What do you think?
Rate this book
科学之路：人、机器与未来 当机器思考时，人类会怎样？
“图灵奖”得主、“深度学习三巨头”之一、“卷积神经网络之父”……由于在人工智能领域的突出贡献，杨立昆被中国计算机科学界和企业界所熟知。
杨立昆的科学之路，谱写了一段关于勇气的宣言。他为了知识本身求学，而不是文凭，他用自己的经历，证明了通过激烈的考试竞争进入名校不是科学成功的窄门。他广泛阅读，为他科学思维的形成奠定了坚实的理论基础。他特立独行，做自己感兴趣的事情，即便那件事在短时间里不被人看好。在人工神经网络研究的低谷期，他寂寞地坚持，终于取得了举世瞩目的成就。
人工智能正在颠覆人类社会，未来机器能思考吗？杨立昆的这部著作，讲述正是人工智能在我们面前崛起——这个历史上仅有的时刻发生的故事。
杨立昆的科学之路，谱写了一段关于勇气的宣言。他为了知识本身求学，而不是文凭，他用自己的经历，证明了通过激烈的考试竞争进入名校不是科学成功的窄门。他广泛阅读，为他科学思维的形成奠定了坚实的理论基础。他特立独行，做自己感兴趣的事情，即便那件事在短时间里不被人看好。在人工神经网络研究的低谷期，他寂寞地坚持，终于取得了举世瞩目的成就。
人工智能正在颠覆人类社会，未来机器能思考吗？杨立昆的这部著作，讲述正是人工智能在我们面前崛起——这个历史上仅有的时刻发生的故事。
340 pages, Paperback
First published October 16, 2019
30 people are currently reading
176 people want to read
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 13 of 13 reviews
May 6, 2020
La lecture est très intéressante. Il y a un réel effort de rendre accessible le propos. C'est, je pense, plutôt réussi.
Le chapitre 3 expliquant la rétropropagation n'est pas des plus évidents. La notation type informatique n'est pas un très bon choix (pour les dérivées partielles notamment, c'est particulièrement indigeste).
J'aurais aimé que l'employeur de l'auteur soit un peu moins cité.
Vers la fin du livre Le Cun aborde tout un tas de sujets ayant trait au futur de l'AI. Il y expose son opinion sans vraiment d'argumentation.
Malgré tout ça, c'est un excellent livre qui demande un réel effort de la part du lecteur.
Le chapitre 3 expliquant la rétropropagation n'est pas des plus évidents. La notation type informatique n'est pas un très bon choix (pour les dérivées partielles notamment, c'est particulièrement indigeste).
J'aurais aimé que l'employeur de l'auteur soit un peu moins cité.
Vers la fin du livre Le Cun aborde tout un tas de sujets ayant trait au futur de l'AI. Il y expose son opinion sans vraiment d'argumentation.
Malgré tout ça, c'est un excellent livre qui demande un réel effort de la part du lecteur.
December 29, 2025
très bon livre sur l‘IA
ne soyons pas prétentieux, je n'ai pas tout compris... mais ce que j'ai compris est passionnant!
ne soyons pas prétentieux, je n'ai pas tout compris... mais ce que j'ai compris est passionnant!
February 13, 2020
(At the moment this is only available in the original French, but I'm sure a translation to English is on its way.)
This book is many things at once.
Firstly, it is a solid, fairly accessible, and very up-to-date introduction to deep learning for people with a STEM background. This said, I may be underestimating the difficulty of certain parts of the book, because I am specialised in machine learning. Some parts were much too easy for me; but LeCun proceeds to go through a wide range of mature AI techniques, some of which I knew very little about, in a simple and intuitive way. So people with some elementary knowledge of maths and programming, and a professional or amateur interest in deep learning, will find that worth their time.
Secondly, it is a professional autobiography: "me, Yann Lecun, my life and work". This will be interesting to those looking for a career in AI or another high-tech field, and probably awfully boring to anybody else.
Thirdly, it's a potpourri of LeCun's opinions on a variety of topics related to AI. Some are clever and interesting. Others are already widely held in the AI community and beyond, and therefore not interesting. Some are, I think, plain wrong, reflecting LeCun's own biases. Finally, some are just rushed sloppily. So for instance Lecun, who works for Facebook, thinks Facebook is great and that the problems caused by AI in social media have been solved. Likewise, there are snippets about machine consciousness, military usage, etc., that reflect a lack of engagement with (or a misunderstanding of) these issues.
All in all, this is definitely worth a read if you work in AI or if you're passionate about the topic. Otherwise, I'm not so sure.
This book is many things at once.
Firstly, it is a solid, fairly accessible, and very up-to-date introduction to deep learning for people with a STEM background. This said, I may be underestimating the difficulty of certain parts of the book, because I am specialised in machine learning. Some parts were much too easy for me; but LeCun proceeds to go through a wide range of mature AI techniques, some of which I knew very little about, in a simple and intuitive way. So people with some elementary knowledge of maths and programming, and a professional or amateur interest in deep learning, will find that worth their time.
Secondly, it is a professional autobiography: "me, Yann Lecun, my life and work". This will be interesting to those looking for a career in AI or another high-tech field, and probably awfully boring to anybody else.
Thirdly, it's a potpourri of LeCun's opinions on a variety of topics related to AI. Some are clever and interesting. Others are already widely held in the AI community and beyond, and therefore not interesting. Some are, I think, plain wrong, reflecting LeCun's own biases. Finally, some are just rushed sloppily. So for instance Lecun, who works for Facebook, thinks Facebook is great and that the problems caused by AI in social media have been solved. Likewise, there are snippets about machine consciousness, military usage, etc., that reflect a lack of engagement with (or a misunderstanding of) these issues.
All in all, this is definitely worth a read if you work in AI or if you're passionate about the topic. Otherwise, I'm not so sure.
June 7, 2021
As a PhD student in deep learning, I really enjoyed reading this book. I think many aspects of it complement what we're usually taught or what we are usually confronted with, and it's inspiring to know the past and possible future of the field.
The only thing I found disappointing was that the explanations of complex algorithms like backpropagation were not very clear, and the "pythonish" notation only complicated things further.
Nevertheless, this is a wonderful book to read for anyone interested or even already working on artificial intelligence.
The only thing I found disappointing was that the explanations of complex algorithms like backpropagation were not very clear, and the "pythonish" notation only complicated things further.
Nevertheless, this is a wonderful book to read for anyone interested or even already working on artificial intelligence.
June 20, 2024
This is yet another book by one of the most renowned AI scientists, and it's somewhat surprising to find that the book (originally written in French) has a Chinese translation but not an English one.
While it was expected to be a biography, the book actually consists mostly of short essays about his personal history and career milestones, along with some introductions to ConvNet and other models. The sections feel somewhat disjointed, and although some personal anecdotes are included, they often feel detached and similar to what you might find in a Wikipedia article. Midway through the book, the discussion on learning models also comes across as a bit scattered. The last chapters don't quite deliver either, with few topics being thoroughly explored.
That said, the book does shed light on the vicissitudes of artificial intelligence research. It outlines how researchers endured the "AI winter" from a personal perspective, recovering from being considered outliers and heretics to becoming some of the most recognized scientists today. Though not recommended for the general audience, if you're a researcher with some preliminary knowledge of deep learning, you may still find some inspiration from this.
While it was expected to be a biography, the book actually consists mostly of short essays about his personal history and career milestones, along with some introductions to ConvNet and other models. The sections feel somewhat disjointed, and although some personal anecdotes are included, they often feel detached and similar to what you might find in a Wikipedia article. Midway through the book, the discussion on learning models also comes across as a bit scattered. The last chapters don't quite deliver either, with few topics being thoroughly explored.
That said, the book does shed light on the vicissitudes of artificial intelligence research. It outlines how researchers endured the "AI winter" from a personal perspective, recovering from being considered outliers and heretics to becoming some of the most recognized scientists today. Though not recommended for the general audience, if you're a researcher with some preliminary knowledge of deep learning, you may still find some inspiration from this.
September 1, 2023
Yann Le Cun est au machine learning ce que le pape est a l'eglise catholique. Du coup, quand il ecrit un bouquin sur le sujet, ca attire les lecteurs.
J'avais tres peur quand j'ai achete le livre. J'avais peur que l'auteur prenne soit tellement de raccourcis que ca devienne risible. Et a l'inverse, j'avais peur que l'auteur veuille rendre la partie technique trop compliquee et detaillee et ne perde les lecteurs.
J'ai ete vraiment tres agreablement surpris et le livre est un excellent melange entre (1) la carriere de Yann dans le domaine (2) les concepts et les explications techniques du machine learning et (3) l'etat de l'art et l'avenir. C'est extremement difficile d'arriver a tel equilibre et Yann a reussi l'exercice.
Je ne peux que recommender ce livre a quiconque souhaite s'informer sur le machine learning. Non seulement l'auteur est tres qualifie mais le contenu et tres accessible. Je pense que cela restera un ouvrage de reference, au moins dans la langue francaise.
J'avais tres peur quand j'ai achete le livre. J'avais peur que l'auteur prenne soit tellement de raccourcis que ca devienne risible. Et a l'inverse, j'avais peur que l'auteur veuille rendre la partie technique trop compliquee et detaillee et ne perde les lecteurs.
J'ai ete vraiment tres agreablement surpris et le livre est un excellent melange entre (1) la carriere de Yann dans le domaine (2) les concepts et les explications techniques du machine learning et (3) l'etat de l'art et l'avenir. C'est extremement difficile d'arriver a tel equilibre et Yann a reussi l'exercice.
Je ne peux que recommender ce livre a quiconque souhaite s'informer sur le machine learning. Non seulement l'auteur est tres qualifie mais le contenu et tres accessible. Je pense que cela restera un ouvrage de reference, au moins dans la langue francaise.
June 6, 2022
Piles of books have been written about AI, primarily by outsiders: philosophers, journalists, linguists, sociologists, business coaches, and the like. This book is an extremely rare case when a generally-accessible text on AI is produced by a world-class computer scientist and one of the inventors of modern neural networks. This fact alone should be a sufficient motivation to peak up the tome when it’s finally translated into English. But there’s one more important reason: being Chief AI Scientist at Meta Yann LeCun has an insider view on Facebook’s policies and infrastructure that defines which news you see or don’t see on your Facebook feed. It's sobering to know what personalities stand behind the algorithms that substituted law these days, which ethical code they subscribe to, and how they imagine our shared future.
November 20, 2019
L'auteur est sans doute un génie Dan's son domaine, mais la lecture de son ouvrage n'est pas aisée. Les concepts s'enchaînent rapidement et rapidement le lecteur, même volontaire et intéressé a de la difficulté à suivre. Rendez-vous raté en ce qui me concerne et je le regrette.
April 26, 2022
Très bon livre qui permet d'avoir une vue globale sur ce que c'est l'IA mais aussi de son histoire avec le point de vue d'un des acteurs majeurs. Quelques passages sont un peu technique mais on peut très bien les sauter.
August 11, 2022
For a non-technical person it offers a couple of chapters of AI for dummies which are very interesting. There is a lot of added value in the enthusiasm of Yann in writing about innovations where he has been a part of or close by.
January 2, 2024
Ah, now I understand why the author is on twitter arguing that books should be free downloads with no payment to authors -- they want the data for training. It seems a short-sighted position, though, still.
November 11, 2021
Ideas for the best reading order: https://mikelaud.github.io/
January 8, 2022
Some parts are quite technical and fascinating. Some are too high level and quite vague.
Displaying 1 - 13 of 13 reviews