What do you think?
Rate this book
Prediction Machines: The Simple Economics of Artificial Intelligence
"What does AI mean for your business? Read this book to find out." -- Hal Varian, Chief Economist, Google Artificial intelligence does the seemingly impossible, magically bringing machines to life--driving cars, trading stocks, and teaching children. But facing the sea change that AI will bring can be paralyzing. How should companies set strategies, governments design policies, and people plan their lives for a world so different from what we know? In the face of such uncertainty, many analysts either cower in fear or predict an impossibly sunny future. But in Prediction Machines , three eminent economists recast the rise of AI as a drop in the cost of prediction. With this single, masterful stroke, they lift the curtain on the AI-is-magic hype and show how basic tools from economics provide clarity about the AI revolution and a basis for action by CEOs, managers, policy makers, investors, and entrepreneurs. When AI is framed as cheap prediction, its extraordinary potential becomes clear: Penetrating, fun, and always insightful and practical, Prediction Machines follows its inescapable logic to explain how to navigate the changes on the horizon. The impact of AI will be profound, but the economic framework for understanding it is surprisingly simple.
250 pages, Kindle Edition
First published April 17, 2018
About the author
Ajay Agrawal
5 books357 followersAjay Agrawal is a professor at the University of Toronto’s Rotman School of Management as the Geoffrey Taber Chair in Entrepreneurship and Innovation as well as the Professor of Strategic Management. Agrawal co-founded NEXT Canada, previously The Next 36 in 2010.
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 30 of 281 reviews
February 5, 2019
This is a pretty good book about how artificial intelligence (AI) can be applied to businesses. It is not a technical book--you won't find any details about the wide range of technologies being used for machine learning. Instead, you will find many ingenious ways to put AI to use, as well as all the business ramifications. Three professionals from Toronto's Rotman School of management collaborated to write this book. The book is unified, and reads as if it were written by a single person. However, it is not a particularly engaging book. There is no entertainment value here, definitely no humor. It is a no-nonsense book--almost in the style of a textbook, with good summaries at the end of each chapter. But, the book is not dry, and is easy to read. It is filled with interesting stories and anecdotes.
The basic premise of the book is that the cost of prediction is dropping. Prediction is at the heart of decision-making, so decisions should, overall improve. And, as decisions improve, so should productivity.
The pitfalls of prediction machines are also described. I just love the story about a chess-playing machine during the early days of AI. The machine was fed games from the great grandmasters of chess. The machine successfully analyzed static board positions and suggested good moves. Then, when the machine was programmed to play complete games, something strange happened. Early in its games, it often would sacrifice its queen with no apparent benefit. It turns out the grandmasters occasionally would sacrifice a queen when a masterful quick checkmate could follow. But, the machine could not see that sacrificing a queen without comparable reward was not a good move.
The basic premise of the book is that the cost of prediction is dropping. Prediction is at the heart of decision-making, so decisions should, overall improve. And, as decisions improve, so should productivity.
The pitfalls of prediction machines are also described. I just love the story about a chess-playing machine during the early days of AI. The machine was fed games from the great grandmasters of chess. The machine successfully analyzed static board positions and suggested good moves. Then, when the machine was programmed to play complete games, something strange happened. Early in its games, it often would sacrifice its queen with no apparent benefit. It turns out the grandmasters occasionally would sacrifice a queen when a masterful quick checkmate could follow. But, the machine could not see that sacrificing a queen without comparable reward was not a good move.
September 9, 2018
This book is one of those “assuming a perfectly spherical cow” things.
If we reduce AI down to ML and ignore the messy realities of the real world (i.e. assume that the curse of dimensionality isn’t a thing and the only limitation on creating perfect predictions is access to sufficient training data), then we get the analysis in this book. It paints a picture of a world where the only barrier to self-driving cars is a sufficiently rich training set so that the AI can predict what a driver would do, This is not a world where pesky problems of feature engineering, embodyment, real world resource constraints, or even ethical concerns will make development of many AI solution challenging, if not impossible.
On the plus side the book acknowledges the difference between ‘decisions’ and ‘judgement’, though it defines judgement very narrowly, and it realises that jobs will be redefined rather than eliminated in many cases. But then it fails to realise that in a world where prediction is cheap, then managing the unpredictable is where all the value is (i.e. business exceptions), and the more we improve prediction the more valuable the unpredictable becomes, it doesn’t even acknowledge the many things that might limit prediction, and completely ignores the importance of learning by doing at the operational coal face.
An overly simplistic and narrowly focused book that will soon be dated.
If we reduce AI down to ML and ignore the messy realities of the real world (i.e. assume that the curse of dimensionality isn’t a thing and the only limitation on creating perfect predictions is access to sufficient training data), then we get the analysis in this book. It paints a picture of a world where the only barrier to self-driving cars is a sufficiently rich training set so that the AI can predict what a driver would do, This is not a world where pesky problems of feature engineering, embodyment, real world resource constraints, or even ethical concerns will make development of many AI solution challenging, if not impossible.
On the plus side the book acknowledges the difference between ‘decisions’ and ‘judgement’, though it defines judgement very narrowly, and it realises that jobs will be redefined rather than eliminated in many cases. But then it fails to realise that in a world where prediction is cheap, then managing the unpredictable is where all the value is (i.e. business exceptions), and the more we improve prediction the more valuable the unpredictable becomes, it doesn’t even acknowledge the many things that might limit prediction, and completely ignores the importance of learning by doing at the operational coal face.
An overly simplistic and narrowly focused book that will soon be dated.
May 7, 2018
More of a 3.5 stars. If you aren't familiar with AI, this is probably a very good introduction, although the examples will date very quickly and some of them are plain incorrect (e.g. face tags now sync across Mac and iOS). The point about prediction being a central part of AI is well-made and important, but like most popular economics books, they take this new insight rather too far and with too much confidence.
February 20, 2020
Mais um livro sobre Inteligência Artificial. Afinal, por que não, né?
Este aqui é bem mais voltado para os mercados e setores da economia que vão ser afetados por computação e como vão ser afetados. É uma discussão bem sóbria sobre o que é substituível ou não, sempre batendo no ponto da diferença entre predição e ação, vai chover/levar um guarda-chuva, para dizer que humanos podem perder as predições para as máquinas, mas ainda vamos ser quem dá as ações.
Começam com a discussão do aumento de computação, como isso vai ficar mais pervasivo conforme fica mais barato e acessível e como AI vai conquistando espaço ou substituindo alternativas anteriores em cada setor. Em seguida, passam por cada parte que vai ser alterada, predições, decisões, ferramentas, estratégias e sociedade. Na ordem de como a informação vai sendo usada, de melhores predições a melhores decisões com consequências em cada setor.
Cai na categoria de livros "positivos" sobre o futuro dos trabalhos, aqueles que acham que com mais informação as pessoas continuam empregadas e tomam mais decisões ou decisões mais rapidamente. Apesar de proporem um livro voltado a negócios, achei interessante como uma discussão sobre AI em geral, e não sei dizer se o aconselhamento que dão para negócios é bom.
Este aqui é bem mais voltado para os mercados e setores da economia que vão ser afetados por computação e como vão ser afetados. É uma discussão bem sóbria sobre o que é substituível ou não, sempre batendo no ponto da diferença entre predição e ação, vai chover/levar um guarda-chuva, para dizer que humanos podem perder as predições para as máquinas, mas ainda vamos ser quem dá as ações.
Começam com a discussão do aumento de computação, como isso vai ficar mais pervasivo conforme fica mais barato e acessível e como AI vai conquistando espaço ou substituindo alternativas anteriores em cada setor. Em seguida, passam por cada parte que vai ser alterada, predições, decisões, ferramentas, estratégias e sociedade. Na ordem de como a informação vai sendo usada, de melhores predições a melhores decisões com consequências em cada setor.
Cai na categoria de livros "positivos" sobre o futuro dos trabalhos, aqueles que acham que com mais informação as pessoas continuam empregadas e tomam mais decisões ou decisões mais rapidamente. Apesar de proporem um livro voltado a negócios, achei interessante como uma discussão sobre AI em geral, e não sei dizer se o aconselhamento que dão para negócios é bom.
November 4, 2018
I've read a lot of book on artificial intelligence--this is by far the best one. The others were good, but mostly focused on defining what it is from a technical standpoint. And they all shared ideas on how to use it. But not until I read this one did I realize that all the others are focused on specific situations. This book is the first I've read that focuses on the principles behind the specifics.
For example, they frame AI as a prediction tool, and then discuss the ways that our life and work with change when prediction is ubiquitous and cheap. This leads to a lot of interesting points, including noticing that prediction is only a part of decision making (not the whole decision). And then you can look at jobs and see what parts of the job are prediction parts and what are not (e.g. truck drivers do more than predict how the other cars on the road will react, they also negotiate with people on both ends of the delivery).
Absolutely wonderful future/strategy book. If that's your jam, you'll love this one.
For example, they frame AI as a prediction tool, and then discuss the ways that our life and work with change when prediction is ubiquitous and cheap. This leads to a lot of interesting points, including noticing that prediction is only a part of decision making (not the whole decision). And then you can look at jobs and see what parts of the job are prediction parts and what are not (e.g. truck drivers do more than predict how the other cars on the road will react, they also negotiate with people on both ends of the delivery).
Absolutely wonderful future/strategy book. If that's your jam, you'll love this one.
October 13, 2018
The book was a good high level introduction to AI and its implications, especially from a business perspective.
It was very repetitive at times but this allows the core message to come across clearly: AI helps with making more accurate and faster predictions. Hence, when you find that the cost of prediction falls, industries are either disrupted or new industries proliferate.
Overall, three stars (more likely makes sense as a three-and-a-half) for being informative but not providing a significant number of fresh insights throughout the book.
It was very repetitive at times but this allows the core message to come across clearly: AI helps with making more accurate and faster predictions. Hence, when you find that the cost of prediction falls, industries are either disrupted or new industries proliferate.
Overall, three stars (more likely makes sense as a three-and-a-half) for being informative but not providing a significant number of fresh insights throughout the book.
June 24, 2018
This review originally appeared on my website: https://daveveitch.wordpress.com/2018...
Recently I read Prediction Machines by Ajay Agrawal, Joshua Gans, and Avi Goldfarb, three professors from UofT’s Rotman School of Management. I recommend this book to anyone who wants to go ‘beyond the headlines’ with respect to what impact on the world machine learning & AI will have.
Topic
“Where others see transformational new innovation, we see a simple fall in price.”
The key idea the book revolves around is that machine learning & AI have brought about a dramatic fall in the price of prediction. The authors argue this fall in price will lead to the emergence of new business models (similar to how new business models emerged as Google search became popular), and it will also increase the value of other things (e.g. sensors which accurately capture data will become more valuable).
Review
This may be the best book yet I have read in the ‘machine learning/AI/robots will transform our world’ genre. Part of the reason why is that Prediction Machines does not try to do too much. The authors focus on a few key ideas (as mentioned above), and then carefully evaluate the ramifications of them. The authors’ writing style is clear (the chapter-by-chapter bullet points help with this), and back their arguments up with numerous real-world examples. The book is similar in some ways to Martin Ford’s Rise of the Robots, and the overall style is reminiscent of Robert Shiller’s Animal Spirits.
Two things I was not wild about. I found the section ‘Part 3: Tools’ a bit dry, and wish the authors had gone into a little bit more detail about how machine learning actually works (potentially via an appendix).
Best Bits
A few sections/arguments/examples I found interesting:
- Rating agencies’ models before the financial crisis did not sufficiently incorporate how housing prices are correlated across regions. “Machine learning enables predictions based on unanticipated correlations”, and this feature could have been helpful at the time. (pg. 37)
- “The value of substitutes to prediction machines, namely human prediction, will decline. However, the value of complements, such as the human skills associated with data collection, judgment, and actions, will become more valuable.” (pg. 81)
- “The recent developments in AI and machine learning have convinced us that this innovation is on par with the great, transformative technologies of the past: electricity, cars, plastics, the microchip, the internet, and the smartphone”. (pg. 155)
- The authors cite an interesting example of how in the 1930s a new strain of higher yielding corn took a very long time to become widely used in some states (e.g. Texas, Alabama). Part of the reason for this was that farms in these states were smaller & less profitable, making experimentation on new corn varieties hard to justify. The authors argue that the large profit margins of firms like Google/Facebook are enabling them to experiment broadly with AI techniques, and “reap huge rewards from successful experiments by applying them across a wide range of products operating at large scale”. (pg. 160)
Recently I read Prediction Machines by Ajay Agrawal, Joshua Gans, and Avi Goldfarb, three professors from UofT’s Rotman School of Management. I recommend this book to anyone who wants to go ‘beyond the headlines’ with respect to what impact on the world machine learning & AI will have.
Topic
“Where others see transformational new innovation, we see a simple fall in price.”
The key idea the book revolves around is that machine learning & AI have brought about a dramatic fall in the price of prediction. The authors argue this fall in price will lead to the emergence of new business models (similar to how new business models emerged as Google search became popular), and it will also increase the value of other things (e.g. sensors which accurately capture data will become more valuable).
Review
This may be the best book yet I have read in the ‘machine learning/AI/robots will transform our world’ genre. Part of the reason why is that Prediction Machines does not try to do too much. The authors focus on a few key ideas (as mentioned above), and then carefully evaluate the ramifications of them. The authors’ writing style is clear (the chapter-by-chapter bullet points help with this), and back their arguments up with numerous real-world examples. The book is similar in some ways to Martin Ford’s Rise of the Robots, and the overall style is reminiscent of Robert Shiller’s Animal Spirits.
Two things I was not wild about. I found the section ‘Part 3: Tools’ a bit dry, and wish the authors had gone into a little bit more detail about how machine learning actually works (potentially via an appendix).
Best Bits
A few sections/arguments/examples I found interesting:
- Rating agencies’ models before the financial crisis did not sufficiently incorporate how housing prices are correlated across regions. “Machine learning enables predictions based on unanticipated correlations”, and this feature could have been helpful at the time. (pg. 37)
- “The value of substitutes to prediction machines, namely human prediction, will decline. However, the value of complements, such as the human skills associated with data collection, judgment, and actions, will become more valuable.” (pg. 81)
- “The recent developments in AI and machine learning have convinced us that this innovation is on par with the great, transformative technologies of the past: electricity, cars, plastics, the microchip, the internet, and the smartphone”. (pg. 155)
- The authors cite an interesting example of how in the 1930s a new strain of higher yielding corn took a very long time to become widely used in some states (e.g. Texas, Alabama). Part of the reason for this was that farms in these states were smaller & less profitable, making experimentation on new corn varieties hard to justify. The authors argue that the large profit margins of firms like Google/Facebook are enabling them to experiment broadly with AI techniques, and “reap huge rewards from successful experiments by applying them across a wide range of products operating at large scale”. (pg. 160)
May 5, 2019
I was recently asked to organize my company’s strategic offsite. I wanted a keynote speaker who could discuss machine learning and its impact on the investment industry. I came across the name, Ajay Agrawal, the founder of Creative Destruction Labs - a remarkable organization at the forefront of science, data, and technology based in Toronto - and the author Prediction Machines. Agrawal’s premise is fascinating: as predictions get cheaper to make it will change how we use them - similar to how cheaper electric lighting changed our working environment in the 1800s, or how cheaper computing power changed the way we used mathematics in the late twentieth century. Apparently predictions will become so cost effective and accurate that Amazon’s algorithms will eventually get so good at figuring out what we want and when we want it, that the company will change its business model from sending us goods after we order and pay for them to before - an idea I find both intriguing and terrifying. While some of the book’s prophesies may be disconcerting, this is hardly an anxious or dispiriting tome. Agrawal offers a more hopeful note by identifying roles for humans to play when machines take over the world, generally to apply judgement and common sense. There are some drier, more policy based sections, but for the most part this is a compelling, very readable book. If at times it was a little slow-going, it was not because it was overly heavy or badly written; but because there was so much thought-provoking material that I paused often - every other sentence at times - to daydream, ponder and contemplate.
March 8, 2019
It is difficult to have a timely and relevant book on such a broad topic as artificial intelligence; some parts will be familiar while other parts are new and fascinating. This book would probably interest knowledge workers or anyone wanting to know how AI is changing our environment and behaviors. For example, we used to review our credit card statements to see if there were any fraudulent charges. Now based on our purchasing habits, credit card companies are flagging possible fraudulent charges and declining those transactions without needing input from us.
Some AI stories are familiar, like the driverless car. Some are new and inconceivable, like Amazon's patent for 'ship and shop.' Currently, we decide what we want to buy, click on those items on Amazon, and they're shipped to us ('shop and ship'). In the future, Amazon expects to know its customers so well that it can ship the items to us (before we've thought about buying them) and we return only those items we don't want.
Some AI stories are familiar, like the driverless car. Some are new and inconceivable, like Amazon's patent for 'ship and shop.' Currently, we decide what we want to buy, click on those items on Amazon, and they're shipped to us ('shop and ship'). In the future, Amazon expects to know its customers so well that it can ship the items to us (before we've thought about buying them) and we return only those items we don't want.
December 4, 2019
อ่านง่ายดีครับ หนังสือเล่าการทำงานง่ายๆ ของ AI และบทบาทของมันในระบบเศรษฐกิจและภาคธุรกิจ
หลักๆ คือเสนอว่าสิ่งที่ AI ทำได้ดีจริงๆ คือการทำนายซึ่งต้องอาศัยข้อมูลปริมาณมากๆ สิ่งที่จำเป็นต่อการทำนาย เช่น ข้อมูลจึงจะมีมูลค่าสูงขึ้นเสมอ แต่ถึงอย่างนั้น แนวโน้มคือกระบวนการตัดสินใจในเรื่องต่่างๆ ยังมีมากกว่าการประมวลข้อมูล คำนวณ และทำนาย พื้นที่เหล่านี้เองที่ทำให้ทักษะบางอย่างของมนุษย์ยังคงจำเป็นอยู่ และทำให้ทักษะเหล่านี้มีมูลค่าเช่นกัน
ผมว่าถ้าใครตามเรื่องพวกนี้อยู่แล้วก็คงเฉยๆ มากๆ เพราะไม่ค่อยมีอะไรใหม่ทั้งที่หนังสือเพิ่งออกมาปีที่แล้ว แต่ถ้าใครไม่เคยตามประเด็น เล่มนี้ก็น่าจะเป็นจุดเริ่มต้นที่ใช้ได้ทีเดียว
หลักๆ คือเสนอว่าสิ่งที่ AI ทำได้ดีจริงๆ คือการทำนายซึ่งต้องอาศัยข้อมูลปริมาณมากๆ สิ่งที่จำเป็นต่อการทำนาย เช่น ข้อมูลจึงจะมีมูลค่าสูงขึ้นเสมอ แต่ถึงอย่างนั้น แนวโน้มคือกระบวนการตัดสินใจในเรื่องต่่างๆ ยังมีมากกว่าการประมวลข้อมูล คำนวณ และทำนาย พื้นที่เหล่านี้เองที่ทำให้ทักษะบางอย่างของมนุษย์ยังคงจำเป็นอยู่ และทำให้ทักษะเหล่านี้มีมูลค่าเช่นกัน
ผมว่าถ้าใครตามเรื่องพวกนี้อยู่แล้วก็คงเฉยๆ มากๆ เพราะไม่ค่อยมีอะไรใหม่ทั้งที่หนังสือเพิ่งออกมาปีที่แล้ว แต่ถ้าใครไม่เคยตามประเด็น เล่มนี้ก็น่าจะเป็นจุดเริ่มต้นที่ใช้ได้ทีเดียว
May 19, 2018
The authors provide a good framework for understanding the changes that Artificial Intelligence effect in businesses and the economy. A book that avoids getting excessively technical (either about economics or computer science) while still avoiding the hype and shallowness of most discussions on the theme. A greater attention to the impacts of AI on outside stakeholders could have improved the overall result, but the book still achieves its framing goals.
January 9, 2020
Topic:
A practical examination of understanding innovations In Artificial Intelligence (specifically Machine Learning) and how commercial applications of AI have already impacted (and will continue to affect) business models, the role of government, and the economic and social relationship between machines and human.
Style:
The three authors are professors of economics at U of Toronto Rotman School of Management who are also founders of an innovation incubator with investments in AI and deep learning. The book is written in less technical language and clearly aimed at consultants and c-suites on how to unpack the full impact of AI for work flow and business models. If you’re interested in deeper dive on the technical side of machine learning, this is not the book for you. For all practical purposes, this book assumes the relevance of AI/machine learning is increased accuracy and cost-efficiency for making predictions.
Organization:
The book is divided into five sections addressing (1) the nature of Prediction and how creating a taxonomy to guide the rest of the discussion (known-knowns, known-unknowns, unknown-unknowns are fairly straightforward; unknown-knowns were new to me and are particularly relevant to the discussion of AI), (2) Decision Making, particularly how it is divided and the role of judgment, (3) Tools and how AI will impact work flows, decision-making, and jobs, (4) Strategy, giving sage advice on preparing one's company for the impact of AI and how to mitigate against certain risks, and (5) Society, a larger discussion on how AI will affect the environment in which people and businesses will operate.
Takeaway:
An accessible introduction to artificial intelligence and machine learning (AI/ML), particularly for unpacking the practical effects AI/ML may (or may not) have on commercial ventures. I initially thought the book's lack of any technical aspects would limit insights to dinner party small talk, but was pleasantly surprised that avoiding scientific lectures (that would likely be outdated by publication) in favor of conversational thought-experiments and economic insights forced me to think critically about why any of this matters.
A recurring example the authors use is improving the accuracy of predicting shopping habits can radically alter a business like Amazon from a "shop-then-ship" to a "ship-then-shop" model; that is, if the likelihood of a customer returning a recommended item falls below a certain threshold, then Amazon's costs for returned items (both for processing returns and customer dissatisfaction) will be outweighed by a greater likelihood the customer is satisfied by this automated convenience and will keep (and pay for) an item they may not yet have shopped for or known about. The authors go back to this hypothetical in exploring various ideas, including:
-how businesses will feel the impact of AI/ML (the short answer is reduction in cost for certain predictions, which will increase the use and demand for predictions and create downstream opportunities; good historical comparisons are the internet reducing the cost of searching or electrical lighting reducing the cost of reading);
-how AI/ML improvements will impact work flows and the division of labor (computers are now better at identifying certain images, e.g., x-rays, than humans are, so the relationship of computers to a human component, e.g., radiologist, may frontload reliance on computers for an initial analysis and shift non-codifiable judgment/decision making to humans to optimize time and error reduction);
-how AI/ML improvements will impact allocation of a company's resources (better predictions could mean less investment in capital equipment since you can predict when to contract out specific capital-intensive tasks; certain risk management solutions like biopsies and airport lounges may be utilized more sparingly with better predictions using non-invasive testing or airport commuting);
-how AI/ML improvements will impact job training (if AI/ML is meant to accurately predict how humans currently behave, e.g., how doctors identify anomalies in medical imagining, at some point machines may become so much better than humans, the programming will become a victim of its own success if there are not enough model humans to retrain the AI if the old data is no longer reliable);
-grappling with when to embrace AI/ML innovation (innovator's dilemma is choosing whether to adopt early to learn from customers and gather data but risk poor performance in the short term OR keep existing customers happy and adopt a more proven model later but risk losing market position to more early-adopter competition; another benefit of early adoption is helping set standards for a nascent technology); and
-how risk tolerance for prediction accuracy across different industries influences customer adoption and incentives even incremental AI/ML improvements (think Google Mail proposed responses vs. computer-navigated driverless cars; increasing predication accuracy from 95 to 96% on when to make a right turn may be the deciding factor of when driverless cars are commercially viable and customers feel safe).
As an attorney who works with many life sciences and technology companies, the importance of navigating data privacy laws (which have only starting gaining attention over the last five years) cannot be emphasized enough as it impacts every industry (finance, tech, life sciences, education, healthcare, etc.). In an era where data can be collected in innumerable and imperceptible ways and the best data gives a huge competitive advantage , citizens in the US and EU have pushed back to assert greater control over such personal information and tech companies like Apple have listened and dedicated considered PR to allay such customer fears. Nevertheless, the authors astutely recognize the inherent innovative advantage that China and other countries without a data privacy legal regime may have in innovating without legal hurdles.
The authors correctly observe the political implications that AI/ML will have on the shift in balance of national income derived from labor vs. capital owners, which we’re already seeing in compensatory policy proposals for “universal basic income” by presidential candidates. But, far from being doomsayers, the authors use historical examples of how innovation has created new opportunities that we could never have predicted. One maxim the author's proffer is how AI/ML actually increases the value of human judgment and decision-making . Certain cognitive tasks, the authors assert, simply cannot be programmed or codified given the inherent limitation of a computer's predictive capacity (think of a "black swan" event). This assumes that computers will never be able to accurately mimick human reasoning. Given how much ground AI/ML has covered over the last five years alone that was previously thought impossible and the ability of computers to crunch quadrillions of quantified past human decisions for any imaginable scenario, I wouldn't be surprised if a sequel book is needed to reassess how inimitable or superior human ingenuity is compared to computers.
A practical examination of understanding innovations In Artificial Intelligence (specifically Machine Learning) and how commercial applications of AI have already impacted (and will continue to affect) business models, the role of government, and the economic and social relationship between machines and human.
Style:
The three authors are professors of economics at U of Toronto Rotman School of Management who are also founders of an innovation incubator with investments in AI and deep learning. The book is written in less technical language and clearly aimed at consultants and c-suites on how to unpack the full impact of AI for work flow and business models. If you’re interested in deeper dive on the technical side of machine learning, this is not the book for you. For all practical purposes, this book assumes the relevance of AI/machine learning is increased accuracy and cost-efficiency for making predictions.
Organization:
The book is divided into five sections addressing (1) the nature of Prediction and how creating a taxonomy to guide the rest of the discussion (known-knowns, known-unknowns, unknown-unknowns are fairly straightforward; unknown-knowns were new to me and are particularly relevant to the discussion of AI), (2) Decision Making, particularly how it is divided and the role of judgment, (3) Tools and how AI will impact work flows, decision-making, and jobs, (4) Strategy, giving sage advice on preparing one's company for the impact of AI and how to mitigate against certain risks, and (5) Society, a larger discussion on how AI will affect the environment in which people and businesses will operate.
Takeaway:
An accessible introduction to artificial intelligence and machine learning (AI/ML), particularly for unpacking the practical effects AI/ML may (or may not) have on commercial ventures. I initially thought the book's lack of any technical aspects would limit insights to dinner party small talk, but was pleasantly surprised that avoiding scientific lectures (that would likely be outdated by publication) in favor of conversational thought-experiments and economic insights forced me to think critically about why any of this matters.
A recurring example the authors use is improving the accuracy of predicting shopping habits can radically alter a business like Amazon from a "shop-then-ship" to a "ship-then-shop" model; that is, if the likelihood of a customer returning a recommended item falls below a certain threshold, then Amazon's costs for returned items (both for processing returns and customer dissatisfaction) will be outweighed by a greater likelihood the customer is satisfied by this automated convenience and will keep (and pay for) an item they may not yet have shopped for or known about. The authors go back to this hypothetical in exploring various ideas, including:
-how businesses will feel the impact of AI/ML (the short answer is reduction in cost for certain predictions, which will increase the use and demand for predictions and create downstream opportunities; good historical comparisons are the internet reducing the cost of searching or electrical lighting reducing the cost of reading);
-how AI/ML improvements will impact work flows and the division of labor (computers are now better at identifying certain images, e.g., x-rays, than humans are, so the relationship of computers to a human component, e.g., radiologist, may frontload reliance on computers for an initial analysis and shift non-codifiable judgment/decision making to humans to optimize time and error reduction);
-how AI/ML improvements will impact allocation of a company's resources (better predictions could mean less investment in capital equipment since you can predict when to contract out specific capital-intensive tasks; certain risk management solutions like biopsies and airport lounges may be utilized more sparingly with better predictions using non-invasive testing or airport commuting);
-how AI/ML improvements will impact job training (if AI/ML is meant to accurately predict how humans currently behave, e.g., how doctors identify anomalies in medical imagining, at some point machines may become so much better than humans, the programming will become a victim of its own success if there are not enough model humans to retrain the AI if the old data is no longer reliable);
-grappling with when to embrace AI/ML innovation (innovator's dilemma is choosing whether to adopt early to learn from customers and gather data but risk poor performance in the short term OR keep existing customers happy and adopt a more proven model later but risk losing market position to more early-adopter competition; another benefit of early adoption is helping set standards for a nascent technology); and
-how risk tolerance for prediction accuracy across different industries influences customer adoption and incentives even incremental AI/ML improvements (think Google Mail proposed responses vs. computer-navigated driverless cars; increasing predication accuracy from 95 to 96% on when to make a right turn may be the deciding factor of when driverless cars are commercially viable and customers feel safe).
As an attorney who works with many life sciences and technology companies, the importance of navigating data privacy laws (which have only starting gaining attention over the last five years) cannot be emphasized enough as it impacts every industry (finance, tech, life sciences, education, healthcare, etc.). In an era where data can be collected in innumerable and imperceptible ways and the best data gives a huge competitive advantage , citizens in the US and EU have pushed back to assert greater control over such personal information and tech companies like Apple have listened and dedicated considered PR to allay such customer fears. Nevertheless, the authors astutely recognize the inherent innovative advantage that China and other countries without a data privacy legal regime may have in innovating without legal hurdles.
The authors correctly observe the political implications that AI/ML will have on the shift in balance of national income derived from labor vs. capital owners, which we’re already seeing in compensatory policy proposals for “universal basic income” by presidential candidates. But, far from being doomsayers, the authors use historical examples of how innovation has created new opportunities that we could never have predicted. One maxim the author's proffer is how AI/ML actually increases the value of human judgment and decision-making . Certain cognitive tasks, the authors assert, simply cannot be programmed or codified given the inherent limitation of a computer's predictive capacity (think of a "black swan" event). This assumes that computers will never be able to accurately mimick human reasoning. Given how much ground AI/ML has covered over the last five years alone that was previously thought impossible and the ability of computers to crunch quadrillions of quantified past human decisions for any imaginable scenario, I wouldn't be surprised if a sequel book is needed to reassess how inimitable or superior human ingenuity is compared to computers.
January 18, 2023
ฉันผู้อยู่ในวงการนี้ ถึงจะไม่ได้ทำกับ AI โดยตรงแต่ก็อยากมาลองฟังบ้าง เห็นคนพูดถึงเล่มนี้มาพักใหญ่แล้ว
เป็นหนังสืออธิบายหลักการ AI ในฉบับภาษาชาวบ้านเข้าใจง่ายสุดๆ พร้อมกับตัวอย่างในอุตสาหกรรมที่เกิดขึ้นจริง ไม่ได้มาในรูปแบบหนังสือเรียนอธิบาย algorithm ชื่อแฟนซีพร้อมสูตรเลขยากๆ แบบนั้น
บอกเล่าว่า AI เติบโตขึ้นได้เพราะมนุษย์ทำการป้อนข้อมูลลงไป (แต่ก็มีประเภทที่ AI สามารถเรียนรู้ด้วยตัวมันเองได้เช่นกัน)
ยกตัวอย่างกรณีต่างๆ ที่เกิดเคยขึ้นเช่น bot บน Twitter อย่างเทย์ผู้ใจแตกที่โดนสั่งปิดไปแล้ว, Alpha Go, เครื่องเสียง Alexa, ระบบ autopilot ของเครื่องยนต์ต่างๆ
ที่น่าทึ่งคือหนังสือเล่มนี้แปลศัพท์เฉพาะเป็นภาษาไทยทั้งหมด เช่น
AI - Artificial Intelligence=ปัญญาประดิษฐ์ อันนี้พบเห็นบ่อย ไม่ค่อยแปลกใจ
Maching Learning= การเรียนรู้ของเครื่องจักร ตามพจนานุกรมแปลแบบนี้จริง แต่เวลาทำงานจริงเค้าก็พูดทับศัพท์ แมชชีนเลิร์นนิ่งหรือ เอ็มแอลไป
การเรียนรู้แบบสะท้อนกลับ(เดาว่าในที่นี่เขาหมายถึง Reinforcement Learning มากกว่า Backpropagation)
จริงๆ คำศัพท์แปลสวย แต่เช่นเดิมว่าชีวิตจริงเค้าก็พูดทับศัพท์กัน
ทั้งนี้ทั้งนั้นใครที่ยังใหม่กับวงการนี้ คงเปิดโลกและเห็นภาพพอสมควรว่า AI เข้ามามีบทบาทได้อย่างไร ซึ่งแน่นอน ณ ตอนนี้มันทำได้แต่งานเฉพาะทาง ไม่ถึงกับจะครองโลกเหมือนในหนังในการ์ตูนได้ขนาดนั้น แต่มันเริ่มขยายสโคปงานมาในจุดที่ล่อแหลมขัดต่อศีลธรรมบ้างเช่น generative AI ที่สร้างงานอาร์ตในเพียงเสี้ยววินาทีจากที่เดิมเป็นทักษะเฉพาะของศิลปิน แต่สำหรับใครที่ศึกษาเรื่องนี้อยู่แล้วหรืออยู่ในคร่ำหวอดในแวดวงการนี้ อาจไม่ได้รู้สึกใหม่อะไรมาก แต่มันก็มีหัวข้อหรือ Business Case ต่างๆ ที่หากเอาไป Google เพิ่ม มันก็อาจไปใช้เล่าใช้พูดตามงานอีเวนท์ต่างๆ ได้นะ
ป.ล. ฟังหนังสือเสียงเอา ด้วยความที่เป็นผลงานของ Hana Studio ที่ชอบใส่เสียง BGM มา
ฉันเมาเสียงปี่ในนี้มาก แอ่ อี้ แอ่ อี๊ แอ่ ทุก ๆ สองนาที 55555
เป็นหนังสืออธิบายหลักการ AI ในฉบับภาษาชาวบ้านเข้าใจง่ายสุดๆ พร้อมกับตัวอย่างในอุตสาหกรรมที่เกิดขึ้นจริง ไม่ได้มาในรูปแบบหนังสือเรียนอธิบาย algorithm ชื่อแฟนซีพร้อมสูตรเลขยากๆ แบบนั้น
บอกเล่าว่า AI เติบโตขึ้นได้เพราะมนุษย์ทำการป้อนข้อมูลลงไป (แต่ก็มีประเภทที่ AI สามารถเรียนรู้ด้วยตัวมันเองได้เช่นกัน)
ยกตัวอย่างกรณีต่างๆ ที่เกิดเคยขึ้นเช่น bot บน Twitter อย่างเทย์ผู้ใจแตกที่โดนสั่งปิดไปแล้ว, Alpha Go, เครื่องเสียง Alexa, ระบบ autopilot ของเครื่องยนต์ต่างๆ
ที่น่าทึ่งคือหนังสือเล่มนี้แปลศัพท์เฉพาะเป็นภาษาไทยทั้งหมด เช่น
AI - Artificial Intelligence=ปัญญาประดิษฐ์ อันนี้พบเห็นบ่อย ไม่ค่อยแปลกใจ
Maching Learning= การเรียนรู้ของเครื่องจักร ตามพจนานุกรมแปลแบบนี้จริง แต่เวลาทำงานจริงเค้าก็พูดทับศัพท์ แมชชีนเลิร์นนิ่งหรือ เอ็มแอลไป
การเรียนรู้แบบสะท้อนกลับ(เดาว่าในที่นี่เขาหมายถึง Reinforcement Learning มากกว่า Backpropagation)
จริงๆ คำศัพท์แปลสวย แต่เช่นเดิมว่าชีวิตจริงเค้าก็พูดทับศัพท์กัน
ทั้งนี้ทั้งนั้นใครที่ยังใหม่กับวงการนี้ คงเปิดโลกและเห็นภาพพอสมควรว่า AI เข้ามามีบทบาทได้อย่างไร ซึ่งแน่นอน ณ ตอนนี้มันทำได้แต่งานเฉพาะทาง ไม่ถึงกับจะครองโลกเหมือนในหนังในการ์ตูนได้ขนาดนั้น แต่มันเริ่มขยายสโคปงานมาในจุดที่ล่อแหลมขัดต่อศีลธรรมบ้างเช่น generative AI ที่สร้างงานอาร์ตในเพียงเสี้ยววินาทีจากที่เดิมเป็นทักษะเฉพาะของศิลปิน แต่สำหรับใครที่ศึกษาเรื่องนี้อยู่แล้วหรืออยู่ในคร่ำหวอดในแวดวงการนี้ อาจไม่ได้รู้สึกใหม่อะไรมาก แต่มันก็มีหัวข้อหรือ Business Case ต่างๆ ที่หากเอาไป Google เพิ่ม มันก็อาจไปใช้เล่าใช้พูดตามงานอีเวนท์ต่างๆ ได้นะ
ป.ล. ฟังหนังสือเสียงเอา ด้วยความที่เป็นผลงานของ Hana Studio ที่ชอบใส่เสียง BGM มา
ฉันเมาเสียงปี่ในนี้มาก แอ่ อี้ แอ่ อี๊ แอ่ ทุก ๆ สองนาที 55555
August 13, 2020
Very good (5* for narrow audience) if niche book on the economics of AI/machine learning.
By narrow audience and mean for business leaders, decision makers, and people learning more on the subject, could be students or economists. It is explicitly aimed at them not a lay audience. It is not a computer science book, rather an economics book on the the impact of the new technology of cheap prediction - wonky but not overly complex or jargon filled. Well structured, some may just read for relevent sections of pushed for time.
It is very good in avoiding buzz and bluster around these tools, seeing them just as that, though like the computing revolution that made computation cheap, this economic change of scale will push profound changes to our lives.
It is also excellent in detailing what won't change and where these technologies have clear weaknesses. In particular better machine prediction will lead to a premium on human judgment in many spaces. That task is often seperated out with good reason - for now!
There are many real world examples and stories that enliven the text, the writers run a AI based resource for new companies so have great access. Interesting that so many start ups are predicated on a single application, though the authors explain the real value is in powering change through the entire work flow.
Interesting that the big tech companies are "AI first" meaning that they prioritize it over even short term customer satisfaction, the big goal is too important to slow development even if that means some of us switch services.
Good and balanced on the future, with the new world being changed by AI but making a clear distinction with any possible development of GAI, that would be different gravy. Good sections on privacy, China, and the EU.
As with all good books references Taversky and Khaneman :)
By narrow audience and mean for business leaders, decision makers, and people learning more on the subject, could be students or economists. It is explicitly aimed at them not a lay audience. It is not a computer science book, rather an economics book on the the impact of the new technology of cheap prediction - wonky but not overly complex or jargon filled. Well structured, some may just read for relevent sections of pushed for time.
It is very good in avoiding buzz and bluster around these tools, seeing them just as that, though like the computing revolution that made computation cheap, this economic change of scale will push profound changes to our lives.
It is also excellent in detailing what won't change and where these technologies have clear weaknesses. In particular better machine prediction will lead to a premium on human judgment in many spaces. That task is often seperated out with good reason - for now!
There are many real world examples and stories that enliven the text, the writers run a AI based resource for new companies so have great access. Interesting that so many start ups are predicated on a single application, though the authors explain the real value is in powering change through the entire work flow.
Interesting that the big tech companies are "AI first" meaning that they prioritize it over even short term customer satisfaction, the big goal is too important to slow development even if that means some of us switch services.
Good and balanced on the future, with the new world being changed by AI but making a clear distinction with any possible development of GAI, that would be different gravy. Good sections on privacy, China, and the EU.
As with all good books references Taversky and Khaneman :)
August 8, 2022
I picked this in continuation of my last read ‘Competing in the Age of AI: Strategy and Leadership When Algorithms and Networks Run the World’ and I am so glad I completed this. While last one was more on implications of AI in general and how it changes operations and strategy of organizations, this book additionally covers details of an important aspect of decision making: prediction. Central theme of the book is prediction. It covers why prediction is central to AI, how its capability is enhanced and limited by data, how it’s changing the division of labor with job of humans shifting towards judgement and data part of decision making, details on how work flows and decisions can be deconstructed leading to automations and redesigning of jobs, how C-suite is forced to change strategies based on data and prediction available, how organizations need to manage risks, etc.
Authors have tried to cover details of prediction and decision making – how it learns from training, input, and feedback data – but at the same time have kept away from technical details to avoid complexity. Examples and reasoning may seem repetitive at times, but it’s an easy read and useful for anyone who is new or understands AI and ML.
Authors have tried to cover details of prediction and decision making – how it learns from training, input, and feedback data – but at the same time have kept away from technical details to avoid complexity. Examples and reasoning may seem repetitive at times, but it’s an easy read and useful for anyone who is new or understands AI and ML.
July 30, 2020
A good read for anyone with the soul of an economist and an interest in AI. The ideas presented are for the most part basic (perhaps 'foundational' is better) but they are clearly described in economic terms and in a sensible order. The authors begin by looking at the fundamental impact that AI in its current form will have on the production function of various goods/services as well as on the demand for labour, before going on to consider how company leaders might evaluate (and manage) the benefits and risks of using AI. They conclude by briefly touching on potential societal issues related to using AI (including discrimination and existential risks) although there are more comprehensive issues that deal entirely with those.
August 12, 2020
What's magical about AI and machine learning? They let us predict better and cheaply. That's the premise of the books. Authors talk about what we as humans can do alongside with these predictions but I'd say we are far off from. Being able to perfectly predict the future. We can't even predict the GDP in 2 years. Have we heard of the pandemic coming 1 year ago? We still need to work on a lot more on being able to predict better.
Overall, I didn't find much depth in the book and it was similar to blogs we can find. AI is a popular topic nowadays anyway.
Overall, I didn't find much depth in the book and it was similar to blogs we can find. AI is a popular topic nowadays anyway.
February 28, 2021
“Um guia de fácil compreensão sobre a IA d os gigantescos efeitos que ela pode ter em nossa economia, sociedade e sistema político.” Robert Rubin
- único ponto negativo quanto ao livro mesmo é a letra muito pequena, deixa a leitura mais difícil.
Como o Rubin disse, o livro possui uma visão bem ampla sobre os efeitos da inteligência artificial. Além disso, ele possui uma linguagem bem acessível.
- único ponto negativo quanto ao livro mesmo é a letra muito pequena, deixa a leitura mais difícil.
Como o Rubin disse, o livro possui uma visão bem ampla sobre os efeitos da inteligência artificial. Além disso, ele possui uma linguagem bem acessível.
June 6, 2021
It is an easy to understand entry-level book about AI and its impact on the economy and society. A group writes it of economist that is closely working with AI-oriented startups, and it will give you good insight about what to expect and how to build a strategy in upcoming's years for business owners, employer and people who want to be ready for years ahead. But, if you are already reading about AI and ML, it might be. A bit boring, even though it's short and easy to listen to/read.
May 9, 2021
This book was fine. I think reading a couple wikipedia pages and watching a 10 minute video could have been just as informative, but the book is good for the general overview it tries to do I suppose
I think the central topic, of focusing on the way that machine learning is fundamentally about predictions is a useful framework to bring to a wider audience. Voice recognition models don't "understand" the words spoken, they predict the words most likely to be associated with the sounds it receives as inputs
I'm not really sure who I would recommend this too. I guess someone who finds a book easier to focus on than just reading wikipedia and does not have much context for the topic otherwise
I think the central topic, of focusing on the way that machine learning is fundamentally about predictions is a useful framework to bring to a wider audience. Voice recognition models don't "understand" the words spoken, they predict the words most likely to be associated with the sounds it receives as inputs
I'm not really sure who I would recommend this too. I guess someone who finds a book easier to focus on than just reading wikipedia and does not have much context for the topic otherwise
August 27, 2022
While this has some enlightening examples of how machine learning currently is being used or potentially will be used, overall it was a bit dry. Also, as I was listening to this on Audible, some of the explanations were a bit number heavy or included charts, which would have been much easier to follow if reading the book.
March 17, 2022
While a little dated. Given the speed of innovation in the machine learning space. I still thoroughly enjoyed the book, with some great practical tips on improving your AI initiatives, like utilising the AI canvas.
February 2, 2019
AI Trong cuộc cách mạng công nghệ 4.0 là một cuốn sách khoa học về đề tài phát triển doanh nghiệp, tổ chức, cá nhân bằng trí tuệ nhân tạo. Bằng cách đặt người đọc là những người chủ doanh nghiệp, công ty, cuốn sách đã diễn tả nội dung một cách chi tiết và gần gũi hơn bao giờ hết.
Có thể nói đây là một trong những cuốn sách hiếm hoi về AI được xuất bản tại Việt Nam. Trước đó phải kể đến Phát minh cuối cùng của James Barrat đã tạo ra một làn sóng dữ dội trong cộng đồng người thích tìm hiểu về AI với nhiều ý kiến trái chiều. Một thời gian sau, AI Trong cuộc cách mạng 4.0 được xuất bản như một điều hiển nhiên và một lần nữa, sức hút của đề tài này lại càng nóng hơn bao giờ hết.
Trước nhất phải nói đây là một cuốn sách pha trộn giữa học thuyết khoa học về trí tuệ nhân tạo với định hướng phát triển doanh nghiệp. Đây chính là điểm nhấn quan trọng và chói sáng của toàn bộ tác phẩm. Đặt nội dung vào trong một ví dụ thực tế, toàn bộ AI được diễn giải một cách không thể rõ ràng, cụ thể hơn. Khác với Phát minh cuối cùng bàn về những kiến thức lý thuyết rất nặng nề thì cuốn sách này lại chi tiết hơn. 3 tác giả đã cố gắng để khiến những người không có nhiều kiến thức về khoa học cũng có thể hiểu được. Đặc biệt là những nhà doanh nghiệp sẽ phải đặc biệt lưu tâm đến cuốn này bởi tính thực tế của nó và coi nó như là "sách giáo khoa" của mình.
Bố cục sách rất rõ ràng được trình bày theo trình tự dễ hiểu. Sự dự đoán là nội dung chính trong vấn đề mà tác giả đưa ra về AI. Kỳ thực ban đầu tôi không đồng tính với quan điểm của tác giả khi cho rằng AI dùng sự dự đoán để tự động hóa chính nó bao gồm công việc, hành động, phản xạ,... Nhưng càng đọc tôi lại càng bị hút vào, thôi thúc đi tìm câu trả lời cho thắc mắc của bản thân và sự không thỏa mãn ban đầu đó. Tôi đã không thất vọng. Càng đào sâu tôi lại càng bất ngờ trước những luận điểm gồm nhiều luận cứ sắc bén của tác giả. Phải nói rằng tôi muốn phản bác lại tất cả những điều đó nhưng không thể, lập luận của tác giả quá chặt chẽ khiến tất cả những sự phản biện trong đầu tôi lần lượt bị bác bỏ qua mỗi lần lật trang.
Nói như vậy không có nghĩa là cuốn sách này không có những lỗ hổng. Ví dụ như việc AI có thể học được cảm xúc của con người. Tôi hoàn toàn không đồng tình với quan điểm này. Hay như AI có thể thay thế một số nghề nghiệp của con người như bác sĩ hay nhà văn,... thông qua sự học hỏi. Giống như một đứa trẻ, AI cần được dạy những gì nên học và không nên học, nên làm và không nên làm, dựa trên ba điều luật mà tôi không nhớ tên do một nhà khoa học đặt ra được lập trình vào AI để đảm bảo AI đi đúng hướng và hoạt động theo đúng ý của con người. Nhưng bản thân 3 điều luật này vẫn chưa hoàn thiện và vẫn còn lỗ hổng. Vậy nên vấn đề AI sẽ thay thế được hoàn toàn con người trong tương lai là một việc theo tôi là không khả thi.
Có một điều tôi khá thích đó là các tác giả đồng tình với nhận định AI sẽ không thể hoạt động tốt nếu không có sự can thiệp của con người. Đó cũng là nhận định của đa số chúng ta về AI.
Tác giả đã móc nối rất tốt giữa hai nội dung về kiến thức AI với việc phát triển doanh nghiệp, mạng, nói chung là tất cả những vấn đề bên lề có liên quan đến AI mà con người cần và sẽ phải đối mặt. Mọi thứ được viết rất thực tế và gần gũi, hoàn toàn không nói quá chút nào. Những nội dung đó khiến chúng ta thực sự phải lưu tâm và chuẩn bị tâm lý vững vàng để làm quen với sự có mặt của AI trong mọi việc của cuộc sống.
Khoa học là vô tận, AI chỉ là một phần của sự vô tận đó nhưng kiến thức về nó lại vô vàn. Chúng ta còn cần phải tìm hiểu rất rất nhiều về nó, còn phải bàn về AI dài dài, bài viết của tôi chỉ là một nhận định rất nhỏ của một người trẻ mới tìm hiểu về AI nên có thể còn những sai sót. Nhưng dẫu sao tất cả mới chỉ bắt đầu, không ai có thể nói trước được điều gì trước sự phát triển như vũ bão của khoa học công nghệ. Những cuộc hội thảo sẽ được tổ chức nhiều hơn, những cuộc trò chuyện về AI sẽ không bao giờ có hồi kết và những cuốn sách sẽ còn tiếp tục được xuất bản. Và nhiệm vụ của chúng ta là phải chuẩn bị một hành trang kiến thức, kỹ năng thật tốt để đối mặt với những điều "chưa biết là chưa biết" sẽ xảy đến trong tương lai.
#AI #technology_revolution_4.0
#ajay_agrawal #joshua_ gans #avi_goldfarb
#1980books
Có thể nói đây là một trong những cuốn sách hiếm hoi về AI được xuất bản tại Việt Nam. Trước đó phải kể đến Phát minh cuối cùng của James Barrat đã tạo ra một làn sóng dữ dội trong cộng đồng người thích tìm hiểu về AI với nhiều ý kiến trái chiều. Một thời gian sau, AI Trong cuộc cách mạng 4.0 được xuất bản như một điều hiển nhiên và một lần nữa, sức hút của đề tài này lại càng nóng hơn bao giờ hết.
Trước nhất phải nói đây là một cuốn sách pha trộn giữa học thuyết khoa học về trí tuệ nhân tạo với định hướng phát triển doanh nghiệp. Đây chính là điểm nhấn quan trọng và chói sáng của toàn bộ tác phẩm. Đặt nội dung vào trong một ví dụ thực tế, toàn bộ AI được diễn giải một cách không thể rõ ràng, cụ thể hơn. Khác với Phát minh cuối cùng bàn về những kiến thức lý thuyết rất nặng nề thì cuốn sách này lại chi tiết hơn. 3 tác giả đã cố gắng để khiến những người không có nhiều kiến thức về khoa học cũng có thể hiểu được. Đặc biệt là những nhà doanh nghiệp sẽ phải đặc biệt lưu tâm đến cuốn này bởi tính thực tế của nó và coi nó như là "sách giáo khoa" của mình.
Bố cục sách rất rõ ràng được trình bày theo trình tự dễ hiểu. Sự dự đoán là nội dung chính trong vấn đề mà tác giả đưa ra về AI. Kỳ thực ban đầu tôi không đồng tính với quan điểm của tác giả khi cho rằng AI dùng sự dự đoán để tự động hóa chính nó bao gồm công việc, hành động, phản xạ,... Nhưng càng đọc tôi lại càng bị hút vào, thôi thúc đi tìm câu trả lời cho thắc mắc của bản thân và sự không thỏa mãn ban đầu đó. Tôi đã không thất vọng. Càng đào sâu tôi lại càng bất ngờ trước những luận điểm gồm nhiều luận cứ sắc bén của tác giả. Phải nói rằng tôi muốn phản bác lại tất cả những điều đó nhưng không thể, lập luận của tác giả quá chặt chẽ khiến tất cả những sự phản biện trong đầu tôi lần lượt bị bác bỏ qua mỗi lần lật trang.
Nói như vậy không có nghĩa là cuốn sách này không có những lỗ hổng. Ví dụ như việc AI có thể học được cảm xúc của con người. Tôi hoàn toàn không đồng tình với quan điểm này. Hay như AI có thể thay thế một số nghề nghiệp của con người như bác sĩ hay nhà văn,... thông qua sự học hỏi. Giống như một đứa trẻ, AI cần được dạy những gì nên học và không nên học, nên làm và không nên làm, dựa trên ba điều luật mà tôi không nhớ tên do một nhà khoa học đặt ra được lập trình vào AI để đảm bảo AI đi đúng hướng và hoạt động theo đúng ý của con người. Nhưng bản thân 3 điều luật này vẫn chưa hoàn thiện và vẫn còn lỗ hổng. Vậy nên vấn đề AI sẽ thay thế được hoàn toàn con người trong tương lai là một việc theo tôi là không khả thi.
Có một điều tôi khá thích đó là các tác giả đồng tình với nhận định AI sẽ không thể hoạt động tốt nếu không có sự can thiệp của con người. Đó cũng là nhận định của đa số chúng ta về AI.
Tác giả đã móc nối rất tốt giữa hai nội dung về kiến thức AI với việc phát triển doanh nghiệp, mạng, nói chung là tất cả những vấn đề bên lề có liên quan đến AI mà con người cần và sẽ phải đối mặt. Mọi thứ được viết rất thực tế và gần gũi, hoàn toàn không nói quá chút nào. Những nội dung đó khiến chúng ta thực sự phải lưu tâm và chuẩn bị tâm lý vững vàng để làm quen với sự có mặt của AI trong mọi việc của cuộc sống.
Khoa học là vô tận, AI chỉ là một phần của sự vô tận đó nhưng kiến thức về nó lại vô vàn. Chúng ta còn cần phải tìm hiểu rất rất nhiều về nó, còn phải bàn về AI dài dài, bài viết của tôi chỉ là một nhận định rất nhỏ của một người trẻ mới tìm hiểu về AI nên có thể còn những sai sót. Nhưng dẫu sao tất cả mới chỉ bắt đầu, không ai có thể nói trước được điều gì trước sự phát triển như vũ bão của khoa học công nghệ. Những cuộc hội thảo sẽ được tổ chức nhiều hơn, những cuộc trò chuyện về AI sẽ không bao giờ có hồi kết và những cuốn sách sẽ còn tiếp tục được xuất bản. Và nhiệm vụ của chúng ta là phải chuẩn bị một hành trang kiến thức, kỹ năng thật tốt để đối mặt với những điều "chưa biết là chưa biết" sẽ xảy đến trong tương lai.
#AI #technology_revolution_4.0
#ajay_agrawal #joshua_ gans #avi_goldfarb
#1980books
September 7, 2018
Rating - 8.5
Fascinating & pointed read that focuses on both the strategic & operational aspect of AI - great use of corporate examples to underly and/or analogize the decisions required to succeed going forward
Writing is very clear & concise, chapters are kept relatively short & summations at the end of each chapter reinforce the key points; Economic hypotheses are consistent, driving home a new, new normal
Key Points
Alexa has replaced the parent as the all-knowing source of information in the eyes of a child
The fundamental of AI is to reduce the cost/improve the accuracy of prediction
Prediction is the process of filling in missing information. Prediction takes the data/information you have and uses it to generate information you don't have
Incremental improvements are beneficial if the costs are great - an improvement from 98% to 99.9% improves performance by a factor of 20
Recent advances in machine learning has improved prediction accuracy so that they can do translation and navigation tasks
Three types of data - Input (for Prediction), Training (train the AI) and Feedback (improve performance
Data has a diminishing rate of return - each additional unit of data improves the prediction less than the last unit
Known Knowns - the things we know we know
Known Unknowns - we know the things that we do not know
Unknown Unknowns - we don't know what we don't know
Prediction machines scale, humans don't
Judgement - determining the relative payoff associated with each possible decision outcome
Reward Function Engineering - job of determining rewards to the various actions based on AI prediction; Must know the organizational needs & capabilities
Protect the bombers where the bullet holes were not example - assessed the bombers not making it back and how they were getting shot down
Satisficing - Humans cannot cope w complexity so they take 'good enough' shortcuts to complete their objectives
Externalities - costs felt by others, not the key decision-makers
People thought the same joke was not as funny if it was created by a machine versus a human
The introduction of AI to a task does not mean that full automation will take place
Implementation of AI has four implications on jobs:
1. Augment (spreadsheets/book-keepers)
2. Contract (fulfilment centres)
3. Re-constituted (tasks added or subtracted)
4. Shift emphasis on specific skills
Training data builds the prediction machine, input data is used to power predictions & feedback data is used to improve
AI has significant impacts on capital, labor & data
An AI-First Strategy is one that places prediction accuracy as the top objective even if there is a trade-off on revenue, volumes & user experience
NYC Fire Department lawsuit against blacks & hispanics - emphasized reading comprehension which is not a required core skill
Six Types of AI Risk
1. Predictions can lead to discrimination
2. AI is ineffective when data is sparse
3. Incorrect input data can fool prediction machines
4. Diversity of prediction machines involve a trade-off between individual & system-level outcomes
5. Prediction machines can be interrogated, exposing the company to IP theft
6. Feedback can be manipulated to teach the AI destructive behaviour
Y Combinator is running experiments where everyone is provided a basic income (offsets job less)
Open AI is designed to prevent a private organization from dominating AI
As AI replaces prediction tasks, more people will focus on judgment tasks
Key policy question is not about whether AI will bring benefits but how they will be distributed
Labor's share of income has consistently been reduced by Income from Capital - leads to further inequality
China will eventually lead the world in AI - investing the most money & building the most infrastructure
Three Trade-offs of AI
1. Productivity vs DIstribution (furthering the wealth divide)
2. Innovation vs Competition (reducing competition)
3. Performance vs Privacy
Fascinating & pointed read that focuses on both the strategic & operational aspect of AI - great use of corporate examples to underly and/or analogize the decisions required to succeed going forward
Writing is very clear & concise, chapters are kept relatively short & summations at the end of each chapter reinforce the key points; Economic hypotheses are consistent, driving home a new, new normal
Key Points
Alexa has replaced the parent as the all-knowing source of information in the eyes of a child
The fundamental of AI is to reduce the cost/improve the accuracy of prediction
Prediction is the process of filling in missing information. Prediction takes the data/information you have and uses it to generate information you don't have
Incremental improvements are beneficial if the costs are great - an improvement from 98% to 99.9% improves performance by a factor of 20
Recent advances in machine learning has improved prediction accuracy so that they can do translation and navigation tasks
Three types of data - Input (for Prediction), Training (train the AI) and Feedback (improve performance
Data has a diminishing rate of return - each additional unit of data improves the prediction less than the last unit
Known Knowns - the things we know we know
Known Unknowns - we know the things that we do not know
Unknown Unknowns - we don't know what we don't know
Prediction machines scale, humans don't
Judgement - determining the relative payoff associated with each possible decision outcome
Reward Function Engineering - job of determining rewards to the various actions based on AI prediction; Must know the organizational needs & capabilities
Protect the bombers where the bullet holes were not example - assessed the bombers not making it back and how they were getting shot down
Satisficing - Humans cannot cope w complexity so they take 'good enough' shortcuts to complete their objectives
Externalities - costs felt by others, not the key decision-makers
People thought the same joke was not as funny if it was created by a machine versus a human
The introduction of AI to a task does not mean that full automation will take place
Implementation of AI has four implications on jobs:
1. Augment (spreadsheets/book-keepers)
2. Contract (fulfilment centres)
3. Re-constituted (tasks added or subtracted)
4. Shift emphasis on specific skills
Training data builds the prediction machine, input data is used to power predictions & feedback data is used to improve
AI has significant impacts on capital, labor & data
An AI-First Strategy is one that places prediction accuracy as the top objective even if there is a trade-off on revenue, volumes & user experience
NYC Fire Department lawsuit against blacks & hispanics - emphasized reading comprehension which is not a required core skill
Six Types of AI Risk
1. Predictions can lead to discrimination
2. AI is ineffective when data is sparse
3. Incorrect input data can fool prediction machines
4. Diversity of prediction machines involve a trade-off between individual & system-level outcomes
5. Prediction machines can be interrogated, exposing the company to IP theft
6. Feedback can be manipulated to teach the AI destructive behaviour
Y Combinator is running experiments where everyone is provided a basic income (offsets job less)
Open AI is designed to prevent a private organization from dominating AI
As AI replaces prediction tasks, more people will focus on judgment tasks
Key policy question is not about whether AI will bring benefits but how they will be distributed
Labor's share of income has consistently been reduced by Income from Capital - leads to further inequality
China will eventually lead the world in AI - investing the most money & building the most infrastructure
Three Trade-offs of AI
1. Productivity vs DIstribution (furthering the wealth divide)
2. Innovation vs Competition (reducing competition)
3. Performance vs Privacy
November 17, 2018
”Prediction machines will have their most immediate impact at the decision level. But decisions have six other key elements. When someone (or something) makes a decision, they take input data from the world that enables a prediction. That prediction is possible because training occurred about relationships between different types of data and which data is most closely associated with a situation. Combining the prediction with judgment on what matters, the decision maker can then choose an action. The action leads to an outcome (which has an associated reward or payoff). The outcome is a consequence of the decision. It is needed to provide a complete picture. The outcome may also provide feedback to help improve the next prediction.”
As the Canadian rock band Rush used to sing: “If you choose not to decide, you still have made a choice.”
Decision making is at the core of most occupations. From doctors deciding what medicine to prescribe, to truck drivers deciding how to respond to route closures and traffic accidents, we are constantly making decisions in our day to day life.
The anatomy of a decision is formed by trying to predict the outcome of the different actions that could be taken given an input. Judgment will decide which is our preferable outcome.
For instance, we can predict that it will rain tomorrow and we can decide if given that we want to take an umbrella with us or not. Since we don’t want to carry with the umbrella all day, our judgment will be not to take it even though there is a high probability that it will rain and we might get wet.
From predicting if an image contains a cat to predicting what is the answer to “Alexa, what is the capital of France?”, AI has become extremely good at guessing what would humans respond to those questions.
Still, AI’s judgment is yet not good enough to substitute us in a lot of cases.
In Prediction Machines, three economists look at the future of AI by framing AI as cheap predictions. If predictions suddenly become cheap, how will businesses that rely on them evolve? What will all these humans that work predicting outcomes focus on instead? Will they focus solely on the judgment part of a decision? How will that affect their jobs?
Probably one of the best books on the topic of AI that I’ve read so far. This is a must read!
As the Canadian rock band Rush used to sing: “If you choose not to decide, you still have made a choice.”
Decision making is at the core of most occupations. From doctors deciding what medicine to prescribe, to truck drivers deciding how to respond to route closures and traffic accidents, we are constantly making decisions in our day to day life.
The anatomy of a decision is formed by trying to predict the outcome of the different actions that could be taken given an input. Judgment will decide which is our preferable outcome.
For instance, we can predict that it will rain tomorrow and we can decide if given that we want to take an umbrella with us or not. Since we don’t want to carry with the umbrella all day, our judgment will be not to take it even though there is a high probability that it will rain and we might get wet.
From predicting if an image contains a cat to predicting what is the answer to “Alexa, what is the capital of France?”, AI has become extremely good at guessing what would humans respond to those questions.
Still, AI’s judgment is yet not good enough to substitute us in a lot of cases.
In Prediction Machines, three economists look at the future of AI by framing AI as cheap predictions. If predictions suddenly become cheap, how will businesses that rely on them evolve? What will all these humans that work predicting outcomes focus on instead? Will they focus solely on the judgment part of a decision? How will that affect their jobs?
Probably one of the best books on the topic of AI that I’ve read so far. This is a must read!
May 30, 2019
Having worked for global eBusiness Agencies (iXL, Pixelpark) of the first peak of the internet innovation cycle 20 years ago I’m still excited about the ongoing disruptive power of digitalization. Using a self-driving Tesla-car, Google Maps or Alexa or just your credit card everyone has had or will soon have his Artificial Intelligence (AI) moment. Therefore, everybody is speculating about the chances but also the threats by AI. But who gives substantial orientation? After reading „Life 3.0: Being Human in the Age of Artificial Intelligence“ by Max Tegmark (a very helpful hint by my former Pixelpark colleague Barbara Daliri Freyduni) I followed the recommendation of serial founder and AI-expert Stephan Uhrenbacher (FLIO.com Digital Airports, Qype, 9flats.com) to read the current Bestseller “Prediction Machines” from Ajay Agrawal, Joshua Gans and Avi Golfarb. As a fellow at Creative Destruction Lab (CDL), a high profiled accelerator focused on Machine Learning in Toronto, Stephan got in personal touch with these three economists from at the University of Toronto's Rotman School of Management, who have cofounded CDL, and supported their publication. Based on various empiric studies and their personal insights from working with AI-pioneers at CDL Agrawal, Gans and Golfarb explain why prediction machines makes predictions better, faster and increasingly cheaper and become extremely powerful but as well with specific limitations. They conclude: The combination of humans and machines generates best results by compensating the differing weaknesses of both. How does such a collaboration translate into a business environment? In two ways: (1) The machine provides an initial prediction that humans can combine and use for their own assessment, judgement and decision making or (2) The machines provide parallel or afterwards a second opinion which helps e.g. higher-level managers to ensure high performance of their employees Agrawal, Gans Golfarb reflect extensively also the different types of risks carried by AI, but they inspire in a pragmatic way to think about the new opportunities in an open way. A must-read for everybody who wants to understand the economic logic of AI-power behind the technical development.
February 2, 2021
I was debating whether this book deserved more than a one-star-rating. But in the end, this book deserves at least some credit. The authors tried to write scientifically, convincingly and I kinda liked the little AI anecdotes (Although I was already familiar with most of them).
Why can I not recommend this book? To literally anyone?
1) It lacks a scientific and thorough explanation of the concept of AI.
Well, I guess you could argue that this book just wants take an economic approach. You could argue that pages are limited and you have to focus on the most important stuff (although 220 pages are not that much).
Then why does it enter the polemic, philosophic sphere in the last chapters? Quote: "Is this the end of the world as we know it? Not yet, but it is the end of this book. Companies are deploying AIs right now."
This does not sound like the rational, economic perspective the authours wanted to convey...
2) The writing style is exhausting
It reads like an essay written by an econ-undergrad student, only a bit elongated. English is not my first language, so I can't criticize grammar and style; but what I can tell is: it switches between overly scientific and colloquial, while at the same time being overly repetitive.
When a chapter consists of 3 pages, you don't need a 1 page summary of that chapter!
3) Who was this book written for?
The authors claim to have written this book for businessmen who plan on incorporating AI in their company.
Yet I can't think of any entrepreneur who knows so little about AI that you have to explain to him what it is used for. If you have heard of AI, you don't need to read this book. It contains no extra knowledge. It also doesn't help you incorporate it.
4) It contains no useful information I could use for "my business".
Interesting questions would be:
- How much does it cost me to hire a company to code my AI?
- Is it worth the money?
- Who do I hire? Do I have to specify the predictions or is it the software-engineers job?
- etc. etc.
What I got:
- Here is a canvas you can fill in to "use AI"
Conclusion:
Too shallow, too much repetition, not enough explanation, too many anecdotes, too many promises
Why can I not recommend this book? To literally anyone?
1) It lacks a scientific and thorough explanation of the concept of AI.
Well, I guess you could argue that this book just wants take an economic approach. You could argue that pages are limited and you have to focus on the most important stuff (although 220 pages are not that much).
Then why does it enter the polemic, philosophic sphere in the last chapters? Quote: "Is this the end of the world as we know it? Not yet, but it is the end of this book. Companies are deploying AIs right now."
This does not sound like the rational, economic perspective the authours wanted to convey...
2) The writing style is exhausting
It reads like an essay written by an econ-undergrad student, only a bit elongated. English is not my first language, so I can't criticize grammar and style; but what I can tell is: it switches between overly scientific and colloquial, while at the same time being overly repetitive.
When a chapter consists of 3 pages, you don't need a 1 page summary of that chapter!
3) Who was this book written for?
The authors claim to have written this book for businessmen who plan on incorporating AI in their company.
Yet I can't think of any entrepreneur who knows so little about AI that you have to explain to him what it is used for. If you have heard of AI, you don't need to read this book. It contains no extra knowledge. It also doesn't help you incorporate it.
4) It contains no useful information I could use for "my business".
Interesting questions would be:
- How much does it cost me to hire a company to code my AI?
- Is it worth the money?
- Who do I hire? Do I have to specify the predictions or is it the software-engineers job?
- etc. etc.
What I got:
- Here is a canvas you can fill in to "use AI"
Conclusion:
Too shallow, too much repetition, not enough explanation, too many anecdotes, too many promises
August 10, 2018
The rundown- this is a really good book about the current positioning of machine learning within a socio-economic framework. It is written from the perspective of a few economists, so there is almost no technical jargon. You only need to know in broad terms what ML is or can do. The book is full, although not overflowing with examples, but the ones used are quite good and illustrate the points the authors are trying to make really well.
Now a little spoiler alert- the main argument of the book is that ML cannot yet produce artificial intelligence ( at least as fas as multiple basic tasks are concerned), but what it currently does really well is fill in some information gaps in our decision-making process, by providing predictions. The authors argue that it is then up to humans to actually make a judgment based on the presumed impact of each prediction. The value of this judgment is almost impossible to code because the combinations are so mamy and sometimes rare.
Ultimately, using AI in the eyes of economists does not mean a new AI economy. All the supply and demand still holds. What changes is the speed of decisions, because predictions are cheaper and faster. This also means, like with amything else that improves in terms of speed and price, that the value of the human employment in that role will decrease, with jobs either focusing on other tasks (the bus driver for a school bus would still be there to supervise the children) or lose considerable income. Additionally, the authors disagree with Nick Bostrom's paperclip AI that will use all of the world's resources. They believe proper in the invisible hand that will lead this AI to trade rather than destroy, i.e. to some sort of an equilibrium.
Bottom line, the book is really good, a welcomed and different view about the place of AI from the perspective of economists, with a lot of food for thought and discussion!
Now a little spoiler alert- the main argument of the book is that ML cannot yet produce artificial intelligence ( at least as fas as multiple basic tasks are concerned), but what it currently does really well is fill in some information gaps in our decision-making process, by providing predictions. The authors argue that it is then up to humans to actually make a judgment based on the presumed impact of each prediction. The value of this judgment is almost impossible to code because the combinations are so mamy and sometimes rare.
Ultimately, using AI in the eyes of economists does not mean a new AI economy. All the supply and demand still holds. What changes is the speed of decisions, because predictions are cheaper and faster. This also means, like with amything else that improves in terms of speed and price, that the value of the human employment in that role will decrease, with jobs either focusing on other tasks (the bus driver for a school bus would still be there to supervise the children) or lose considerable income. Additionally, the authors disagree with Nick Bostrom's paperclip AI that will use all of the world's resources. They believe proper in the invisible hand that will lead this AI to trade rather than destroy, i.e. to some sort of an equilibrium.
Bottom line, the book is really good, a welcomed and different view about the place of AI from the perspective of economists, with a lot of food for thought and discussion!
This entire review has been hidden because of spoilers.
September 25, 2019
I’ve been putting this book off for a little while, and I’m not really sure why. I think I thought it was going to be slow going, but it turns out that I whizzed through it in a day or so. I wasn’t too worried about the subject matter, even though it’s non-fiction about artificial intelligence from the Harvard Business Review press.
This was actually recommended to me by a client of mine, Emmanuel Fombu, who specialises in writing and talking about the future of healthcare. It was a good recommendation on his part, especially because there is some stuff here that focusses directly on the impact of artificial intelligence in the healthcare industry. But it’s not just healthcare that’s covered here, and indeed I think the authors did a good job of covering a wide variety of different use cases.
It was also interesting to read right now because my current “bedtime book” is The Enigma by Andrew Hodges, a biography of Alan Turing. His work had a huge impact on the development of artificial intelligence and what the authors here call “prediction machines”, and indeed one of the main tests that an AI must pass if it’s to be able to pass as a human is named after Turing: the Turing Test.
All in all then, if you have an interest in AI and the way in which it’s changing our society, read this.
This was actually recommended to me by a client of mine, Emmanuel Fombu, who specialises in writing and talking about the future of healthcare. It was a good recommendation on his part, especially because there is some stuff here that focusses directly on the impact of artificial intelligence in the healthcare industry. But it’s not just healthcare that’s covered here, and indeed I think the authors did a good job of covering a wide variety of different use cases.
It was also interesting to read right now because my current “bedtime book” is The Enigma by Andrew Hodges, a biography of Alan Turing. His work had a huge impact on the development of artificial intelligence and what the authors here call “prediction machines”, and indeed one of the main tests that an AI must pass if it’s to be able to pass as a human is named after Turing: the Turing Test.
All in all then, if you have an interest in AI and the way in which it’s changing our society, read this.
Displaying 1 - 30 of 281 reviews