What do you think?
Rate this book
The AI Delusion
We live in an incredible period in history. The Computer Revolution may be even more life-changing than the Industrial Revolution. We can do things with computers that could never be done before, and computers can do things for us that could never be done before.
But our love of computers should not cloud our thinking about their limitations.
We are told that computers are smarter than humans and that data mining can identify previously unknown truths, or make discoveries that will revolutionize our lives. Our lives may well be changed, but not necessarily for the better. Computers are very good at discovering patterns, but are useless in judging whether the unearthed patterns are sensible because computers do not think the way humans think.
We fear that super-intelligent machines will decide to protect themselves by enslaving or eliminating humans. But the real danger is not that computers are smarter than us, but that we think computers are smarter than us and, so, trust computers to make important decisions for us.
The AI Delusion explains why we should not be intimidated into thinking that computers are infallible, that data-mining is knowledge discovery, and that black boxes should be trusted.
But our love of computers should not cloud our thinking about their limitations.
We are told that computers are smarter than humans and that data mining can identify previously unknown truths, or make discoveries that will revolutionize our lives. Our lives may well be changed, but not necessarily for the better. Computers are very good at discovering patterns, but are useless in judging whether the unearthed patterns are sensible because computers do not think the way humans think.
We fear that super-intelligent machines will decide to protect themselves by enslaving or eliminating humans. But the real danger is not that computers are smarter than us, but that we think computers are smarter than us and, so, trust computers to make important decisions for us.
The AI Delusion explains why we should not be intimidated into thinking that computers are infallible, that data-mining is knowledge discovery, and that black boxes should be trusted.
256 pages, Hardcover
Published October 1, 2018
17 people are currently reading
330 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 - 30 of 35 reviews
October 1, 2018
This is a very important little book ('little' isn't derogatory - it's just quite short and in a small format) - it gets to the heart of the problem with applying artificial intelligence techniques to large amounts of data and thinking that somehow this will result in wisdom.
Gary Smith as an economics professor who teaches statistics, understands numbers and, despite being a self-confessed computer addict, is well aware of the limitations of computer algorithms and big data. What he makes clear here is that we forget at our peril that computers do not understand the data that they process, and as a result are very susceptible to GIGO - garbage in, garbage out. Yet we are increasingly dependent on computer-made decisions coming out of black box algorithms which mine vast quantities of data to find correlations and use these to make predictions. What's wrong with this? We don't know how the algorithms are making their predictions - and the algorithms don't know the difference between correlation and causality.
The scientist's (and statistician's) mantra is often 'correlation is not causality.' What this means is that if we have two things happening in the world we choose to measure - let's call them A (it could be banana imports) and B (it could number of pregnancies in the country) and if B rises and falls as A does, it doesn't mean that B is caused by A. It could be that A is caused B, A and B are both caused by C, or it's just a random coincidence. The banana import/pregnancy correlation actually happened in the UK for a number of years after the second world war. Human statisticians would never think the pregnancies were caused by banana imports - but an algorithm would not know any better.
In the banana case there was probably a C linking the two, but because modern data mining systems handle vast quantities of data and look at hundreds or thousands of variables, it is almost inevitable that they will discover apparent links between two sets of information where the coincidence is totally random. The correlation happens to work for the data being mined, but is totally useless for predicting the future.
This is the thesis at the heart of this book. Smith makes four major points that really should be drummed into all stock traders, politicians, banks, medics, social media companies... and anyone else who is tempted to think that letting a black box algorithm loose on vast quantities of data will make useful predictions. First, there are patterns in randomness. Given enough values, totally random data will have patterns embedded within it - it's easy to assume that these have a meaning, but they don't. Second, correlation is not causality. Third, cherry picking is dangerous. Often these systems pick the bits of the data that work and ignore the bits that don't - an absolute no-no in proper analysis. And finally, data without theory is treacherous. You need to have a theory and test it against the data - if you try to derive the theory from the data with no oversight, it will always fit that data, but is very unlikely to be correct.
My only problems with book is that Smith insists for some reason on making databases two words ('data bases' - I know, not exactly terrible), and the book can feel a bit repetitious because most of it consists of repeated examples of how the four points above lead AI systems to make terrible predictions - from Hillary Clinton's system mistakenly telling her team where to focus canvassing effort to the stock trading systems produced by 'quants'. But I think that repetition is important here because it shows just how much we are under the sway of these badly thought-out systems - and how much we need to insist that algorithms that affect our lives are transparent and work from knowledge, not through data mining.
As Smith points out, we regularly hear worries that AI systems are going to get so clever that they will take over the world. But actually the big problem is that our AI systems are anything but intelligent: 'In the age of Big Data, the real danger is not that computers are smarter than us, but that we think computers are smarter than us and therefore trust computers to make important decisions for us.’
This should be big-selling book. A plea to the publisher: change the cover (it just looks like it's badly printed and smudged) and halve the price to give it wider appeal.
Gary Smith as an economics professor who teaches statistics, understands numbers and, despite being a self-confessed computer addict, is well aware of the limitations of computer algorithms and big data. What he makes clear here is that we forget at our peril that computers do not understand the data that they process, and as a result are very susceptible to GIGO - garbage in, garbage out. Yet we are increasingly dependent on computer-made decisions coming out of black box algorithms which mine vast quantities of data to find correlations and use these to make predictions. What's wrong with this? We don't know how the algorithms are making their predictions - and the algorithms don't know the difference between correlation and causality.
The scientist's (and statistician's) mantra is often 'correlation is not causality.' What this means is that if we have two things happening in the world we choose to measure - let's call them A (it could be banana imports) and B (it could number of pregnancies in the country) and if B rises and falls as A does, it doesn't mean that B is caused by A. It could be that A is caused B, A and B are both caused by C, or it's just a random coincidence. The banana import/pregnancy correlation actually happened in the UK for a number of years after the second world war. Human statisticians would never think the pregnancies were caused by banana imports - but an algorithm would not know any better.
In the banana case there was probably a C linking the two, but because modern data mining systems handle vast quantities of data and look at hundreds or thousands of variables, it is almost inevitable that they will discover apparent links between two sets of information where the coincidence is totally random. The correlation happens to work for the data being mined, but is totally useless for predicting the future.
This is the thesis at the heart of this book. Smith makes four major points that really should be drummed into all stock traders, politicians, banks, medics, social media companies... and anyone else who is tempted to think that letting a black box algorithm loose on vast quantities of data will make useful predictions. First, there are patterns in randomness. Given enough values, totally random data will have patterns embedded within it - it's easy to assume that these have a meaning, but they don't. Second, correlation is not causality. Third, cherry picking is dangerous. Often these systems pick the bits of the data that work and ignore the bits that don't - an absolute no-no in proper analysis. And finally, data without theory is treacherous. You need to have a theory and test it against the data - if you try to derive the theory from the data with no oversight, it will always fit that data, but is very unlikely to be correct.
My only problems with book is that Smith insists for some reason on making databases two words ('data bases' - I know, not exactly terrible), and the book can feel a bit repetitious because most of it consists of repeated examples of how the four points above lead AI systems to make terrible predictions - from Hillary Clinton's system mistakenly telling her team where to focus canvassing effort to the stock trading systems produced by 'quants'. But I think that repetition is important here because it shows just how much we are under the sway of these badly thought-out systems - and how much we need to insist that algorithms that affect our lives are transparent and work from knowledge, not through data mining.
As Smith points out, we regularly hear worries that AI systems are going to get so clever that they will take over the world. But actually the big problem is that our AI systems are anything but intelligent: 'In the age of Big Data, the real danger is not that computers are smarter than us, but that we think computers are smarter than us and therefore trust computers to make important decisions for us.’
This should be big-selling book. A plea to the publisher: change the cover (it just looks like it's badly printed and smudged) and halve the price to give it wider appeal.
June 5, 2019
The AI part of this book could fit into a sentence; computers use statistics and statistics without context is nonsense.
The remaining 99% of the book is a (sometimes verbatim) repeat of the material in "Standard Deviations", which is a better written book.
At no time does the author actually discuss the state of AI research beyond the context of narrow expert systems.
The remaining 99% of the book is a (sometimes verbatim) repeat of the material in "Standard Deviations", which is a better written book.
At no time does the author actually discuss the state of AI research beyond the context of narrow expert systems.
February 5, 2022
An econometrics professor's understanding of AI - while in principle correct and in the right direction, limited and not very deep. A much better title for this book would be something like: “the statistical pitfalls of data mining techniques - for college students”. The AI project implies far more than a specified domain like statistics/data mining; while a good book about AI should be more than a correct and conservative summary/limitations of such a field. For example, Bostrom's Superintelligence is quite brilliant and deep - even if the main assumption about the imminence of AI is wrong.
July 26, 2019
While an important premise, arguing that the AI is not the panacea to all the worlds issues and it has problems, that should be discussed, it falls short of its goal. I guess I'm giving 3 stars to it just because it argues this point.
The author mostly discusses the pitfalls of statistical analysis, which, while an important part, is not all that modern AI systems do. Problems with statistical models from suspect data have been known for a very long time, and there is not that much new information about them or how they differ when AI systems are involved. A lot of examples have very little to do with artificial intelligence as we know it today.
If you are interested in a book about statistics, then this is not a bad one, but the same ideas can be found in something like "Weapons of Math Destruction", albeit in a more entertaining way.
If you want to know about the problems with AI, then this is not it.
The author mostly discusses the pitfalls of statistical analysis, which, while an important part, is not all that modern AI systems do. Problems with statistical models from suspect data have been known for a very long time, and there is not that much new information about them or how they differ when AI systems are involved. A lot of examples have very little to do with artificial intelligence as we know it today.
If you are interested in a book about statistics, then this is not a bad one, but the same ideas can be found in something like "Weapons of Math Destruction", albeit in a more entertaining way.
If you want to know about the problems with AI, then this is not it.
October 14, 2024
Written from the perspective of an economist, The AI Delusion seeks to present reason why we ought not fall to the trap of thinking artificial intelligence is infallible. Divided into twelve relatively short chapters, this book is both suitable for and aimed at non-experts in AI. Smith starts with more definitional chapters, focusing on things like the difference between true intelligence and robotic obedience, action without thinking, and the importance of context, which computers can't isolate. He then moves to explain data manipulation through several chapters, focusing on things like bad data (garbage in, garbage out), the meaninglessness of patterns derived from random data, data torturing, and he gives a special focus to the Texas Sharpshooter fallacy - namely, where differences are underplayed and similarities are overemphasized, leading to false conclusions. Smith ends by taking the last few chapters to focus in on specific industries, such as healthcare, financial markets (he does this topic twice), and surveillance.
This has a lot of good information for a non-expert, and seems to me like this could be aimed at a more popular audience without requiring any modifications. The scope was also good for the size of the book and the level of detail in each chapter. However, my biggest problem is that it is very cursory and clearly written by someone who uses AI but doesn't design AI. This leads Smith to be more critical than I think is warranted, especially since the majority of the book feels like it is more of a commentary on statistics and statistical manipulation. The AI elements of the middle and later chapters felt more tacked on - understandably, statistical errors will of course be a problem if we rely on AI that uses flawed data. However, the focus just wasn't as advertised. Additionally, as the book proceeded it became less about information and more about examples - some sections were entirely example without anything else added in between examples.
This has a lot of good information for a non-expert, and seems to me like this could be aimed at a more popular audience without requiring any modifications. The scope was also good for the size of the book and the level of detail in each chapter. However, my biggest problem is that it is very cursory and clearly written by someone who uses AI but doesn't design AI. This leads Smith to be more critical than I think is warranted, especially since the majority of the book feels like it is more of a commentary on statistics and statistical manipulation. The AI elements of the middle and later chapters felt more tacked on - understandably, statistical errors will of course be a problem if we rely on AI that uses flawed data. However, the focus just wasn't as advertised. Additionally, as the book proceeded it became less about information and more about examples - some sections were entirely example without anything else added in between examples.
January 30, 2024
هذا الكتاب أحد أفضل الكتب التقنية التي لابد لكل مثقف من الإطلاع عليها وليس بالضرورة أن تكون متخصصا فقد جنح الكاتب - وهو العالم الكبير - إلى لغة مبسطة وسلسلة تتميز بوضوح العبارة و حسن المثال و صحة التعبير وسلامة النتيجة .. في هذا الكتاب تدرك الحقيقة الواضحة والتي يحاول الجميع تناسيها في خضم موجة الانبهار بالذكاء الصناعي و تطبيقاته ، وهي أن الخوارزميات والحاسوب بكليّته لا يفهم البيانات ولا يعرف معاني الألفاظ و لا يدرك حقائق الأشياء ولا يعي جوابه .. وأن المسألة عبارة عن محركات رياضية تبحث عن الأنماط و تقارب بين الأشباه والنظائر وتتحسس السمت العام دون ادراك أو وعي وإنما في آلية تامة .. وهذه النقطة هي لب الكتاب … بعد ذلك غاص العالم الرياضي والاقتصادي الكبير في عوالم الاحتمالات و دراسة الأنماط من البيانات الضخمة والمتضخمة كل لحظة وبين بالدليل والمثال أن الركون إليها يعني تدمير البشرية بقرارات غير سوية يظنها الجاهل صوابا باعتماد على الأرقام والأنماط دون اعتبار المسببات الغير محدودة -ربما- في كل حالة ، والتي ويا للعجب يقوم دماغنا البشري رغم محدوديته بتحليلها واستيعابها و استبعادها أو تقريبها بصورة سريعة غير ملحوظة وكلما أزداد الإنسان علما تحول عقله إلى ما يفوق الحواسيب …
الذكاء الصناعي والحواسيب والخوارزميات ممتازة في تسهيل الأعمال و تقديم قدرات بحثية لا يستطيعها الإنسان في عمره القصير وجهده القليل لكنها أبدا لن تقارب المعجزة التي نحملها على أكتافنا وربما لن يكون لها وعي وإدراك أبدا - ربما - ..
في الحقيقة لا أحب الخوض في تفاصيل الكتاب و فصوله فهذا نوع من اللغو في حضرة الكتاب ..لكن ما يسعني قوله أن هذا الكتاب لابد أن يطلع عليه الجميع خصوصا من أنصاف المتعلمين الجالسين على كراسي اتخاذ القرار في عالمنا العربي .. ولله في خلقه شؤون
الذكاء الصناعي والحواسيب والخوارزميات ممتازة في تسهيل الأعمال و تقديم قدرات بحثية لا يستطيعها الإنسان في عمره القصير وجهده القليل لكنها أبدا لن تقارب المعجزة التي نحملها على أكتافنا وربما لن يكون لها وعي وإدراك أبدا - ربما - ..
في الحقيقة لا أحب الخوض في تفاصيل الكتاب و فصوله فهذا نوع من اللغو في حضرة الكتاب ..لكن ما يسعني قوله أن هذا الكتاب لابد أن يطلع عليه الجميع خصوصا من أنصاف المتعلمين الجالسين على كراسي اتخاذ القرار في عالمنا العربي .. ولله في خلقه شؤون
August 5, 2019
AI is cool, some people think AI will "save us from our own stupidity" thats just not there yet.
With the help of AI and some random statistics like goat prices in the middle east, best buy's top sold items we can get a patterns that we can use to buy and sell stocks on s&p 500, brilliant.
Anyway good book about the misconceptions and the missuses of current AI systems.
With the help of AI and some random statistics like goat prices in the middle east, best buy's top sold items we can get a patterns that we can use to buy and sell stocks on s&p 500, brilliant.
Anyway good book about the misconceptions and the missuses of current AI systems.
January 12, 2022
I recommend this book for its first half. The latter part isn’t worth listening to. Probably reading it as a text book instead of listening to it makes a difference, because of all the equations, figures and tables that takes up most of the latter part.
October 20, 2019
Książka dobrze wyjaśnia ograniczenia i problemy związane z data mining i uczeniem maszynowym. Problemem jest powtarzalność poszczególnych wątków i słabe wejście w obecne trendy rozwoju metod AI.
January 18, 2020
A.I. is the modern snake oil solution to all your work needs and it's likely to screw things up. Or at least, that's Smith's general argument as he breaks down the fundamental problems with artificial intelligence. One such problem is the fact that even labeling it "intelligence" tricks people into a false sense of security about what AI does. A.I. does not have intelligence or intention and therefore, its ability to do the things that so many people claim it can do is disconcerting and misleading. Essentially, Smith argues that most of AI works akin the Texas Sharp-Shooter sham wherein either a gunslinger shoots a bunch of holes into a wall and then puts the target over one of the shots and say "look, I got a bullseye" or points and shoots at a bullseye and misses, but then moves the target over. Because computers just crunch numbers without any sense of relation or meaning, the end result is that it evitably finds connections. Essentially, all that big data does is torture the data until it confesses with some kind of relationship. The problem as Smith highlights here is that if you put together enough data and mix it up enough, it will create patterns but some of that pattern is the nature of randomization (that is, part of the randomness of randomness is to occasionally have patterns). He demonstrates this consistently throughout by reminding that correlation is not causation but that also correlation is often spurious at best and A.I. does not have intelligence enough to know what is evidently spurious and what isn't. What I like about Smith's book is the critical way he breaks down A.I. and makes its underpinnings a bit more comprehensive. Now, this isn't consistent throughout the book and there are places where I get lost and am just along for the ride. But much of his work is consistently accessible. It also helps that he uses lots of examples to demonstrate his points and shows time and again, how patterns even in meaningless data could be found to be correlative (he routinely shares statistical relationships among different things, only to reveal that he made up the initial numbers and the computer "found" the relationship). However, I think I also get a bit concerned that his argument seems to largely say that all research is bunk or that we should trust any research. On the one hand, I get that, but on the other, it means that we all need to become specialists in everything that we might need to rely on and that just doesn't seem feasible either. Finally, I wish he would have better discussed or addressed the idea that if his argument is true, then how does he explain the places where it at least appears that A.I. is getting things right. Regardless, it's a useful book to consider for anyone who might be encountering AI in their realm and see people blindly accepting it.
August 15, 2019
The AI Delusion (2018) by Gary Smith is an attempt to show how overblown the hype about Artificial Intelligence is. In part it succeeds with some interesting tales of how various AI systems have failed and also by pointing out flaws in just picking correlations. But it also fails by seriously considering the way modern Machine Learning (ML) systems have solved problems that people thought intractable 20 years ago and to consider if there is more progress where it might lead. It is something of an antidote to people who think ML is about to put everyone out of work and lead to the Singularity but it goes far to far the other way.
There are some fascinating stories in the book. In 2008 and 2012 Barack Obama's data analytics team was regarded as great and crucial in helping him win. Few people know that in 2016 Hilary Clinton had an ML system that instructed her that it wasn't worth campaigning in Michigan and a number of states where Obama had done well but she'd done poorly against Bernie Sanders. The book alleges that Bill Clinton thought she was in trouble in these states and was furious but was over ruled by the computer.
The book has other stories like that about trading and other things. But also many that are repeated in other books like the one about where to armor aircraft in WWII.
Smith puts forward the case about how ML can learn specific things but can be easily fooled. He also spends far too many chapters on ML that is over fitted. Introduction to ML courses teach students the dangers and how to avoid that.
The chapters on finance go to far. Correlation and chartism are well known. I've also heard quants on podcasts calmly say quite happily that ML doesn't really work on markets because of false positives. However, something quants do does work. Human traders are being replaced, but no by ML.
The AI Delusion does make a few good points and will have some good stories for most people. It also contains a lot on how statistics can lead people to make mistakes. Smith's recommendation that we don't defer to ML is also pretty wise. However, the case it makes is overdone and Smith spends too much time attacking a straw man of what. Whether ML will lead to job losses and the Singularity is highly questionable, but there is more to this than Smith lets on.
There are some fascinating stories in the book. In 2008 and 2012 Barack Obama's data analytics team was regarded as great and crucial in helping him win. Few people know that in 2016 Hilary Clinton had an ML system that instructed her that it wasn't worth campaigning in Michigan and a number of states where Obama had done well but she'd done poorly against Bernie Sanders. The book alleges that Bill Clinton thought she was in trouble in these states and was furious but was over ruled by the computer.
The book has other stories like that about trading and other things. But also many that are repeated in other books like the one about where to armor aircraft in WWII.
Smith puts forward the case about how ML can learn specific things but can be easily fooled. He also spends far too many chapters on ML that is over fitted. Introduction to ML courses teach students the dangers and how to avoid that.
The chapters on finance go to far. Correlation and chartism are well known. I've also heard quants on podcasts calmly say quite happily that ML doesn't really work on markets because of false positives. However, something quants do does work. Human traders are being replaced, but no by ML.
The AI Delusion does make a few good points and will have some good stories for most people. It also contains a lot on how statistics can lead people to make mistakes. Smith's recommendation that we don't defer to ML is also pretty wise. However, the case it makes is overdone and Smith spends too much time attacking a straw man of what. Whether ML will lead to job losses and the Singularity is highly questionable, but there is more to this than Smith lets on.
August 1, 2019
The AI Delusion by Gary Smith is a welcome and compelling assessment of the core logic of how computers operate and the limits and cautions we must apply when using these magnificent machines. Not hypothetical about controlling the Terminator in 2050, but the role of computers today, here and now.
Smith, a Professor of economics, shows in detail that there are fundamental limits to how well computers can interpret the world, and in turn, significant problems if we seek them to ask more than they can do. The heart of his analysis is examining how computers sort data.
Across scientific and financial fields, he shows, in a very engaging and easy to follow way, how data mining, mathematical analysis and the search for 'statistically significance' can easily lead us astray. And how many people who either don't know better, or worse do know but see profit in the distortions, promote such schemes. Often based on the mistaken idea that because it is 'just data' computer driven results are somehow objective and more reliable than humans. As such, the book serves as one of the best critiques of the cult of quantification that I have encountered.
Smith doesn't want us to stop using machines, nor give up on numbers for our analysis. Rather, he argues the big problem is not that the machines will do things we don't want (ie, Artificial Intelligence taking over the world) and more that machines today can't do what far too many people already expect of them (i.e China's Social Credit system is based on dubious judgements of the characteristics of good or bad people, Wall Street has financial scams which treat inevitable coincidences as tricks to the stock market, and Silicon Valley apps and search systems hide and discriminate, sometimes without even realising it.
To deal with today's problems, and help forestall the worst outcomes tomorrow, we need a more realistic assessment of what computers can, and often can't do. The delusion is not in the AI, but ourselves.
Strongly Recommended.
Smith, a Professor of economics, shows in detail that there are fundamental limits to how well computers can interpret the world, and in turn, significant problems if we seek them to ask more than they can do. The heart of his analysis is examining how computers sort data.
Across scientific and financial fields, he shows, in a very engaging and easy to follow way, how data mining, mathematical analysis and the search for 'statistically significance' can easily lead us astray. And how many people who either don't know better, or worse do know but see profit in the distortions, promote such schemes. Often based on the mistaken idea that because it is 'just data' computer driven results are somehow objective and more reliable than humans. As such, the book serves as one of the best critiques of the cult of quantification that I have encountered.
Smith doesn't want us to stop using machines, nor give up on numbers for our analysis. Rather, he argues the big problem is not that the machines will do things we don't want (ie, Artificial Intelligence taking over the world) and more that machines today can't do what far too many people already expect of them (i.e China's Social Credit system is based on dubious judgements of the characteristics of good or bad people, Wall Street has financial scams which treat inevitable coincidences as tricks to the stock market, and Silicon Valley apps and search systems hide and discriminate, sometimes without even realising it.
To deal with today's problems, and help forestall the worst outcomes tomorrow, we need a more realistic assessment of what computers can, and often can't do. The delusion is not in the AI, but ourselves.
Strongly Recommended.
January 12, 2020
Buen libro con consideraciones muy relevantes para el uso cuidadoso y productivo de la Inteligencia Artificial en la toma de decisiones. A pesar del título provocador, el objetivo del autor no es desacreditar por completo el uso de estas poderosas herramientas sino por el contrario entender sus alcances, en qué situaciones pueden ser útiles y en cuáles no y, cuando sea pertinente su uso, qué debemos tomar en cuenta para sacarles el máximo provecho.
Los avances que se han tenido en Inteligencia Artificial en los últimos años han sido impresionantes y tienen el potencial de afectar prácticamente todas las areas de la actividad humana. Sin embargo, también es cierto que se ha puesto de moda ofrecer servicios de ciencia de datos, usando la etiqueta de IA como herramienta de mercadotecnia, pero que no son más que ejercicios de búsqueda de correlaciones entre variables que pueden no ser relevantes. El peligro a medida que tenemos acceso a más datos (el famoso big data) es que las correlaciones que se puedan encontrar la mayor parte de las veces serán espurias y por lo tanto será una pérdida de tiempo y recursos, o incluso peligroso, el tomar decisiones basadas en los resultados de estos modelos analíticos.
La única razón por la que no le doy cinco estrellas es que el libro puede llegar a ser repetitivo; sin embargo, creo que es lectura obligada para cualquier que quiera desarrollar o utilizar herramientas de Inteligencia Artificial en la toma de decisiones.
Los avances que se han tenido en Inteligencia Artificial en los últimos años han sido impresionantes y tienen el potencial de afectar prácticamente todas las areas de la actividad humana. Sin embargo, también es cierto que se ha puesto de moda ofrecer servicios de ciencia de datos, usando la etiqueta de IA como herramienta de mercadotecnia, pero que no son más que ejercicios de búsqueda de correlaciones entre variables que pueden no ser relevantes. El peligro a medida que tenemos acceso a más datos (el famoso big data) es que las correlaciones que se puedan encontrar la mayor parte de las veces serán espurias y por lo tanto será una pérdida de tiempo y recursos, o incluso peligroso, el tomar decisiones basadas en los resultados de estos modelos analíticos.
La única razón por la que no le doy cinco estrellas es que el libro puede llegar a ser repetitivo; sin embargo, creo que es lectura obligada para cualquier que quiera desarrollar o utilizar herramientas de Inteligencia Artificial en la toma de decisiones.
August 16, 2023
Overall, this book was ok, but I came away feeling a bit let down. There were interesting and valuable chapters, but much of the middle of the book felt quite repetitive (like Smith's mantra about data mining: "Make up model. Test. Change Model. Repeat.") Smith's argument boils down to claiming that artificial intelligence relies on correlations to make predictions. Relying on correlations amidst Big Data leads to many spurious correlations, which result in poor out-of-sample predictions. Therefore, artificial intelligence should not be trusted on it's own. Fair enough, but this is a fairly simple argument which didn't require nearly so many simulation examples to prove how commonplace spurious correlations can be. It didn't help that the writing was mediocre and there were more typos/grammatical mistakes than I am used to encountering in a book published by a reputable press.
One strength of the book was the number of practical examples of AI being used blindly with negative consequences. Some of these were classic examples (e.g. Google Flu), but there were many from the world of finance and investing that I was unfamiliar with and found interesting.
With the recent explosion of LLMs, I was particularly interested in reading this book. Despite my critiques of the book, I am inclined to agree with Smith's general take on AI and as far as I can tell, I think that his argument still holds up.
One strength of the book was the number of practical examples of AI being used blindly with negative consequences. Some of these were classic examples (e.g. Google Flu), but there were many from the world of finance and investing that I was unfamiliar with and found interesting.
With the recent explosion of LLMs, I was particularly interested in reading this book. Despite my critiques of the book, I am inclined to agree with Smith's general take on AI and as far as I can tell, I think that his argument still holds up.
May 24, 2023
This book may be five years old now, so from before the Generative AI / Chat GPT excitement reached full fever pitch, but as this is about the core fundamentals (statistics, probability, contextual understanding), this doesn't really matter. The goal here is to demonstrate why and how machines think the way they do - and so to demonstrate their limitations and encourage rational/critical thinking when working with them.
The basic argument: Correlation does not equal causation; humans are bad at remembering this, but AI bots are even worse, because they have no true conception of wider reality, and so don't really understand what "cause" means. But they're getting very good at acting like they do...
Combine that with the surprising statistical frequency of things that seem like meaningful patterns that tend to crop up when analysing large datasets, and there's a lot of potential for AI to identify false positives without being able to self-correct. And as most AI models are black box algorithms that even their human creators don't fully understand, it's almost impossible for humans to effectively counter this tendency.
The basic argument: Correlation does not equal causation; humans are bad at remembering this, but AI bots are even worse, because they have no true conception of wider reality, and so don't really understand what "cause" means. But they're getting very good at acting like they do...
Combine that with the surprising statistical frequency of things that seem like meaningful patterns that tend to crop up when analysing large datasets, and there's a lot of potential for AI to identify false positives without being able to self-correct. And as most AI models are black box algorithms that even their human creators don't fully understand, it's almost impossible for humans to effectively counter this tendency.
January 23, 2023
Weapons Of Math Destruction by Cathy O'Neil is a better book on this topic, so please go read that book instead. To be clear, the ACTUAL topic of this book is statistics, data mining, Big Data, financial 'quants,' technical analysis, and dumb algorithms given broad trust over human judgement. This is NOT a book about AI or machine learning per se, so much as a book about the layers of data and data manipulation utilized by AI systems, regular computer algorithms, and even regular humans.
The best parts of the book are where the author sets up the reader with patterns in data and then reveals that the data is completely random. It's a gag that's funny the first three times. The next three times he persists in this demonstration, it's funny on a meta level of being abused by the author as the reader willing to keep on reading 🥸
The theme of the book is important but can really be summarized in a short paragraph. The book was more a tediously long essay than a (hopefully) short book. I gave it 3 stars as I was feeling kind ✨
The best parts of the book are where the author sets up the reader with patterns in data and then reveals that the data is completely random. It's a gag that's funny the first three times. The next three times he persists in this demonstration, it's funny on a meta level of being abused by the author as the reader willing to keep on reading 🥸
The theme of the book is important but can really be summarized in a short paragraph. The book was more a tediously long essay than a (hopefully) short book. I gave it 3 stars as I was feeling kind ✨
November 27, 2023
The book's title is misleading because the author's criticism is not specifically targeted at A.I. Instead, it focuses on the practice of “data mining” and the discovery of spurious correlations. The author emphasizes that humans are still superior to machines in contextualizing and evaluating statistical outcomes in realistic situations. This critique applies generally to science, not just to A.I, however. The book offers a collection of stories that highlight how data mining has led people astray and provides better criticism of data science and the hype around "big data" as the next big thing. However, if you are looking for content specifically related to A.I., there may not be much in the book. Reading it as a collection of essays raising concerns around data science, it is enjoyable.
July 31, 2019
Starts off on the wrong foot -- and then shifts into being about big data, randomness, and people's absolute shit intuition.
It's funny that he keeps saying humans can recognize when a correlation doesn't pass the sniff test, tho, since this book is about how, in practice, no, we cannot.
Also reminds me of a smart-ass thing I like to say: So, You're telling me randomness is the most efficient way to get a model of a random system?"
(note the sarcasm, even if it lacks wit).
It's funny that he keeps saying humans can recognize when a correlation doesn't pass the sniff test, tho, since this book is about how, in practice, no, we cannot.
Also reminds me of a smart-ass thing I like to say: So, You're telling me randomness is the most efficient way to get a model of a random system?"
(note the sarcasm, even if it lacks wit).
July 15, 2020
Really good book, although the author does become a bit repetitive in the second half. Author makes an excellent point in explaining that patterns in data are not necessarily meaningful, an observation often ignored by many AI and Data Mining enthousiasts.
I really like this book and also recommend Rebooting AI by Gary Marcus and Ernest Davis; I do *not* recommend Artificial Unintelligence by Meredith Broussard
I really like this book and also recommend Rebooting AI by Gary Marcus and Ernest Davis; I do *not* recommend Artificial Unintelligence by Meredith Broussard
August 16, 2019
Remember all those books hyping machine learning and neural networks? This is the opposite but equally vacuous. I lost count of how many times the author uses random data to prove algorithms can be fooled. We get it. You don't need a book to make a simple argument like that. The rest is just the author being grumpy and glorifying human intelligence.
October 31, 2019
Många och låååånga exempel på varför en generell AI inte fungerar och inte kommer att fungera. Så många och långa exempel att jag faktiskt hoppade någon sida här och där.
Smith gör en stor sak av att det går att hitta mönster som inte finns men tänk om vi istället skulle ta och leta efter mönster som faktiskt finns.
Nä, jag tror iallafall att AGI är något som är oundvikligt för framtiden.
Smith gör en stor sak av att det går att hitta mönster som inte finns men tänk om vi istället skulle ta och leta efter mönster som faktiskt finns.
Nä, jag tror iallafall att AGI är något som är oundvikligt för framtiden.
February 3, 2020
Well written reminder that AI should really be called CU - Computer Usefulness - and that Big Data is only useful when someone with some sense is interpreting it - correlations of data can be random, and humans are very good at attaching a story to them, whether true or false.
September 28, 2022
Great warnings about algorithms and our blind trust in them from finance to politics. Easy to read and easy to understand, without any technical jargon or assumed knowledge of computers or programming.
January 5, 2024
Far more depth than just focusing on computers and AI. The book had great examples and stories explaining the shortfall in big data and the pitfalls we see with utilizing statistics to make decisions.
January 24, 2025
Great book
Great book!
Belief in AI reminds me of athlete worship - just because someone is a great athlete doesn't mean that person is a great person.
Just because a computer is involved doesn't mean you get a great answer.
Great book!
Belief in AI reminds me of athlete worship - just because someone is a great athlete doesn't mean that person is a great person.
Just because a computer is involved doesn't mean you get a great answer.
March 1, 2019
Great book about some of the pitfalls of AI and technology and the problems that come along with over-reliance on technology. Insightful book.
March 29, 2019
Expertly crafted, well-explained and a thoroughly fascinating account of AI contrary to the mainstream outlook.
December 16, 2019
A great overview of the current opportunities and limitations of the current level of Ai.
Read
July 29, 2020It's so complicated to rate non-fiction books honestly...
January 4, 2021
Too much bias..
Displaying 1 - 30 of 35 reviews