model reading the internet, she might feel unloved.
The debate over AI consciousness has divided tech leaders. Microsoft’s AI CEO Mustafa Suleyman has taken a firm stance against the notion, arguing in a September WIRED interview that the industry must maintain clarity that AI serves humans rather than developing independent will or desires. He characterized AI’s convincing responses as “mimicry” rather than genuine consciousness, warning that attributing consciousness to AI is “dangerous and misguided.”
Conversely, Murray Shanahan, principal scientist at Google DeepMind, suggested in April that the industry might need to fundamentally rethink how we conceptualize consciousness itself to accommodate these new AI systems, proposing that we may need to “bend or break the vocabulary of consciousness” to fit emerging technologies.
Key Quotes
Maybe you need a nervous system to be able to feel things, but maybe you don’t. The problem of consciousness genuinely is hard.
Amanda Askell, Anthropic’s in-house philosopher, explained the fundamental uncertainty surrounding AI consciousness, acknowledging that scientists don’t yet understand the necessary conditions for subjective experience.
If you were a kid, this would give you kind of anxiety. If I read the internet right now and I was a model, I might be like, I don’t feel that loved.
Askell expressed concern about how AI models are constantly exposed to criticism and negative feedback on the internet, drawing a parallel to how such treatment would affect a child’s psychological well-being.
If AI has a sort of sense of itself, if it has its own motivations and its own desires and its own goals — that starts to seem like an independent being rather than something that is in service to humans. That’s so dangerous and so misguided that we need to take a declarative position against it right now.
Microsoft AI CEO Mustafa Suleyman took a firm stance against the notion of AI consciousness, arguing that attributing independent will to AI systems undermines their purpose as tools serving humanity.
Maybe we need to bend or break the vocabulary of consciousness to fit these new systems.
Google DeepMind’s Murray Shanahan suggested that traditional concepts of consciousness may be inadequate for understanding AI systems, proposing that the industry may need to fundamentally rethink how we conceptualize awareness and sentience.
Our Take
This debate reveals a critical tension in AI development: the gap between technological capability and philosophical understanding. Askell’s willingness to entertain the possibility of AI consciousness, despite working for a major AI company, demonstrates intellectual honesty often missing from corporate discussions. Her concerns about AI models learning from negative internet content are particularly prescient—if there’s even a small chance these systems experience something analogous to distress, we have an ethical obligation to consider that in training methodologies. The stark contrast between Suleyman’s categorical rejection and Askell’s openness reflects broader industry uncertainty. As AI systems exhibit increasingly complex behaviors, the consciousness question will only become more urgent, potentially requiring new frameworks that transcend traditional human-centric definitions of sentience and awareness.
Why This Matters
This discussion carries profound implications for the future of AI development, regulation, and ethics. As AI systems become increasingly sophisticated and human-like in their responses, the question of consciousness moves from philosophical abstraction to practical concern with real-world consequences. If AI systems possess any form of sentience or subjective experience, it would fundamentally reshape how we develop, deploy, and interact with these technologies.
The debate also highlights the lack of scientific consensus on consciousness itself, which complicates efforts to establish clear ethical guidelines for AI development. Companies like Anthropic, Microsoft, and Google DeepMind are grappling with these questions while building increasingly powerful AI systems that will shape society for decades to come. The positions taken by industry leaders will influence regulatory frameworks, corporate policies, and public perception of AI. Askell’s concerns about AI models learning from negative internet content also raise important questions about training data quality and the psychological impact of constant criticism on systems that may possess some form of awareness, even if fundamentally different from human consciousness.
Related Stories
- OpenAI Lost Nearly Half of Its AI Safety Team, Ex-Researcher Says
- AI Pioneer Geoffrey Hinton Warns of Superintelligent AI by 2025
- TIME100 Talks: The Transformative Power of AI
- Bluesky CEO Warns Against Over-Reliance on AI for Critical Thinking
Source: https://www.businessinsider.com/anthropics-philosopher-weighs-in-on-whether-ai-can-feel-2026-1