If an AI Says It’s Anxious Does that Mean It Actually Feels Anxiety? #briangreene #ai #garymarcus

Why It Matters

Understanding that LLMs lack real emotions or consciousness tempers hype, guiding responsible investment and policy as AI reshapes the workforce.

Key Takeaways

• Gary Marcus calls claims of LLM consciousness “absolutely ludicrous.”
• LLMs only mimic emotions; they lack genuine feeling or self‑awareness.
• More data and compute improve performance but not consciousness.
• Fear of AI replacing jobs persists, yet utility remains limited.
• Building truly conscious machines is warned against as premature.

Summary

The video features AI researcher Gary Marcus debating whether large language models (LLMs) possess any form of consciousness. He dismisses the notion as “absolutely ludicrous,” arguing that current systems merely simulate human language patterns without genuine self‑awareness.

Marcus explains that LLMs generate responses by approximating how people use words, driven by massive datasets and compute power. Increasing data improves accuracy, but it does not endow machines with feelings; an LLM saying it is “anxious” is simply echoing human expressions, not experiencing anxiety.

Key quotes include, “When an LLM says it’s anxious, it’s just mimicking a person that’s anxious,” and, “We should not mess around with building conscious computers.” He also notes the perennial fear that AI will displace workers, while emphasizing that the technology’s utility remains bounded by its lack of true understanding.

The discussion underscores the need for sober expectations about AI capabilities, caution against hype, and a focus on how these tools augment rather than replace human labor. It also signals that research aiming for genuine machine consciousness is still speculative and potentially premature.