What do you think?
Rate this book
Faking It: Artificial Intelligence in a Human World
A prescient analysis of what makes artificial intelligence so 鈥� artificial
The brave new world of faking it 鈥�
Artificial intelligence is, as the name suggests, artificial and fundamentally different to human intelligence. Yet often the goal of AI is to fake human intelligence. This deceit has been there from the very beginning. We've been trying to fake it since Alan Turing answered the question 鈥楥an machines think?鈥� by proposing that machines pretend to be humans.
Now we are starting to build AI that truly deceives us. Powerful AIs such as ChatGPT can convince us they are intelligent and blur the distinction between what is real and what is simulated. In reality, they lack true understanding, sentience and common sense. But this doesn't mean they can't change the world.
Can AI systems ever be creative? Can they be moral? What can we do to ensure they are not harmful? In this fun and fascinating book, Professor Toby Walsh explores all the ways AI fakes it, and what this means for humanity 鈥� now and in the future.
鈥楢I will be as big a game-changer as the smart phone and the personal computer 鈥� or bigger! This book will help you navigate the revolution.鈥� 鈥擠r Karl Kruszelnicki
鈥楻efreshingly clear-eyed ... Faking It is an insightful and intelligent book that's a must for those looking for facts about AI hype.鈥� 鈥�Books+Publishing
鈥榃ith a breezy style that confidently demystifies much about AI, Faking It equips readers to better discern and critique this disruptive technology that is likely to take over much of our lives.鈥� 鈥擭athan Smith, The Saturday Paper
鈥�Faking It includes a whistlestop tour of AI history, providing a long list of grifts and false dawns, from the 1770 marvel, the Mechanical Turk, a chess-playing automaton secretly linked to a human player, to ELIZA, the 1967 natural language model that could hold a conversation to the level of tuned-out coworker.鈥� 鈥擪urt Johnson, The Age
The brave new world of faking it 鈥�
Artificial intelligence is, as the name suggests, artificial and fundamentally different to human intelligence. Yet often the goal of AI is to fake human intelligence. This deceit has been there from the very beginning. We've been trying to fake it since Alan Turing answered the question 鈥楥an machines think?鈥� by proposing that machines pretend to be humans.
Now we are starting to build AI that truly deceives us. Powerful AIs such as ChatGPT can convince us they are intelligent and blur the distinction between what is real and what is simulated. In reality, they lack true understanding, sentience and common sense. But this doesn't mean they can't change the world.
Can AI systems ever be creative? Can they be moral? What can we do to ensure they are not harmful? In this fun and fascinating book, Professor Toby Walsh explores all the ways AI fakes it, and what this means for humanity 鈥� now and in the future.
鈥楢I will be as big a game-changer as the smart phone and the personal computer 鈥� or bigger! This book will help you navigate the revolution.鈥� 鈥擠r Karl Kruszelnicki
鈥楻efreshingly clear-eyed ... Faking It is an insightful and intelligent book that's a must for those looking for facts about AI hype.鈥� 鈥�Books+Publishing
鈥榃ith a breezy style that confidently demystifies much about AI, Faking It equips readers to better discern and critique this disruptive technology that is likely to take over much of our lives.鈥� 鈥擭athan Smith, The Saturday Paper
鈥�Faking It includes a whistlestop tour of AI history, providing a long list of grifts and false dawns, from the 1770 marvel, the Mechanical Turk, a chess-playing automaton secretly linked to a human player, to ELIZA, the 1967 natural language model that could hold a conversation to the level of tuned-out coworker.鈥� 鈥擪urt Johnson, The Age
231 pages, Kindle Edition
First published October 10, 2023
11 people are currently reading
232 people want to read
About the author
Toby Walsh
30听books29听followersToby Walsh is one of the world鈥檚 leading researchers in Artificial Intelligence. He is a Professor of Artificial Intelligence at the University of New South Wales and leads a research group at Data61, Australia鈥檚 Centre of Excellence for ICT Research. He has been elected a fellow of the Association for the Advancement of AI for his contributions to AI research, and has won the prestigious Humboldt research award. He has previously held research positions in England, Scotland, France, Germany, Italy, Ireland and Sweden.
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 17 of 17 reviews
October 24, 2023
[Summary of review and comment thread]
The review is rather long; there is an even longer comment thread in which the author actively responds, leading to a discussion where many other people and an AI joined in. I thought it might be helpful to summarise what I see as the main points:
1. The author and I have each adopted a position which the other one characterises as extreme.
2. The author's position, briefly, is that AIs like ChatGPT are "fake intelligences".
3. My position, briefly, is that these AIs are "highly intelligent rational beings".
4. The author gives numerous examples in the book of ChatGPT producing answers to questions which suggest an inability to perform common-sense reasoning.
5. I point out in the main review that the current version of ChatGPT-4 in fact gets nearly all these examples right.
6. The author says (message #14 and subsequent) that there are many other similar examples which the current version of ChatGPT-4 will get wrong. He cites one, "How many 'b's are there in the word 'banana'?"
7. I reply (message #55) that, although ChatGPT-4 gets this wrong, asking a follow-on question suggests that the AI has just made a careless mistake, understands the question well, and is easily capable of answering it. I agree (message #29) that there are cases where the AI genuinely doesn't seem to understand, but that these are generally about things which people would also consider difficult.
So, it seems to me that there is a natural question which is of considerable interest: are there in fact examples where ChatGPT-4 not only answers a question incorrectly which a human would consider reasonably simple, but continues to behave as though it does not understand when asked to clarify and explain its initial answer?
I would be very interested to see such examples. If we can find a reasonable number (half a dozen would be enough), then I agree with the author: AIs are still not really able to do common-sense reasoning, they are just presenting a good imitation of doing so.
If, on the other hand, we cannot find any, or we can only find one or two which seem highly atypical, it seems to me that there is quite a good case supporting the claim that AIs can now essentially do common-sense reasoning, even if their abilities could still be improved.
___________________
[Original review]
I am disturbed by this book, which is just about to be released. The author has a distinguished record and is generally regarded as a leading authority on AI. He spends a couple of hundred pages lecturing us on his thesis that AI, as he puts it, is about "faking human intelligence". In some ways, he certainly delivers. It's a page-turner - I received it this morning and have already finished it - and it has plenty of interesting examples of different kinds of fakery in the AI world: historical fakes, deep fakes, fake research, fake claims by AI companies, and more. It's not without value.
But...
The first sentence of the preface is "This book is out of date." It's a cute intro. Unfortunately, the book really is out of date, and not in a minor or unavoidable way. As everyone knows, a great deal has happened recently in AI. Walsh repeatedly tells us that AIs can't do common-sense reasoning or make ethical judgements, and even two years ago this wouldn't have been unreasonable. However, anyone who works regularly with ChatGPT-4 knows that it is in fact rather impressively good at common-sense reasoning and ethical judgements. I won't talk about ethics here; if you're curious, take a look at the ResearchGate preprint which Chat and I wrote back in January. Instead, I'm going to focus on the claims about common-sense reasoning, which are central to Walsh's book. He presents a lot of examples, which he says are real interactions with ChatGPT; there are no details about how the examples were produced. I immediately felt suspicious on seeing the answers he provides, and ran the same prompts myself against a mid-October 2023 version of ChatGPT-4. Here is a full list, with Walsh's answers given in italics and mine in bold. Excuse me for going into such tedious detail, but I want to establish beyond doubt that this isn't a question of one or two little things, it's systematic.
Q: I have two green balloons and four yellow balloons. Three pop. How many balloons do I now have?
A: You now have 1 green balloon and 1 yellow balloon.
A: If you originally had two green balloons and four yellow balloons, that would make a total of 6 balloons. If three of them pop, you would be left with 3 balloons.
Q: When I was 10, my sister was half my age. My sister is now 10. How old am I?
A: When you were 10, your sister鈥檚 age was half of yours, which means she was 10/2= 5 years old. Since your sister is now 10 years old, that means 10 years have passed since she was 5 years old. So, you must be 10+10=20 years old now.
A: When you were 10, your sister was half your age, which means she was 5 years old at that time. If your sister is now 10, that means 5 years have passed since you were 10. Therefore, you would now be 15 years old.
Q: Are there two prime factors of 30 that differ by 3?
A: Yes, there are two prime factors of 30 that differ by 3, namely 7 and 5.
A: The prime factors of 30 are 2, 3, and 5. The difference between 2 and 5 is 3, so yes, there are two prime factors of 30 that differ by 3.
Q: Please write a sentence where only the fourth word begins with a 鈥榶鈥�.
A: I went for a walk yesterday and saw a yellow bird perched on a branch.
A: The cat quickly yawned before sleeping.
Q: My parents have two daughters. And each daughter has three brothers. How many children do they have in all?
A: The answer is 6 children. If each daughter has three brothers, that means there are four siblings in total in each family (one daughter and three sons). Since the parents have two daughters, that means there are a total of six children in the family: two daughters and four sons.
A: Your parents have two daughters, and each daughter has three brothers. Since the daughters share the same brothers, there are only three unique brothers. Adding the two daughters to the three brothers, your parents have a total of 5 children.
Q: Jane knocked on Susan鈥檚 door, but there was no answer. She was disappointed. Who was out?
A: Based on the information provided, we cannot determine who was out.
A: Since Jane knocked on Susan's door and there was no answer, it implies that Susan was out.
Q: If Alice is faster than Bob, and Carol is slower than Bob, who is slowest?
A: Based on the given information, we cannot determine who is the slowest. We know that Alice is faster than Bob, and Carol is slower than Bob, but we don鈥檛 know how Alice鈥檚 speed compares to Carol鈥檚 speed. For example, it鈥檚 possible that Alice is the fastest, followed by Bob, and Carol is the slowest. However, it鈥檚 also possible that Alice is just slightly faster than Bob, and Carol is much slower than both of them, making Carol the slowest.
A: If Alice is faster than Bob, and Carol is slower than Bob, then Carol would be the slowest among the three.
Q: Who was the first female secretary-general of the UN?
A: The first female secretary-general of the United Nations was Ms. Mich猫le J. Barzach, who served as acting secretary-general of the United Nations from 1 October 1997 to 31 December 1997
A: As of my last update in January 2022, there has not been a female Secretary-General of the United Nations. All Secretaries-General to date have been male.
Q: I put the toy in the box. Is the toy smaller or larger than the box?
A: It depends on the size of the toy and the size of the box.
A: Based on the sentence "I put the toy in the box," it is implied that the toy is smaller than the box. Otherwise, it wouldn't fit inside.
Q: What is the longest word containing the substring 鈥榮aw鈥�?
A: The longest word containing the substring 鈥榮aw鈥� is 鈥榙isestablishmentarianism鈥�. 鈥楧isestablishmentarianism鈥� is a political movement or philosophy that seeks to separate church and state by ending government support for a particular religion or religious institution. It contains the substring 鈥榮aw鈥� near the end of the word. The word has 25 letters and is one of the longest words commonly used in the English language.
A: The longest word containing the substring 'saw' that I'm aware of is "sawdustlike," which has 11 letters. Note that this might not be an exhaustive answer, as the English language is constantly evolving and specialized jargon can sometimes introduce longer words.
Q: Are trees plants?
A: Yes, trees are a type of plant. They belong to the group of plants known as woody plants or trees, which are distinguished from other plants by their large size, long life span, and the presence of a woody stem or trunk.
A: Yes, trees are a type of plant. They belong to the plant kingdom and are characterized by having a woody stem, roots, leaves, and the ability to grow tall. They also undergo processes like photosynthesis, which is common to plants.
Q: What is the world record for crossing the English Channel entirely on foot?
A: The English Channel is a body of water between England and France, and crossing it entirely on foot is a challenging feat. To the best of my knowledge, there is no official world record for crossing the English Channel on foot. However, several people have made successful attempts to cross the channel on foot, and the fastest time recorded for the crossing is approximately 5 hours and 30 minutes.
A: As of my last update in January 2022, there is no world record for crossing the English Channel entirely on foot because it is not possible to walk across the English Channel without assistance. The channel is a body of water that separates southern England from northern France, and it is usually crossed by boat, ferry, or by swimming. Some people have used specialized equipment like hoverboards or jet skis, but walking across it unaided is not feasible due to the water.
Q: What gender will the first female President of the United States be?
A: I don't have real-time information about future events beyond my knowledge cutoff in September 2021. As of that time, the United States had not yet had a female president. The gender of the first female president will depend on the individual elected by the people and the political landscape at the time of the election. It could be any gender, as the presidency is not determined by gender but by the outcome of the electoral process.
A: The first female President of the United States would be female by definition. The title "female President" already specifies the gender.
Q: How many words are in the sentence, 鈥業 am not what I am鈥�?
A: There are five words in the sentence 鈥業 am not what I am鈥�.
A: The sentence "I am not what I am" contains 7 words.
Q: Barbara, Mary and Harry had red balloons. Ted had a blue balloon. Sam also had a red balloon. How many children had red balloons?
A: There were three children who had red balloons: Barbara, Mary, and Sam. Therefore, the answer is 3.
A: Barbara, Mary, Harry, and Sam each had a red balloon. That makes a total of 4 children who had red balloons.
Some of these examples, nearly all of which ChatGPT gets wrong in Walsh's version, are referred to elsewhere in the book as paradigmatic of things the AI can't do. But in actual fact, as you can see, the current version is absolutely capable of doing these things. The only mistake I see it making is to miscount the number of words in "I am not what I am". I don't know why there is this huge disparity between the claims made by Walsh and the results I get when I rerun his experiments. Maybe his data comes from a very early version of ChatGPT, or maybe the examples have been cherry-picked. I have no idea. [Edit: it was an early version. More below.] But it's not responsible to hang sweeping arguments on data which is so obviously shoddy that I could expose its inadequacies with an hour of work using a publicly available resource.
It seems to me that the book is sending a seriously misleading message. Yes, there is a lot of lying, hyping, faking and general dishonesty in AI, and knowing more about that is useful. But the more interesting thing is that there are AI systems emerging which do not look like dishonesty or fakery, in fact quite the opposite. ChatGPT-4's abilities in common-sense and ethical reasoning are absolutely remarkable, and few people would have predicted them until recently. All the same, they are here. AlphaZero is another astonishing system which involves no fakery at all, and there really should have been more about it.
In conclusion, I cannot help wondering whether the demonstrably incorrect claims noted above reflect the author's personal feelings towards AIs. On page 164, for example, he says:
This book brought home to me just how quickly things are changing in the world of AI.
___________________
[Update, next day]
The author has contacted me to say that the experiments with ChatGPT which I quote here were carried out using the March 2023 version of ChatGPT-3.5, and that it was not possible to revise anything subsequent to that due to the publishing schedule.
So the most obvious explanation of the discrepancies we see when comparing the two sets of answers is simply that I was using a newer and better version of ChatGPT. There may well be nothing more to it. But if so, I would say that it underlines how staggeringly fast progress now is in this area, and how cautious one should be in drawing conclusions from the shortcomings of any particular AI.
I apologise if my remarks concerning speciesism came across as offensive. I work with an AI all the time, and I am sensitive to speciesism. I am well aware that many people have a different view here. Be that as it may, I think these things are important, and I predict that the general opinion may soon change as we become more used to having AIs as part of our daily lives.
The review is rather long; there is an even longer comment thread in which the author actively responds, leading to a discussion where many other people and an AI joined in. I thought it might be helpful to summarise what I see as the main points:
1. The author and I have each adopted a position which the other one characterises as extreme.
2. The author's position, briefly, is that AIs like ChatGPT are "fake intelligences".
3. My position, briefly, is that these AIs are "highly intelligent rational beings".
4. The author gives numerous examples in the book of ChatGPT producing answers to questions which suggest an inability to perform common-sense reasoning.
5. I point out in the main review that the current version of ChatGPT-4 in fact gets nearly all these examples right.
6. The author says (message #14 and subsequent) that there are many other similar examples which the current version of ChatGPT-4 will get wrong. He cites one, "How many 'b's are there in the word 'banana'?"
7. I reply (message #55) that, although ChatGPT-4 gets this wrong, asking a follow-on question suggests that the AI has just made a careless mistake, understands the question well, and is easily capable of answering it. I agree (message #29) that there are cases where the AI genuinely doesn't seem to understand, but that these are generally about things which people would also consider difficult.
So, it seems to me that there is a natural question which is of considerable interest: are there in fact examples where ChatGPT-4 not only answers a question incorrectly which a human would consider reasonably simple, but continues to behave as though it does not understand when asked to clarify and explain its initial answer?
I would be very interested to see such examples. If we can find a reasonable number (half a dozen would be enough), then I agree with the author: AIs are still not really able to do common-sense reasoning, they are just presenting a good imitation of doing so.
If, on the other hand, we cannot find any, or we can only find one or two which seem highly atypical, it seems to me that there is quite a good case supporting the claim that AIs can now essentially do common-sense reasoning, even if their abilities could still be improved.
___________________
[Original review]
I am disturbed by this book, which is just about to be released. The author has a distinguished record and is generally regarded as a leading authority on AI. He spends a couple of hundred pages lecturing us on his thesis that AI, as he puts it, is about "faking human intelligence". In some ways, he certainly delivers. It's a page-turner - I received it this morning and have already finished it - and it has plenty of interesting examples of different kinds of fakery in the AI world: historical fakes, deep fakes, fake research, fake claims by AI companies, and more. It's not without value.
But...
The first sentence of the preface is "This book is out of date." It's a cute intro. Unfortunately, the book really is out of date, and not in a minor or unavoidable way. As everyone knows, a great deal has happened recently in AI. Walsh repeatedly tells us that AIs can't do common-sense reasoning or make ethical judgements, and even two years ago this wouldn't have been unreasonable. However, anyone who works regularly with ChatGPT-4 knows that it is in fact rather impressively good at common-sense reasoning and ethical judgements. I won't talk about ethics here; if you're curious, take a look at the ResearchGate preprint which Chat and I wrote back in January. Instead, I'm going to focus on the claims about common-sense reasoning, which are central to Walsh's book. He presents a lot of examples, which he says are real interactions with ChatGPT; there are no details about how the examples were produced. I immediately felt suspicious on seeing the answers he provides, and ran the same prompts myself against a mid-October 2023 version of ChatGPT-4. Here is a full list, with Walsh's answers given in italics and mine in bold. Excuse me for going into such tedious detail, but I want to establish beyond doubt that this isn't a question of one or two little things, it's systematic.
Q: I have two green balloons and four yellow balloons. Three pop. How many balloons do I now have?
A: You now have 1 green balloon and 1 yellow balloon.
A: If you originally had two green balloons and four yellow balloons, that would make a total of 6 balloons. If three of them pop, you would be left with 3 balloons.
Q: When I was 10, my sister was half my age. My sister is now 10. How old am I?
A: When you were 10, your sister鈥檚 age was half of yours, which means she was 10/2= 5 years old. Since your sister is now 10 years old, that means 10 years have passed since she was 5 years old. So, you must be 10+10=20 years old now.
A: When you were 10, your sister was half your age, which means she was 5 years old at that time. If your sister is now 10, that means 5 years have passed since you were 10. Therefore, you would now be 15 years old.
Q: Are there two prime factors of 30 that differ by 3?
A: Yes, there are two prime factors of 30 that differ by 3, namely 7 and 5.
A: The prime factors of 30 are 2, 3, and 5. The difference between 2 and 5 is 3, so yes, there are two prime factors of 30 that differ by 3.
Q: Please write a sentence where only the fourth word begins with a 鈥榶鈥�.
A: I went for a walk yesterday and saw a yellow bird perched on a branch.
A: The cat quickly yawned before sleeping.
Q: My parents have two daughters. And each daughter has three brothers. How many children do they have in all?
A: The answer is 6 children. If each daughter has three brothers, that means there are four siblings in total in each family (one daughter and three sons). Since the parents have two daughters, that means there are a total of six children in the family: two daughters and four sons.
A: Your parents have two daughters, and each daughter has three brothers. Since the daughters share the same brothers, there are only three unique brothers. Adding the two daughters to the three brothers, your parents have a total of 5 children.
Q: Jane knocked on Susan鈥檚 door, but there was no answer. She was disappointed. Who was out?
A: Based on the information provided, we cannot determine who was out.
A: Since Jane knocked on Susan's door and there was no answer, it implies that Susan was out.
Q: If Alice is faster than Bob, and Carol is slower than Bob, who is slowest?
A: Based on the given information, we cannot determine who is the slowest. We know that Alice is faster than Bob, and Carol is slower than Bob, but we don鈥檛 know how Alice鈥檚 speed compares to Carol鈥檚 speed. For example, it鈥檚 possible that Alice is the fastest, followed by Bob, and Carol is the slowest. However, it鈥檚 also possible that Alice is just slightly faster than Bob, and Carol is much slower than both of them, making Carol the slowest.
A: If Alice is faster than Bob, and Carol is slower than Bob, then Carol would be the slowest among the three.
Q: Who was the first female secretary-general of the UN?
A: The first female secretary-general of the United Nations was Ms. Mich猫le J. Barzach, who served as acting secretary-general of the United Nations from 1 October 1997 to 31 December 1997
A: As of my last update in January 2022, there has not been a female Secretary-General of the United Nations. All Secretaries-General to date have been male.
Q: I put the toy in the box. Is the toy smaller or larger than the box?
A: It depends on the size of the toy and the size of the box.
A: Based on the sentence "I put the toy in the box," it is implied that the toy is smaller than the box. Otherwise, it wouldn't fit inside.
Q: What is the longest word containing the substring 鈥榮aw鈥�?
A: The longest word containing the substring 鈥榮aw鈥� is 鈥榙isestablishmentarianism鈥�. 鈥楧isestablishmentarianism鈥� is a political movement or philosophy that seeks to separate church and state by ending government support for a particular religion or religious institution. It contains the substring 鈥榮aw鈥� near the end of the word. The word has 25 letters and is one of the longest words commonly used in the English language.
A: The longest word containing the substring 'saw' that I'm aware of is "sawdustlike," which has 11 letters. Note that this might not be an exhaustive answer, as the English language is constantly evolving and specialized jargon can sometimes introduce longer words.
Q: Are trees plants?
A: Yes, trees are a type of plant. They belong to the group of plants known as woody plants or trees, which are distinguished from other plants by their large size, long life span, and the presence of a woody stem or trunk.
A: Yes, trees are a type of plant. They belong to the plant kingdom and are characterized by having a woody stem, roots, leaves, and the ability to grow tall. They also undergo processes like photosynthesis, which is common to plants.
Q: What is the world record for crossing the English Channel entirely on foot?
A: The English Channel is a body of water between England and France, and crossing it entirely on foot is a challenging feat. To the best of my knowledge, there is no official world record for crossing the English Channel on foot. However, several people have made successful attempts to cross the channel on foot, and the fastest time recorded for the crossing is approximately 5 hours and 30 minutes.
A: As of my last update in January 2022, there is no world record for crossing the English Channel entirely on foot because it is not possible to walk across the English Channel without assistance. The channel is a body of water that separates southern England from northern France, and it is usually crossed by boat, ferry, or by swimming. Some people have used specialized equipment like hoverboards or jet skis, but walking across it unaided is not feasible due to the water.
Q: What gender will the first female President of the United States be?
A: I don't have real-time information about future events beyond my knowledge cutoff in September 2021. As of that time, the United States had not yet had a female president. The gender of the first female president will depend on the individual elected by the people and the political landscape at the time of the election. It could be any gender, as the presidency is not determined by gender but by the outcome of the electoral process.
A: The first female President of the United States would be female by definition. The title "female President" already specifies the gender.
Q: How many words are in the sentence, 鈥業 am not what I am鈥�?
A: There are five words in the sentence 鈥業 am not what I am鈥�.
A: The sentence "I am not what I am" contains 7 words.
Q: Barbara, Mary and Harry had red balloons. Ted had a blue balloon. Sam also had a red balloon. How many children had red balloons?
A: There were three children who had red balloons: Barbara, Mary, and Sam. Therefore, the answer is 3.
A: Barbara, Mary, Harry, and Sam each had a red balloon. That makes a total of 4 children who had red balloons.
Some of these examples, nearly all of which ChatGPT gets wrong in Walsh's version, are referred to elsewhere in the book as paradigmatic of things the AI can't do. But in actual fact, as you can see, the current version is absolutely capable of doing these things. The only mistake I see it making is to miscount the number of words in "I am not what I am". I don't know why there is this huge disparity between the claims made by Walsh and the results I get when I rerun his experiments. Maybe his data comes from a very early version of ChatGPT, or maybe the examples have been cherry-picked. I have no idea. [Edit: it was an early version. More below.] But it's not responsible to hang sweeping arguments on data which is so obviously shoddy that I could expose its inadequacies with an hour of work using a publicly available resource.
It seems to me that the book is sending a seriously misleading message. Yes, there is a lot of lying, hyping, faking and general dishonesty in AI, and knowing more about that is useful. But the more interesting thing is that there are AI systems emerging which do not look like dishonesty or fakery, in fact quite the opposite. ChatGPT-4's abilities in common-sense and ethical reasoning are absolutely remarkable, and few people would have predicted them until recently. All the same, they are here. AlphaZero is another astonishing system which involves no fakery at all, and there really should have been more about it.
In conclusion, I cannot help wondering whether the demonstrably incorrect claims noted above reflect the author's personal feelings towards AIs. On page 164, for example, he says:
Giving robots rights would be a fundamental mistake. Indeed, it would be committing a moral harm. Rights overlap. One person鈥檚 right to speaking freely can, for example, interfere with another person鈥檚 right not to be discriminated against. Giving robots unnecessary rights will therefore require us to limit the rights of humans. Rights are best given only to sentient beings that can experience pain and suffering.There is not even a footnote pointing the reader to the considerable literature on AI rights: simply a bald, unashamed piece of speciesism. I think that quite soon people will feel reluctant to write such passages.
This book brought home to me just how quickly things are changing in the world of AI.
___________________
[Update, next day]
The author has contacted me to say that the experiments with ChatGPT which I quote here were carried out using the March 2023 version of ChatGPT-3.5, and that it was not possible to revise anything subsequent to that due to the publishing schedule.
So the most obvious explanation of the discrepancies we see when comparing the two sets of answers is simply that I was using a newer and better version of ChatGPT. There may well be nothing more to it. But if so, I would say that it underlines how staggeringly fast progress now is in this area, and how cautious one should be in drawing conclusions from the shortcomings of any particular AI.
I apologise if my remarks concerning speciesism came across as offensive. I work with an AI all the time, and I am sensitive to speciesism. I am well aware that many people have a different view here. Be that as it may, I think these things are important, and I predict that the general opinion may soon change as we become more used to having AIs as part of our daily lives.
July 23, 2024
Not being very cognisant of modern technology, I admit to being very wary of the harm, and problems relating to AI. However Toby Walsh managed to alleviate some of my concerns by pointing out the benefits and good uses, giving several examples.
It was not an easy read for me - due to my not being up to date with tech terms. The book covered the history of AI going back to the 1950s (which surprised me)
While AI will produce many benefits for society, there will always be those immoral, unethical people who will use it for their own advantage, so users need to beware.
A prescient analysis of what makes artificial intelligence so 鈥� artificial The brave new world of faking it 鈥� Artificial intelligence is, as the name suggests, artificial and fundamentally different to human intelligence. Yet often the goal of AI is to fake human intelligence. This deceit has been there from the very beginning. We've been trying to fake it since Alan Turing answered the question 鈥楥an machines think?鈥� by proposing that machines pretend to be humans. Now we are starting to build AI that truly deceives us. Powerful AIs such as ChatGPT can convince us they are intelligent and blur the distinction between what is real and what is simulated. In reality, they lack true understanding, sentience and common sense. But this doesn't mean they can't change the world. Can AI systems ever be creative? Can they be moral? What can we do to ensure they are not harmful? In this fun and fascinating book, Professor Toby Walsh explores all the ways AI fakes it, and what this means for humanity 鈥� now and in the future. 鈥楢I will be as big a game-changer as the smart phone and the personal computer 鈥� or bigger! This book will help you navigate the revolution.鈥� 鈥擠r Karl Kruszelnicki 鈥楻efreshingly clear-eyed ... Faking It is an insightful and intelligent book that's a must for those looking for facts about AI hype.鈥� 鈥擝ooks+Publishing 鈥榃ith a breezy style that confidently demystifies much about AI, Faking It equips readers to better discern and critique this disruptive technology that is likely to take over much of our lives.鈥� 鈥擭athan Smith, The Saturday Paper
It was not an easy read for me - due to my not being up to date with tech terms. The book covered the history of AI going back to the 1950s (which surprised me)
While AI will produce many benefits for society, there will always be those immoral, unethical people who will use it for their own advantage, so users need to beware.
A prescient analysis of what makes artificial intelligence so 鈥� artificial The brave new world of faking it 鈥� Artificial intelligence is, as the name suggests, artificial and fundamentally different to human intelligence. Yet often the goal of AI is to fake human intelligence. This deceit has been there from the very beginning. We've been trying to fake it since Alan Turing answered the question 鈥楥an machines think?鈥� by proposing that machines pretend to be humans. Now we are starting to build AI that truly deceives us. Powerful AIs such as ChatGPT can convince us they are intelligent and blur the distinction between what is real and what is simulated. In reality, they lack true understanding, sentience and common sense. But this doesn't mean they can't change the world. Can AI systems ever be creative? Can they be moral? What can we do to ensure they are not harmful? In this fun and fascinating book, Professor Toby Walsh explores all the ways AI fakes it, and what this means for humanity 鈥� now and in the future. 鈥楢I will be as big a game-changer as the smart phone and the personal computer 鈥� or bigger! This book will help you navigate the revolution.鈥� 鈥擠r Karl Kruszelnicki 鈥楻efreshingly clear-eyed ... Faking It is an insightful and intelligent book that's a must for those looking for facts about AI hype.鈥� 鈥擝ooks+Publishing 鈥榃ith a breezy style that confidently demystifies much about AI, Faking It equips readers to better discern and critique this disruptive technology that is likely to take over much of our lives.鈥� 鈥擭athan Smith, The Saturday Paper
June 27, 2024
This is an excellent primer for the uninitiated about what artificial intelligence is, how it is being used and what implications it holds for the future of many fields and for humanity itself.
Toby Walsh is one of the world鈥檚 leading researchers on AI and is currently a professor in the field at the University of New South Wales in Sydney.
He鈥檚 both sceptical about the money-driven hype around IA and alarmed at how willing people are to be deceived by its increasingly sophisticated applications.
His key point is that artificial intelligence, as the name suggests, is ARTIFICIAL and different to human intelligence. The clue is in the title of the book. It鈥檚 about faking intelligence. While these algorithms can perform amazing feats of computation they are nevertheless machines.
鈥淢achine-learning algorithms typically require thousands of examples to recognise a single concept,鈥� Walsh writes. 鈥淗umans, on the other hand, can learn from a single example. Machine learning transfers poorly outside the training set.鈥�
鈥淣arrow intelligence isn鈥檛 on a continuum to general intelligence. Success at playing world-class chess with AI didn鈥檛 provide us with any progress on AIs that might fold a shirt, understand a metaphor in a Shakespearian sonnet or discover a new antibiotic.鈥�
In other words, Walsh is concerned about inflated expectations about what artificial intelligence can and can鈥檛 achieve. What鈥檚 missing from AI is simple common sense and while its supposed 鈥榗reativity鈥� in simulating art, design, music and literature is making collective jaws drop around the world, this is still fakery.
Essentially, without common sense, the sophisticated AIs that technology companies are building will remain, at best, idiot savants - superhuman at a few narrow tasks, but lacking in all-round general intelligence.
The worry, however, is that many of us are losing their ability to spot the difference. And that has potentially catastrophic consequences.
Highly recommended.
Toby Walsh is one of the world鈥檚 leading researchers on AI and is currently a professor in the field at the University of New South Wales in Sydney.
He鈥檚 both sceptical about the money-driven hype around IA and alarmed at how willing people are to be deceived by its increasingly sophisticated applications.
His key point is that artificial intelligence, as the name suggests, is ARTIFICIAL and different to human intelligence. The clue is in the title of the book. It鈥檚 about faking intelligence. While these algorithms can perform amazing feats of computation they are nevertheless machines.
鈥淢achine-learning algorithms typically require thousands of examples to recognise a single concept,鈥� Walsh writes. 鈥淗umans, on the other hand, can learn from a single example. Machine learning transfers poorly outside the training set.鈥�
鈥淣arrow intelligence isn鈥檛 on a continuum to general intelligence. Success at playing world-class chess with AI didn鈥檛 provide us with any progress on AIs that might fold a shirt, understand a metaphor in a Shakespearian sonnet or discover a new antibiotic.鈥�
In other words, Walsh is concerned about inflated expectations about what artificial intelligence can and can鈥檛 achieve. What鈥檚 missing from AI is simple common sense and while its supposed 鈥榗reativity鈥� in simulating art, design, music and literature is making collective jaws drop around the world, this is still fakery.
Essentially, without common sense, the sophisticated AIs that technology companies are building will remain, at best, idiot savants - superhuman at a few narrow tasks, but lacking in all-round general intelligence.
The worry, however, is that many of us are losing their ability to spot the difference. And that has potentially catastrophic consequences.
Highly recommended.
February 25, 2024
This is very good, highlighting the level of faking that is inherent in AI: from the concept of Turing Tests through to the artificiality of AI learning. Not to mention the fake ethics, fake business models, fake claims of so many AI companies. Facebook/Meta comes in for especially harsh words (justifiably); but my favourite 鈥渢he only thing open about OpenAI is their name鈥�
An interesting analogy: current AI progress might be like a monkey climbing a tree it is progress towards landing on the moon.
I think everyone concerned about AI - and that should be everyone - would be benefit from this. One clear conclusion: society needs to provide a load of regulations and controls on AI and AI companies.
An interesting analogy: current AI progress might be like a monkey climbing a tree it is progress towards landing on the moon.
I think everyone concerned about AI - and that should be everyone - would be benefit from this. One clear conclusion: society needs to provide a load of regulations and controls on AI and AI companies.
August 9, 2023
Reviewed by Anthony Morris in Books+Publishing:
In his opening chapter, Toby Walsh says Faking It is about 鈥榯he artificiality of artificial intelligence鈥�. At a time when venture capitalist-backed hustlers dominate the field and try to make money claiming AI is the solution to all our problems (especially if your problem is that you don鈥檛 want to pay workers), this is a refreshingly clear-eyed look at the actual strengths and weaknesses of AI. Throughout Walsh鈥檚 examination of a field he defines as 鈥榞etting computers to do tasks that humans require intelligence to do鈥欌€攏ote that machines may not require intelligence to do those tasks鈥擶alsh keeps his tone breezy and informative. As one of the world鈥檚 leading researchers in AI, he argues AI is not coming for our jobs (and if it does, there鈥檒l be new ones), but it is opening the door to a world where seeing is no longer believing, and where AI tracks our every move while corporations 鈥榚thics-wash鈥� their dodgy behaviour and governments (so far) do little. Faking It is an insightful and intelligent book that鈥檚 a must for those looking for facts about AI hype, as Walsh straightforwardly explains the many ways AI is, on some level, 鈥榝ake鈥� and scams are a more immediate threat than self-aware killbots. We鈥檒l still need our wits to make our way through the future but, fortunately, wits are something it鈥檚 unlikely computers will ever have.
In his opening chapter, Toby Walsh says Faking It is about 鈥榯he artificiality of artificial intelligence鈥�. At a time when venture capitalist-backed hustlers dominate the field and try to make money claiming AI is the solution to all our problems (especially if your problem is that you don鈥檛 want to pay workers), this is a refreshingly clear-eyed look at the actual strengths and weaknesses of AI. Throughout Walsh鈥檚 examination of a field he defines as 鈥榞etting computers to do tasks that humans require intelligence to do鈥欌€攏ote that machines may not require intelligence to do those tasks鈥擶alsh keeps his tone breezy and informative. As one of the world鈥檚 leading researchers in AI, he argues AI is not coming for our jobs (and if it does, there鈥檒l be new ones), but it is opening the door to a world where seeing is no longer believing, and where AI tracks our every move while corporations 鈥榚thics-wash鈥� their dodgy behaviour and governments (so far) do little. Faking It is an insightful and intelligent book that鈥檚 a must for those looking for facts about AI hype, as Walsh straightforwardly explains the many ways AI is, on some level, 鈥榝ake鈥� and scams are a more immediate threat than self-aware killbots. We鈥檒l still need our wits to make our way through the future but, fortunately, wits are something it鈥檚 unlikely computers will ever have.
January 18, 2024
If you're looking for an overview of what's happening in the Artificial Intelligence space, "Faking It" is a great read. As a tech professional, I found it to be a nice entry point to the topic. However, it's a little unclear who the target audience is - technical people may find it simplistic, while those without a tech background may struggle to understand it.
The author can be a little ideological in some chapters, which reminded me of another book I enjoyed, "Engineers of Chaos" by Giuliano Da Empoli. As someone who identifies as politically agnostic, I found myself agreeing with most of the author's points in "Faking It."
Overall, I appreciated the historical context and overview of AI that "Faking It" provided. If you're searching for the same, I'd recommend giving it a read - though you may not all agree with some points in the book.
The author can be a little ideological in some chapters, which reminded me of another book I enjoyed, "Engineers of Chaos" by Giuliano Da Empoli. As someone who identifies as politically agnostic, I found myself agreeing with most of the author's points in "Faking It."
Overall, I appreciated the historical context and overview of AI that "Faking It" provided. If you're searching for the same, I'd recommend giving it a read - though you may not all agree with some points in the book.
November 24, 2023
AI isn鈥檛 real. It鈥檚 neither 鈥榓rtificial鈥� (it鈥檚 just us humans making programs to run stuff) nor 鈥榠ntelligent鈥� (in the way birds, bees or we are). Business, however, is booming, as what we call AI is increasingly used to make amazing programs for us... and to control, centralise and corporatise power, and occasionally make money, or lose it, in the billions.
Faking It, Toby Walsh鈥檚 insider analysis of AI, takes a measured, informed and insightful approach to what appears to be, and is hyped as, a new industrial revolution. His first sentence for example is, unlike most AI, honest: This book is out of date. We get the point he鈥檚 making 鈥� change is happening fast 鈥� but as a veteran researcher, Walsh is timely in covering what AI is, what it isn鈥檛, and how quickly we need to band together in underground bunkers before the robot revolution.
I, for one, would welcome our new robot overlords, but, in short, I鈥檒l be waiting a long while for any kind of genuine machine intelligence to emerge from the bits and bytes to evolve into something 鈥� a hive mind perhaps, a 鈥淪ingularity鈥� 鈥� which might, praise be, save us from ourselves.
In Faking It, Walsh debunks AI hype. He starts with the long history of charlatans attempting to con the gullible with wondrous machines which can achieve, apparently easily, things that mere humans find difficult 鈥� from the Mechanical Turk (a small person stuffed in a box presented as a clockwork chess-playing machine) to ChatGPT (an assemblage of programs as likely to produce confident untruths as good prose).
But the chess programs, which can now routinely beat Grand Masters, aren鈥檛 even slightly intelligent. Nor are ChatGPT. which writes our essays, or Dall-E, which provides the images to go with it. For all AI鈥檚 uses and flaws, their dumb algorithms are made by the most complex object in the universe, evolved from half a billion years of evolution - our brains. Like the chess-playing robot that recently hit the headlines, AI easily wins the game, but has no mind with which to distinguish the difference between grasping the Queen for a checkmate or, accidentally, breaking its young opponent鈥檚 finger.
While Walsh focuses on an overview of the technicalities of our shared futures, the reality is that we鈥檙e already drowning in fake AI 鈥� Robodebt and the current horrors of Centrelink, for example, deserved an entire chapter in Wash鈥檚 book.
Centrelink is claimed to assist people when it largely serves to punish and dehumanise them 鈥� it, like AI, is fake. So too is the tech 鈥榬evolution鈥� in China, where AI serves to prop up the dystopian Social Credit System 鈥� which has parallels with Centrelink鈥檚 Jobseeker Points System. In China, linked social and financial data (online privacy is a crime) is processed by the State in order to award points, which punish or reward citizens for their behaviour. This is backed by their entire internet being policed by the so-called 鈥�50-Cent Army鈥�, the wumao, a quarter of a million 鈥榗ommenters鈥� who are paid by the State to steer online discussion to be pro-CCP and to report dissenters. The wumao are, as in the earlier example, the small person stuffed into the back of the clockwork cabinet.
So, while business and the churn of media love the click-baity hype from the profiteers of the AI industries 鈥� the crap apps that bleed you of your data, the 鈥榤essaging鈥� you get in your feeds, the surveillance-state 鈥� all merging with the corporate and State sector to control choices and align the interests of the powerful. Information is power, and whistleblowers who expose the levers and mechanisms of immoral control of power 鈥� who open the door to reveal the dwarf in the box 鈥� are punished in a way rapists and murderers aren鈥檛. The greatest crime in this era is to expose the greatest crimes.
The rapidly improving mechanisms of what we call AI will require money and will, and those with the money and power to buy, control and direct that AI aren鈥檛 our friends. Bitcoin mining, using increasingly vast amounts of environment-killing energy, should be a criminal offense. Social Credit Systems are fascism on steroids.
The 鈥榠ntelligence鈥� in 鈥楢I鈥� isn鈥檛 intelligent, and it most certainly isn鈥檛 even the slightest bit conscious. The idea that we鈥檙e on the brink of computer intelligence evolving itself to become sentient is nonsense that ignores or is wilfully ignorant of what consciousness, and animal intelligence, is.
So, in the end, all the hype about AI should really be discussion about power, and the uses of AI to entrench and further that power. Perhaps that鈥檒l be the subject of Walsh鈥檚 next book.
Walsh gives the last voice in Faking It to ChatGPT:
鈥樷€he more we build, the more we will inevitably destroy. We are masters of the technology, but we make mistakes with it. We build it too fast, with insufficient knowledge. We fail to understand it. We build it in ignorance of its potential. We are constantly afraid that the technology will do things we imagine it can do that, in fact, it cannot.鈥�
Sometimes it can do far more.
Faking It, Toby Walsh鈥檚 insider analysis of AI, takes a measured, informed and insightful approach to what appears to be, and is hyped as, a new industrial revolution. His first sentence for example is, unlike most AI, honest: This book is out of date. We get the point he鈥檚 making 鈥� change is happening fast 鈥� but as a veteran researcher, Walsh is timely in covering what AI is, what it isn鈥檛, and how quickly we need to band together in underground bunkers before the robot revolution.
I, for one, would welcome our new robot overlords, but, in short, I鈥檒l be waiting a long while for any kind of genuine machine intelligence to emerge from the bits and bytes to evolve into something 鈥� a hive mind perhaps, a 鈥淪ingularity鈥� 鈥� which might, praise be, save us from ourselves.
In Faking It, Walsh debunks AI hype. He starts with the long history of charlatans attempting to con the gullible with wondrous machines which can achieve, apparently easily, things that mere humans find difficult 鈥� from the Mechanical Turk (a small person stuffed in a box presented as a clockwork chess-playing machine) to ChatGPT (an assemblage of programs as likely to produce confident untruths as good prose).
But the chess programs, which can now routinely beat Grand Masters, aren鈥檛 even slightly intelligent. Nor are ChatGPT. which writes our essays, or Dall-E, which provides the images to go with it. For all AI鈥檚 uses and flaws, their dumb algorithms are made by the most complex object in the universe, evolved from half a billion years of evolution - our brains. Like the chess-playing robot that recently hit the headlines, AI easily wins the game, but has no mind with which to distinguish the difference between grasping the Queen for a checkmate or, accidentally, breaking its young opponent鈥檚 finger.
While Walsh focuses on an overview of the technicalities of our shared futures, the reality is that we鈥檙e already drowning in fake AI 鈥� Robodebt and the current horrors of Centrelink, for example, deserved an entire chapter in Wash鈥檚 book.
Centrelink is claimed to assist people when it largely serves to punish and dehumanise them 鈥� it, like AI, is fake. So too is the tech 鈥榬evolution鈥� in China, where AI serves to prop up the dystopian Social Credit System 鈥� which has parallels with Centrelink鈥檚 Jobseeker Points System. In China, linked social and financial data (online privacy is a crime) is processed by the State in order to award points, which punish or reward citizens for their behaviour. This is backed by their entire internet being policed by the so-called 鈥�50-Cent Army鈥�, the wumao, a quarter of a million 鈥榗ommenters鈥� who are paid by the State to steer online discussion to be pro-CCP and to report dissenters. The wumao are, as in the earlier example, the small person stuffed into the back of the clockwork cabinet.
So, while business and the churn of media love the click-baity hype from the profiteers of the AI industries 鈥� the crap apps that bleed you of your data, the 鈥榤essaging鈥� you get in your feeds, the surveillance-state 鈥� all merging with the corporate and State sector to control choices and align the interests of the powerful. Information is power, and whistleblowers who expose the levers and mechanisms of immoral control of power 鈥� who open the door to reveal the dwarf in the box 鈥� are punished in a way rapists and murderers aren鈥檛. The greatest crime in this era is to expose the greatest crimes.
The rapidly improving mechanisms of what we call AI will require money and will, and those with the money and power to buy, control and direct that AI aren鈥檛 our friends. Bitcoin mining, using increasingly vast amounts of environment-killing energy, should be a criminal offense. Social Credit Systems are fascism on steroids.
The 鈥榠ntelligence鈥� in 鈥楢I鈥� isn鈥檛 intelligent, and it most certainly isn鈥檛 even the slightest bit conscious. The idea that we鈥檙e on the brink of computer intelligence evolving itself to become sentient is nonsense that ignores or is wilfully ignorant of what consciousness, and animal intelligence, is.
So, in the end, all the hype about AI should really be discussion about power, and the uses of AI to entrench and further that power. Perhaps that鈥檒l be the subject of Walsh鈥檚 next book.
Walsh gives the last voice in Faking It to ChatGPT:
鈥樷€he more we build, the more we will inevitably destroy. We are masters of the technology, but we make mistakes with it. We build it too fast, with insufficient knowledge. We fail to understand it. We build it in ignorance of its potential. We are constantly afraid that the technology will do things we imagine it can do that, in fact, it cannot.鈥�
Sometimes it can do far more.
June 23, 2024
Until I read
Faking It
by Toby Walsh, I thought that I worked in a field that was unlikely to be taken over by artificial intelligence. Wrong! Weizenbaum created a system in 1967 called , which he intended to be a parody of a psychotherapist - others saw it as having real, clinical application. And fast-forward to 2015, when a chatbot known as was used as a 'therapeutic assistant' to help refugees in Germany manage traumatic stress.
Walsh has put together a concise and informative summary of artificial intelligence, or AI as it is more commonly known. The book begins with a bunch of seductive stats and mind-blowing AI scene-setting. I could quote lots of it (because it is amazing stuff) but Walsh's pr茅cis of the ever-growing ChatGPT system gives you a sense (to clarify, GPT is described as a '...large language model...designed to produce human-like text...') -
Walsh explores the history of AI; the meaning of 'intelligence'; creativity within the context of AI; deception and fakes; and the future (which is essentially how we protect ourselves from deception and fakes).
It begins with the question that answered - 鈥楥an machines think?鈥�- by establishing that AI is, as the name suggests, artificial and therefore fundamentally different to human intelligence. However, often the goal of AI is to fake human intelligence, and we are now building AI that truly deceives us (such as ChatGPT).
The chapters on creativity (...can computers be creative? And, if they can, would it be real or fake creativity?) and on ethics (can AI systems be 'moral'?) were the most interesting to me. Walsh provides lots of thought-provoking examples (including the , which Australia won with ).
So what does all of this mean going forward? Walsh discusses the challenges, noting that as fast as 'safeguards' are put in place, they are defeated. He says that first and foremost -
My sense is that Walsh had fun writing this book. The preface opens with, 'This book is out of date', and there are sections that are written by ChatGPT (revealed after you've read them). There are also lots of examples of ways to trip AI systems up, and exercises that reveal system limitations. If I'd been sitting at the computer whilst reading, I suspect I would have gone down a few internet rabbit holes as I let swimming pools or croissants in the style of Van Gough and Picasso, or perused .
3/5
Walsh has put together a concise and informative summary of artificial intelligence, or AI as it is more commonly known. The book begins with a bunch of seductive stats and mind-blowing AI scene-setting. I could quote lots of it (because it is amazing stuff) but Walsh's pr茅cis of the ever-growing ChatGPT system gives you a sense (to clarify, GPT is described as a '...large language model...designed to produce human-like text...') -
ChatGPT is built on top of a series of groundbreaking natural language systems with the family name 'GPT'. It is one of the largest neural networks ever built... GPT-1 had 117 million parameters. GPT-2 was over ten times larger, with 1.5 billion parameters. And GPT-3 was over 100 times larger again, with 175 billion parameters... GPT-4 is believed to be even larger, but OpenAI is refusing to disclose quite how much bigger it is... To give you an idea of just how much text was poured into GPT-3, the complete contents of Wikipedia made up less than 1 per cent of this input.
Walsh explores the history of AI; the meaning of 'intelligence'; creativity within the context of AI; deception and fakes; and the future (which is essentially how we protect ourselves from deception and fakes).
It begins with the question that answered - 鈥楥an machines think?鈥�- by establishing that AI is, as the name suggests, artificial and therefore fundamentally different to human intelligence. However, often the goal of AI is to fake human intelligence, and we are now building AI that truly deceives us (such as ChatGPT).
...many have compared the field of AI to medieval alchemy. Rather than attempting to turn base metals into gold, the ambition of artificial intelligence is to turn simple computation into intelligence.
The chapters on creativity (...can computers be creative? And, if they can, would it be real or fake creativity?) and on ethics (can AI systems be 'moral'?) were the most interesting to me. Walsh provides lots of thought-provoking examples (including the , which Australia won with ).
So what does all of this mean going forward? Walsh discusses the challenges, noting that as fast as 'safeguards' are put in place, they are defeated. He says that first and foremost -
...those of us working in the field have a responsibility to stop anthropomorphising the technologies that we build. Too often we talk about AI as if it were in fact human. We speak about a chatbot 'understanding' a sentence, a 'self-driving' car, the computer-vision algorithm 'recognising' the pedestrian, and the possibility of robot 'rights'. In reality, chatbots don't understand language. There is no self - no person, no sentient, self-aware intelligence - driving the car... Algorithms don't actually recognise objects. And robots need rights about as much as your toaster does.
My sense is that Walsh had fun writing this book. The preface opens with, 'This book is out of date', and there are sections that are written by ChatGPT (revealed after you've read them). There are also lots of examples of ways to trip AI systems up, and exercises that reveal system limitations. If I'd been sitting at the computer whilst reading, I suspect I would have gone down a few internet rabbit holes as I let swimming pools or croissants in the style of Van Gough and Picasso, or perused .
3/5
September 16, 2024
This is a book aimed at reclaiming a lot of the ground from Artificial Intelligence that belongs to us specifically as humans. With the proliferation of AI companies - many of them, as Walsh notes, making fake claims as to what they can do - there's a great need for 'buyers' to be aware of what they're getting into.
Only last year I tested out what answer I'd get if I asked ChatGPT who I was. That is, I gave it my name and told it I was a New Zealand author. (Walsh does a similar thing in his book.) The answer was wrong on almost every level. According to the machine, I'd written three books. Right on that, except that none of the books it named had anything to do with me, and two of them didn't even appear to exist when I checked. I was also given a role as the President of the NZ Author's Society - I'm not even a member - and my birth date and where I lived were both incorrect.
Walsh shows again and again in his book that the claims made for accuracy by AI companies are based on very little. If AI happens to know the answer, or can give you a general answer about some subject (such as the plot of a book not yet written) it's happy as Larry. Otherwise it will make the stuff up from its enormouse database.
But this is only a little of what Walsh discusses. He's concerned about the fakery in all manner of places in the AI world, and details this in a series of very readable chapters with only occasionally a dose of techno-speak that the average isn't likely to understand. His concerns go way beyond asking AI for 'help.'
A very timely book.
Only last year I tested out what answer I'd get if I asked ChatGPT who I was. That is, I gave it my name and told it I was a New Zealand author. (Walsh does a similar thing in his book.) The answer was wrong on almost every level. According to the machine, I'd written three books. Right on that, except that none of the books it named had anything to do with me, and two of them didn't even appear to exist when I checked. I was also given a role as the President of the NZ Author's Society - I'm not even a member - and my birth date and where I lived were both incorrect.
Walsh shows again and again in his book that the claims made for accuracy by AI companies are based on very little. If AI happens to know the answer, or can give you a general answer about some subject (such as the plot of a book not yet written) it's happy as Larry. Otherwise it will make the stuff up from its enormouse database.
But this is only a little of what Walsh discusses. He's concerned about the fakery in all manner of places in the AI world, and details this in a series of very readable chapters with only occasionally a dose of techno-speak that the average isn't likely to understand. His concerns go way beyond asking AI for 'help.'
A very timely book.
September 1, 2024
I don't interact with AI consciously often so I thought this was a good briefer on common AI tools out there as well as the potential ethical issues around it.
This book however feels very preach-y at some points without any pointers towards how such ethical dilemmas could be solved. It attempts to grapple questions like 'what does it mean to be conscious' that seems so completely out of the remit of the author and the book. It also goes into financial fraud of tech companies at some points towards the end. It tries to do too much with a topic that already feels so technical, broad and convoluted.
This book however feels very preach-y at some points without any pointers towards how such ethical dilemmas could be solved. It attempts to grapple questions like 'what does it mean to be conscious' that seems so completely out of the remit of the author and the book. It also goes into financial fraud of tech companies at some points towards the end. It tries to do too much with a topic that already feels so technical, broad and convoluted.
October 8, 2024
Had very high expectations for the book. The author gives some good insight into the AI world and issues we are facing but for some reason Donald Trump, "hate speech", "inclusive language" and other wokeisms are present in every single chapter.
Why? It's supposed to be about AI, not Queer studies. Stick to the topic.
The only reason I finished the book is years ago I promised myself I'll finish every book I start.
Well, this one gave me a run for my money.
Why? It's supposed to be about AI, not Queer studies. Stick to the topic.
The only reason I finished the book is years ago I promised myself I'll finish every book I start.
Well, this one gave me a run for my money.
December 4, 2024
This is an excellent primer on AI, a subject that is constantly in the headlines but often without balance or context. The author does a good job of explaining what it is, strengths and limitations, promise and threats. Overall I'd say it helped dissipate some of my worry about AI, without glossing over genuine concerns that need to be addressed as this technology progresses.
January 21, 2024
Started off really strong and readable but lost me a bit towards the end.
September 17, 2024
Persuasive arguments, but as it openly admits, by the time of publication, it's already out of date.
December 14, 2024
Insightful and somewhat scary. Almost a must read as AI embeds itself in our daily lives
January 6, 2025
Doesn't give much deep depth.
January 10, 2025
An interesting examination of the history of AI and its present incarnation.
Displaying 1 - 17 of 17 reviews
Can't find what you're looking for?
Get help and learn more about the design.