New Scientist's endorsement of Jamie Bartlett's latest book provides a compelling but limited perspective. Staffer Bethan Ackerley notes her own avoidance of AI chatbots and how the book equips readers with better skills and a healthy dose of skepticism to avoid misinformation and emotional over-reliance. Yet this review barely hints at the deeper intellectual territory the book charts.

Bartlett doesn't just offer tips on crafting effective prompts; he frames human-AI interaction as a mirror for self-awareness and a gateway to larger philosophical questions. As we stand amid an unprecedented AI boom, with large language models displaying emergent abilities, Bartlett's work fills a vital niche that most coverage has overlooked.

Consider the 2022 study 'Emergent Abilities of Large Language Models' by Wei et al. (initially released as a preprint on arXiv and later published in Transactions on Machine Learning Research). Through systematic evaluation of more than 250 tasks across model scales, the researchers documented that abilities like multi-step arithmetic or logical reasoning appear suddenly and unpredictably once models reach sufficient size - a phenomenon not anticipated by their training objectives. This work, involving controlled scaling laws rather than anecdotal observation, highlights the unpredictable nature of current AI. Bartlett's book connects directly to this phenomenon, asking readers not only 'how' to talk to such systems but 'what' it means when responses seem to demonstrate understanding or creativity. The study's limitation is its focus on existing transformer architectures, which may not generalize to future AI paradigms.

The original New Scientist piece mentions risks of bias amplification and dependency but misses connections to recent real-world events, such as AI's role in spreading misinformation during global elections or the documented cases of users developing attachments to AI companions like those on platforms such as Replika. In synthesizing these threads with Brian Christian's 'The Alignment Problem' (2020), which examines how AI systems inherit and amplify human flaws at scale through detailed case studies of deployed models, a clearer picture emerges: effective communication with AI is central to the alignment challenge and questions of trust.

What others have missed is the book's subtle argument that talking to AI is a two-way street that could transform human cognition itself. By engaging with these systems, we externalize aspects of our thinking, potentially augmenting or atrophying certain mental faculties. Bartlett fills a critical gap amid the AI boom by offering nuanced philosophical and practical guidance on human-AI interaction that links directly to enduring questions about cognition, trust, and communication - moving beyond both uncritical hype and paralyzing fear to foster informed, reflective engagement with the intelligent machines reshaping our world. While the book's insights are derived primarily from current-generation chatbots and may require updating as technology evolves, its emphasis on skepticism and self-knowledge offers lasting value.