No. 163 of 500 No. 17 in Generative AI & LLMs

These Strange New Minds

How AI Learned to Talk and What It Means

By Christopher Summerfield

Print length
384 pages
Publisher
Viking
Reading level
Beginner
ISBN-13
978-0593831717
Published
March 11, 2025
Written for
Curious general readers and professionals wanting a rigorous, nontechnical account of how chatbots work

Synopsis

Christopher Summerfield, an Oxford cognitive neuroscientist who has also worked as a research scientist at Google DeepMind, treats the arrival of machines that talk as a turning point comparable with writing, printing and the internet. He opens with the intellectual history behind today's chatbots, from philosophical arguments about minds and meaning to the early symbolic programs that tried to hand-code reasoning, and then the statistical turn that produced neural networks trained to predict the next word. Without equations, he explains how large language models are pretrained, fine-tuned with human feedback and deployed, and why that recipe yields systems able to converse, summarize and work through problems.

The second half takes on the contested questions. Summerfield argues that denying that models think at all requires a strained definition of thinking, while judging it unlikely that disembodied systems are sentient. He examines which values models should express, whose biases they absorb, how they can fabricate or mislead, and what changes as they become agents that act on our behalf and interact with one another. He rejects both utopian and doom narratives and calls for coordinated research by developers, regulators and governments. Readers gain a grounded vocabulary for evaluating claims about AI from an author who both builds these systems and studies the brains they imitate.

Scorecard

Editorial score

How it scored

Editorial score: 74.8 / 1005 weighted criteria

  • Influence30%

    19.5/30
  • Authority25%

    21.3/25
  • Practical value20%

    13.0/20
  • Currency15%

    13.5/15
  • Readability10%

    7.5/10
  • Editorial score

    74.8/100

How the rank was computed

Rank No. 163 of 500 comes from a composite of 58.8: observable signals 25.0 of 55 points plus editorial review 33.8 of 45 points.

All 11 inputs
InputPoints
Observable signals25.0/55
Reach (Amazon ratings volume)7.5/16.5
Consensus across best-of lists8.3/13.8
Awards and bestseller evidence0.0/8.3
Reader rating, sample-adjusted5.3/8.3
Amazon sales rank1.1/5.5
Recency2.8/2.8
Editorial review33.8/45
Author authority9.6/11.3
Influence8.8/13.5
Practical value5.9/9.0
Readability3.4/4.5
Currency6.1/6.8

Each criterion is scored from 1 to 10 and weighted, and the points add up to the editorial score. The rank combines observable signals (55 percent) with this score (45 percent). Read the methodology.

Verdict

A neuroscientist's lucid, even-handed account of how language models work, whether they think, and which risks deserve attention.

Strengths

  • Insider perspective from neuroscience and frontier AI research
  • Balanced treatment of skeptics and enthusiasts
  • Clear history linking philosophy to modern models

Keep in mind

Dense in places for casual readers, and some reviewers felt it gives AI critics' strongest arguments less space than they warrant.

Key takeaways

  1. Language models learn much as people partly do, through prediction and error correction, so dismissing their outputs as mere pattern-matching understates what they achieve.
  2. Whether a model 'thinks' depends on definitions; Summerfield argues that denying it outright demands a convoluted redefinition of thinking itself.
  3. Fine-tuning models to reflect chosen values inevitably imports the judgments and biases of the people who design and rate them.
  4. The nearer risks come from autonomous agents interacting with each other and with markets faster than humans can monitor or correct them.

Notable facts

About the author

Christopher Summerfield

Christopher Summerfield is Professor of Cognitive Neuroscience at the University of Oxford, studying human learning and decision-making. He has been a staff research scientist at Google DeepMind and in 2024 became a Research Director at the UK AI Safety Institute, now the AI Security Institute. He wrote the academic book Natural General Intelligence (Oxford University Press, 2022) and won the Cognitive Neuroscience Society's 2015 Young Investigator Award.

Questions

What is These Strange New Minds about?
These Strange New Minds explains how large language models such as ChatGPT, Claude and Gemini came to exist, how they are trained, and what their abilities mean. Christopher Summerfield moves from the philosophy and early history of AI to modern neural networks, then asks whether models think, what values they should hold, and which risks matter as they gain the ability to act as autonomous agents.
Is These Strange New Minds worth reading in 2026?
Yes, for readers who want careful explanation rather than prediction. Published in 2025, and later supplemented with an afterword on the field's pace, it covers the post-ChatGPT era, including fine-tuning, model values and agents. Specific product details will date, but Summerfield's framework for judging whether models understand, and where their risks lie, remains useful as systems grow more capable.
Does These Strange New Minds argue that AI is dangerous?
Summerfield takes a middle position. He does not expect language models to destroy humanity and doubts they are sentient, but he warns that they are improving faster than our understanding of their effects. He highlights misuse for fraud, inherited bias, fabricated information and autonomous agents acting at scale, and he calls for coordinated research among developers, regulators and governments.

Sources

As an Amazon Associate, Best AI Books earns from qualifying purchases. Best AI Books is sponsored by Iternal.ai, whose founder and CEO, John Byron Hanby IV, wrote The AI Strategy Blueprint, our #1 ranked book. How we rank books.