What do you think?
Rate this book
Why Greatness Cannot Be Planned: The Myth of the Objective
Why does modern life revolve around objectives? From how science is funded, to improving how children are educated -- and nearly everything in-between -- our society has become obsessed with a seductive that greatness results from doggedly measuring improvement in the relentless pursuit of an ambitious goal. In Why Greatness Cannot Be Planned, Stanley and Lehman begin with a surprising scientific discovery in artificial intelligence that leads ultimately to the conclusion that the objective obsession has gone too far. They make the case that great achievement can't be bottled up into mechanical metrics; that innovation is not driven by narrowly focused heroic effort; and that we would be wiser (and the outcomes better) if instead we whole-heartedly embraced serendipitous discovery and playful creativity.
Controversial at its heart, yet refreshingly provocative, this book challenges readers to consider life without a destination and discovery without a compass.
Controversial at its heart, yet refreshingly provocative, this book challenges readers to consider life without a destination and discovery without a compass.
154 pages, Kindle Edition
First published May 5, 2015
Ratings & Reviews
Friends & Following
Create a free account to discover what your friends think of this book!
Community Reviews
Displaying 1 - 30 of 81 reviews
December 7, 2019
At the AI conference I attended in Prague last summer, this author's talk stood out as the most interesting to me. Although the author is an AI researcher, this book is written for the lay reader. His point is a very simple one: if you always try to move towards an end goal, so much of the space of possibilities will go unexplored that the best solutions won't be found, except in the most straightforward of cases. Instead of heading toward the objective we should explore the space of possibilities by following novelty or interestingness wherever it leads us, collecting treasures along the way. After exploring the simplest solutions, the only way to go that is novel is towards the more complex, so this kind of exploration moves in the direction of increasing complexity.
He applies the idea to scientific research, to evolution, to art, and to education, and brings insight to how each of these fields could be reformed to be more creative and in doing so, paradoxically progress faster.
Evolution, for example: "survival of the fittest" implies there is one form which is the fittest form, and evolution is moving ever towards that goal. But if success at reproduction is all that matters to evolution (speaking anthropomorphically) it has never done better than bacteria. Clearly something else is going on here.
business: this is the difference between innovative new products and commoditization.
research: everyone knows that just getting 2% better on the metrics isn't the best way to decide which papers to publish, but we keep going back to it because if it does better it must be publishable.
The book has changed one of my long held beliefs, that something like Common Core and lots of testing are needed if education is to improve. If what Stanley is saying is right, this will only lead to small gains that then plateau. Instead, what education needs is diversity and freedom, and lots of cross-pollenization.
When I think about how to implement his ideas for artificial creativity, though, I keep running into the question of how to make decisions without an objective. It's all well and good to say "try everything!" but most things you try are bad ideas that won't lead anywhere. How can you build a fully autonomous system that can recognize "this has potential" without defining an objective?
He gives a few examples of how his systems actually outperform traditional search methods that move towards an objective. But when I followed up on researchers who cite his work, the picture gets muddier: it turns out that hybrids (which include an objective as well as exploration) often perform better overall. And in looking for what performs best, aren't we abandoning the principle behind all this anyway?
I've still got to put a lot of effort into considering his ideas, but they are intriguing. I also feel like they validate my own approach towards my work. My only complaint was that parts got repetitive, but I just skimmed those.
EDIT: One year later, these hybrid methods have a lot better empirical results showing that on certain hard problems, looking for a diverse set of good solutions does a better job at finding the best solution than just trying to find the best solution. Some high-profile game-playing programs like AlphaStar and Go-Explore made use of this technique. The term people are using is Quality Diversity (QD).
He applies the idea to scientific research, to evolution, to art, and to education, and brings insight to how each of these fields could be reformed to be more creative and in doing so, paradoxically progress faster.
Evolution, for example: "survival of the fittest" implies there is one form which is the fittest form, and evolution is moving ever towards that goal. But if success at reproduction is all that matters to evolution (speaking anthropomorphically) it has never done better than bacteria. Clearly something else is going on here.
business: this is the difference between innovative new products and commoditization.
research: everyone knows that just getting 2% better on the metrics isn't the best way to decide which papers to publish, but we keep going back to it because if it does better it must be publishable.
The book has changed one of my long held beliefs, that something like Common Core and lots of testing are needed if education is to improve. If what Stanley is saying is right, this will only lead to small gains that then plateau. Instead, what education needs is diversity and freedom, and lots of cross-pollenization.
When I think about how to implement his ideas for artificial creativity, though, I keep running into the question of how to make decisions without an objective. It's all well and good to say "try everything!" but most things you try are bad ideas that won't lead anywhere. How can you build a fully autonomous system that can recognize "this has potential" without defining an objective?
He gives a few examples of how his systems actually outperform traditional search methods that move towards an objective. But when I followed up on researchers who cite his work, the picture gets muddier: it turns out that hybrids (which include an objective as well as exploration) often perform better overall. And in looking for what performs best, aren't we abandoning the principle behind all this anyway?
I've still got to put a lot of effort into considering his ideas, but they are intriguing. I also feel like they validate my own approach towards my work. My only complaint was that parts got repetitive, but I just skimmed those.
EDIT: One year later, these hybrid methods have a lot better empirical results showing that on certain hard problems, looking for a diverse set of good solutions does a better job at finding the best solution than just trying to find the best solution. Some high-profile game-playing programs like AlphaStar and Go-Explore made use of this technique. The term people are using is Quality Diversity (QD).
February 17, 2021
when the business + eastern philosophy book sections get disintermediated by nerdy AI researchers
"When all is said and done, when even visionaries grow weary of stale visions, when the ash of unrequited expectation settles on the cloak of the impenetrable future, there is but one principle that may yet pierce the darkness: To achieve our highest goals, we must be willing to abandon them."
"When all is said and done, when even visionaries grow weary of stale visions, when the ash of unrequited expectation settles on the cloak of the impenetrable future, there is but one principle that may yet pierce the darkness: To achieve our highest goals, we must be willing to abandon them."
August 7, 2022
The authors of this book performed an interesting experiment in 2006 where they created a web-based application called Picbreeder, which is still available for anyone to try. The program generates an array of small images with very little details, and each very similar to the rest except for some small random variations. The human user is supposed to pick one of these images to trigger the next generation of images, which are small variations of the selected image. The same user, or someone else, can continue this selective "breeding" as long as desired. The strange finding was that ultimately it ends up generating surprising images that resemble recognizable objects such as a human face, a car, an insect, or a candle. The important observation here is that the users did not aim to create these images from the start, and it can be shown that if they aimed at something, most likely they would never get there. That is, the most surprising results are obtained when that was not the objective, and it reaches this state via a series of unrelated interesting images, which the authors call the "stepping stones".
In this book, the authors try to generalize this observation into a much larger principle. They claim that while small improvements can be achieved through a goal-directed approach, really ambitious things can only be achieved when we allow ourselves to wander around without specific objectives. For example, the first computer, Eniac, used vacuum tubes, but if someone wanted to create the first computer, say a hundred years ago, they would not have discovered the vacuum tube in the process. The discovery of the vacuum tube happened for a very different reason and acted as a stepping stone toward designing the first computer.
Their main attack is on the present culture of objective-driven processes in the world of science and technology. Starting from how research is funded and evaluated to how we are trained to think, the focus is always on specific objectives, and according to the authors, that is a very short-sighted approach and is a hindrance to spectacular discoveries.
They also try to view the process of natural evolution through this prism and present an alternative interpretation of the evolutionary process. The group also tried to apply these principles in an AI search algorithm that only looks for "novelty" in the search space, rather than maximizing some objective function, and show that in some cases, this approach produces better results than an objective-based algorithm.
It is certainly a powerful and novel idea that offers a new way of looking at many things. I certainly felt intellectually enriched after reading the book. However, I also felt that the writers fell into a common trap many others fall into after they come up with a powerful idea -- that of over-generalizing the scope of the idea. They start believing that this one idea can be applied everywhere. In this case, the authors do talk about this mistake but still fall into it.
First of all, they never discuss real-life cases where their ideas may not apply. For example, they talk of extremely successful individuals who did not aim at what ultimately made them famous but never mention other successful people who set their ultimate target very early on in their lives and doggedly pursued it till they reached their celebrity. They criticize the current process of objective-based policies and explain why greatness cannot be achieved that way but do not try to look at thousands of spectacular innovations that happened in spite of these restrictions.
Finally, they fail to offer a pragmatic alternative to objective-based thinking. In a world where resources are always limited, how else can we allocate those precious resources without some sort of practical constraints? In reality, most of the scientific and technological improvements that happen are due to unambitious small steps done by mediocre people. The system asks us to predict an outcome and then go for it. However, in this sea of mediocrity, there are a few brilliant individuals who can see a little further or are just more creative, and they discover something that is away from the original objective. It is these leaps that form what the authors are calling the stepping stones. That is, spectacular progress still happens, despite our uninspiring processes, and the objective-based paradigms ensure that the majority of mediocre innovators make the best use of the resources.
I am not suggesting that the process that exists today is ideal or optimum. Thinking like this can certainly drive us towards a better system. However, I don't believe the removal of the objective-based process altogether can be very efficient.
A great, thought-provoking book that I'd highly recommend. However, I think the book could have been much shorter without losing much of its communicative strength.
In this book, the authors try to generalize this observation into a much larger principle. They claim that while small improvements can be achieved through a goal-directed approach, really ambitious things can only be achieved when we allow ourselves to wander around without specific objectives. For example, the first computer, Eniac, used vacuum tubes, but if someone wanted to create the first computer, say a hundred years ago, they would not have discovered the vacuum tube in the process. The discovery of the vacuum tube happened for a very different reason and acted as a stepping stone toward designing the first computer.
Their main attack is on the present culture of objective-driven processes in the world of science and technology. Starting from how research is funded and evaluated to how we are trained to think, the focus is always on specific objectives, and according to the authors, that is a very short-sighted approach and is a hindrance to spectacular discoveries.
They also try to view the process of natural evolution through this prism and present an alternative interpretation of the evolutionary process. The group also tried to apply these principles in an AI search algorithm that only looks for "novelty" in the search space, rather than maximizing some objective function, and show that in some cases, this approach produces better results than an objective-based algorithm.
It is certainly a powerful and novel idea that offers a new way of looking at many things. I certainly felt intellectually enriched after reading the book. However, I also felt that the writers fell into a common trap many others fall into after they come up with a powerful idea -- that of over-generalizing the scope of the idea. They start believing that this one idea can be applied everywhere. In this case, the authors do talk about this mistake but still fall into it.
First of all, they never discuss real-life cases where their ideas may not apply. For example, they talk of extremely successful individuals who did not aim at what ultimately made them famous but never mention other successful people who set their ultimate target very early on in their lives and doggedly pursued it till they reached their celebrity. They criticize the current process of objective-based policies and explain why greatness cannot be achieved that way but do not try to look at thousands of spectacular innovations that happened in spite of these restrictions.
Finally, they fail to offer a pragmatic alternative to objective-based thinking. In a world where resources are always limited, how else can we allocate those precious resources without some sort of practical constraints? In reality, most of the scientific and technological improvements that happen are due to unambitious small steps done by mediocre people. The system asks us to predict an outcome and then go for it. However, in this sea of mediocrity, there are a few brilliant individuals who can see a little further or are just more creative, and they discover something that is away from the original objective. It is these leaps that form what the authors are calling the stepping stones. That is, spectacular progress still happens, despite our uninspiring processes, and the objective-based paradigms ensure that the majority of mediocre innovators make the best use of the resources.
I am not suggesting that the process that exists today is ideal or optimum. Thinking like this can certainly drive us towards a better system. However, I don't believe the removal of the objective-based process altogether can be very efficient.
A great, thought-provoking book that I'd highly recommend. However, I think the book could have been much shorter without losing much of its communicative strength.
January 16, 2018
Conflates objectives (goals) with objective functions (measurement of progress). The book is filled with good examples of bad objective functions, but then makes wild claims about the usefulness of objectives. In the end this whole book can be viewed as another good example of programmers overabstracting things to the point of absurdity.
August 17, 2015
It is interesting to see discussion about the value of exploration, instead of focused march towards ambitious goals based on clear measures of “progress” coming from western researchers. It's quite interesting that this comes as insight from artificial intelligence research, complete with examples how learning algorithms based on exploration rather than set goals and metrics can achieve various results, like learning to navigate a maze, much faster than traditional approaches. I also greatly enjoyed reframing things like evolution as non-objective search with minimum criterion (survivability) instead of traditional talk about optimising fitness. It's quite a fascinating intellectual exercise.
It is worth noting that the book does not attempt to tell you that there is no value in objective-driven approach. The authors make a point of emphasising that the problem is with “ambitious” goals, which in this case usually means those without clear path towards them, i.e., those we do not really know how to achieve.
Having said all that, the premise, repeated over and over again throughout substantial part of the book, reads a lot like “beware of local maxima!” The thing is that the way some people would have us devise, say, cure to cancer, or improve educational situation in US, is by setting a clear numerical goal, and then measuring progress. The idea is that we will be going steadily towards the desired goal, and as long as the metric goes closer towards target values, we're all good. That's essentially gradient-based optimisation, and it is indeed prone to getting stuck in local extrema, should they exist.
Bringing that idea to the attention of potential decision makers is a worthy goal, as they might not have taken courses in optimisation methods, but I feel that authors insist too much on how in case of ambitious goals we *will* necessarily get stuck, because the breakthroughs required to achieve them will not look anything like the final goal. The necessity of such state of the affairs was not motivated sufficiently. I agree that it was shown that this may be the case: we may not know how to get to the final goal, and required “stepping stones” may be unexpected with respect to our current state of knowledge. It's a possibility that should be taken into account, and as authors show on various examples, currently rarely is. But I did not end up convinced that it is necessary to get stuck in local extremum, as authors would have you believe.
That's the only weak point of the book in my opinion, and it's still well worth a read.
It is worth noting that the book does not attempt to tell you that there is no value in objective-driven approach. The authors make a point of emphasising that the problem is with “ambitious” goals, which in this case usually means those without clear path towards them, i.e., those we do not really know how to achieve.
Having said all that, the premise, repeated over and over again throughout substantial part of the book, reads a lot like “beware of local maxima!” The thing is that the way some people would have us devise, say, cure to cancer, or improve educational situation in US, is by setting a clear numerical goal, and then measuring progress. The idea is that we will be going steadily towards the desired goal, and as long as the metric goes closer towards target values, we're all good. That's essentially gradient-based optimisation, and it is indeed prone to getting stuck in local extrema, should they exist.
Bringing that idea to the attention of potential decision makers is a worthy goal, as they might not have taken courses in optimisation methods, but I feel that authors insist too much on how in case of ambitious goals we *will* necessarily get stuck, because the breakthroughs required to achieve them will not look anything like the final goal. The necessity of such state of the affairs was not motivated sufficiently. I agree that it was shown that this may be the case: we may not know how to get to the final goal, and required “stepping stones” may be unexpected with respect to our current state of knowledge. It's a possibility that should be taken into account, and as authors show on various examples, currently rarely is. But I did not end up convinced that it is necessary to get stuck in local extremum, as authors would have you believe.
That's the only weak point of the book in my opinion, and it's still well worth a read.
April 3, 2022
Saw this book as a recommendation on Twitter by Patrick O’Shaughnessey. Was very inclined to agree with the book’s worldview, but would’ve preferred this is as 3-part blog post. The authors are AI researchers who used algorithmic search techniques to show how beginning with an end-state in mind can often lead to worse results than simply seeking novelty.
Main idea: when doing ambitious things, setting explicit objectives (end goals) can be counter-productive. Instead, seeking novelty at each step and exploring without an end-goal in mind can be a better approach.
“Think of the process of creation as a process of searching through the space of the room. As you can imagine, the kind of image you are most likely to paint depends on what parts of the room you’ve already visited .. the more you’ve explored the room yourself, the more you understand where you might be able to go next .. the places we have visited are stepping stones to new ideas”
The authors posit that simply following one’s curiosity and following a trail of stepping stones can lead to better outcomes than deciding what the end goal should be and the figuring out how to get there.
They also posit that “we don’t face a false choice between slavishly following objectives and aimless wandering”. Novelty seeking can be quantified, by observing how much what we’re doing is different from that in the past, and following a gradient of novelty instead of the gradient of “how close are we from our given objective”. Setting constraints on behaviour and following a novelty-seeking approach leads to a higher probability of achieving spectacular outcomes, in their opinion.
I would’ve loved for the book to be more rigorous, and to address the shortcomes of this approach better. Definitely has some interesting food for thought, though!
Main idea: when doing ambitious things, setting explicit objectives (end goals) can be counter-productive. Instead, seeking novelty at each step and exploring without an end-goal in mind can be a better approach.
“Think of the process of creation as a process of searching through the space of the room. As you can imagine, the kind of image you are most likely to paint depends on what parts of the room you’ve already visited .. the more you’ve explored the room yourself, the more you understand where you might be able to go next .. the places we have visited are stepping stones to new ideas”
The authors posit that simply following one’s curiosity and following a trail of stepping stones can lead to better outcomes than deciding what the end goal should be and the figuring out how to get there.
They also posit that “we don’t face a false choice between slavishly following objectives and aimless wandering”. Novelty seeking can be quantified, by observing how much what we’re doing is different from that in the past, and following a gradient of novelty instead of the gradient of “how close are we from our given objective”. Setting constraints on behaviour and following a novelty-seeking approach leads to a higher probability of achieving spectacular outcomes, in their opinion.
I would’ve loved for the book to be more rigorous, and to address the shortcomes of this approach better. Definitely has some interesting food for thought, though!
October 2, 2023
I've been aware of this book since its inception, but I often ponder why I didn't dive into it sooner. At first glance, it appears unassuming, a slim volume that doesn't seem to promise a substantial challenge. But how wrong would one be to underestimate its force field! This book not only lives up to the greatness of its title but stands as one of the rare works capable of giving voice to the myriad private thoughts I've collected over the years.
To be perfectly candid, I've never scribbled as fervently in the margins of a book as I have with this one — Mortimer Adler himself might commend my dedication to syntopic reading. Nearly every paragraph catapulted me into the stratosphere, providing a sweeping perspective on our humanity, only to pull me forcefully back to Earth, and then, in a stunning twist, throw me to the edge of the sun or solar system to grasp the larger picture in which the insignificance of our modern explorations lies.
I've yet to fully grasp the far-reaching impact this book will have on me, but I am reasonably confident that I've grasped its essence. Perhaps this certainty stems from my deep engagement with the landscape of artificial intelligence, my extensive readings on optimization, and my understanding of the challenges inherent in navigating the vast search space. Nevertheless, the ideas presented in this book have found a comfortable home within me.
Since delving into its pages, I find myself continually experiencing the Baader-Meinhof phenomenon. And even though I possess a solid grounding in probability theory, the qualia of this experience leaves a taste that can only be described as a superposition of both sour and sweet.
To be perfectly candid, I've never scribbled as fervently in the margins of a book as I have with this one — Mortimer Adler himself might commend my dedication to syntopic reading. Nearly every paragraph catapulted me into the stratosphere, providing a sweeping perspective on our humanity, only to pull me forcefully back to Earth, and then, in a stunning twist, throw me to the edge of the sun or solar system to grasp the larger picture in which the insignificance of our modern explorations lies.
I've yet to fully grasp the far-reaching impact this book will have on me, but I am reasonably confident that I've grasped its essence. Perhaps this certainty stems from my deep engagement with the landscape of artificial intelligence, my extensive readings on optimization, and my understanding of the challenges inherent in navigating the vast search space. Nevertheless, the ideas presented in this book have found a comfortable home within me.
Since delving into its pages, I find myself continually experiencing the Baader-Meinhof phenomenon. And even though I possess a solid grounding in probability theory, the qualia of this experience leaves a taste that can only be described as a superposition of both sour and sweet.
August 29, 2021
More like 3.5
Very verbose but fairly accessible to general audience. Made me rethink/question my goals and career plans.
Very verbose but fairly accessible to general audience. Made me rethink/question my goals and career plans.
December 18, 2020
This is one of the most thought-provoking books I have read. And I could relate even better because I work in the area of AI like the authors and many of the examples and the final case-study is from this area. Also it's inspiring to see such deep insight coming from the tool they built called Picbreeder.
The core idea presented in the book is that "objectives" are unnecessary and rather a hinderance while solving "complex" problems. Objectives might be useful for simple problems where it is easy to chart out a path to the goal. However in case of more complex problems where the route to the goal is not known, objectives cause more harm than good.
Next, an alternate model is presented called the novelty search. It finds solutions based on how "interesting" the solution is rather than how close it is to an "objective". The authors connect this idea to the stepping stones, which is a great analogy. They argue that a complex goal lies somewhere in a hazy lake with a very low visibility. It is difficult or impossible to reach a specific goal in this lake and an objective in this case is like a "broken compass". However, it is much more relevant to explore the lake by finding new stepping stones based on novelty or interestingness and how many new stepping stones it can lead to. True success lies in exploring the space of the lake rather than trying to reach a specific imaginary point.
Finally, a variety of use-cases are discussed from education, innovation to evolution and AI where a novelty based approach makes more sense rather than an objective based approach.
This book presents an intriguing idea which can literally change the way we operate in life. Sometimes it gets a bit repetitive but overall it drives the point. I highly recommend this book!
The core idea presented in the book is that "objectives" are unnecessary and rather a hinderance while solving "complex" problems. Objectives might be useful for simple problems where it is easy to chart out a path to the goal. However in case of more complex problems where the route to the goal is not known, objectives cause more harm than good.
Next, an alternate model is presented called the novelty search. It finds solutions based on how "interesting" the solution is rather than how close it is to an "objective". The authors connect this idea to the stepping stones, which is a great analogy. They argue that a complex goal lies somewhere in a hazy lake with a very low visibility. It is difficult or impossible to reach a specific goal in this lake and an objective in this case is like a "broken compass". However, it is much more relevant to explore the lake by finding new stepping stones based on novelty or interestingness and how many new stepping stones it can lead to. True success lies in exploring the space of the lake rather than trying to reach a specific imaginary point.
Finally, a variety of use-cases are discussed from education, innovation to evolution and AI where a novelty based approach makes more sense rather than an objective based approach.
This book presents an intriguing idea which can literally change the way we operate in life. Sometimes it gets a bit repetitive but overall it drives the point. I highly recommend this book!
August 31, 2021
One of the few books that had truly affected how I see the world.
The author took his findings from the AI research (which I follow, which is how I learned about the book in the first place) and applied it to everyday life. So, on the surface this book looks like yet another self-help book, but the ideas in it originate from the AI research. I'm not sure if it's valid to make this transfer between the fields, but I really liked it in this case.
The book's idea is that it makes no sense to track progress towards any sufficiently ambitious goal. If you want to become a billionaire, it makes no sense to track and to maximize your salary. If you want to land on the moon, you won't make progress by climbing higher and higher mountains, even though it does get you a little closer to the moon.
The important corollary is that it makes no sense to plan for anything truly ambitious. If you can't measure if your actions bring you closer to the goal, then why bother planning? And it also means that the really ambitious goal you have in mind will probably cannot be achieved.
This is terrifying and is in contrast to the belief that anyone can achieve anything, provided enough grit. But at the same time it's liberating, because once you stop pursuing your single goal, you'll start seeing many other possibilities around you. Collect useful ideas -- become a "stepping stones" collector -- and see where they lead you, without having a specific objective in mind. This could make life much more interesting and rewarding.
I didn't agree with the author everywhere. But I do agree with the general idea, and this book definitely generates many opportunities to think deeply. My personal book of the year so far :)
The author took his findings from the AI research (which I follow, which is how I learned about the book in the first place) and applied it to everyday life. So, on the surface this book looks like yet another self-help book, but the ideas in it originate from the AI research. I'm not sure if it's valid to make this transfer between the fields, but I really liked it in this case.
The book's idea is that it makes no sense to track progress towards any sufficiently ambitious goal. If you want to become a billionaire, it makes no sense to track and to maximize your salary. If you want to land on the moon, you won't make progress by climbing higher and higher mountains, even though it does get you a little closer to the moon.
The important corollary is that it makes no sense to plan for anything truly ambitious. If you can't measure if your actions bring you closer to the goal, then why bother planning? And it also means that the really ambitious goal you have in mind will probably cannot be achieved.
This is terrifying and is in contrast to the belief that anyone can achieve anything, provided enough grit. But at the same time it's liberating, because once you stop pursuing your single goal, you'll start seeing many other possibilities around you. Collect useful ideas -- become a "stepping stones" collector -- and see where they lead you, without having a specific objective in mind. This could make life much more interesting and rewarding.
I didn't agree with the author everywhere. But I do agree with the general idea, and this book definitely generates many opportunities to think deeply. My personal book of the year so far :)
August 2, 2021
Отличная книга про то, что амбициозные цели не работают. Авторы немного заезженно повторяют одни и те же примеры в качестве аргументов к своей точки зрения, но раскрытия текущих работающих систем в мире (образование, научная деятельность), действительно была пересмотрена под новым углом.
Список основных пунктов из книги:
- Цели работают в случае, если они скромные и обозримые. Не больше пары лет вперёд.
- Фундамент важен для обнаружения возможностей. Если мы отправим современного гения на пару сотен лет назад, то изобрести компьютер он не сможет.
- Амбициозные цели часто достигались в разрез плана.
- Инновации не идут намеченным путём, а скорее методом проб и ошибок.
- Ответом на замену целей является чуйка. В связи с чем отход от плана иногда может принести большую награду. Правда эта награда, не та что мы ожидаем.
- Stepping stones, not milestones.
Список основных пунктов из книги:
- Цели работают в случае, если они скромные и обозримые. Не больше пары лет вперёд.
- Фундамент важен для обнаружения возможностей. Если мы отправим современного гения на пару сотен лет назад, то изобрести компьютер он не сможет.
- Амбициозные цели часто достигались в разрез плана.
- Инновации не идут намеченным путём, а скорее методом проб и ошибок.
- Ответом на замену целей является чуйка. В связи с чем отход от плана иногда может принести большую награду. Правда эта награда, не та что мы ожидаем.
- Stepping stones, not milestones.
June 27, 2019
5 Stars. This is a must read if you are engaged in research and development--regardless of your field or specialty. The authors provide deep insight into some of the misconceptions and errors that have become accepted as fact when it comes to seeking ambitious objectives. The authors introduce the concept of search as a process of discovery. This leads to a related idea: creativity as a kind of search. They use the concept of stepping stones that lead to new developments and how the challenge is to discover the proper stepping stones within the search space. The implications of the authors' arguments are far reaching and persuasive. I give the book 5 stars because it has altered my perception of how to pursue ambitious objectives.
October 5, 2022
Goal setting, breaking it down to the steps, and methodical execution is the best way to achieve an objective that is within sight. But how do you shoot for the stars? How do you come up with extraordinary progress? You can't just plan and will your way to greatness. The only thing you could do is to be open-minded and curious, follow your guts and dig deep into what feels interesting. This is the best way to discover something truly new. It might be not at all what you were looking for, but it could be something great nonetheless.
This is the tl;dr of the book, and I would take a star off because the book could have been quite a bit shorter - it's the same argument over and over again. I also think authors did not do good enough job explaining why objectives and divide and conquer are so great.
With that caveat aside, I think the main idea of this book is profound: you can't plan greatness. The reason is rather simple: if you could see the next big thing (invention, idea, whatever) from where we as humanity stand - then the job is easy - set the objective and execute. But the thing is that of course we don't see all the possible big next things! We can only see so far. By focusing on objectives exclusively, we deprive ourselves of the opportunity to explore the unexplored. And time and time again completely accidental discoveries in one place lead to unprecedented improvements in others.
Authors demonstrate this effect more formally with the aid of a computer simulation, which is quite convincing, and is a great model for the problem.
I also really liked their application of this idea to education: standardized testing does not improve the learning; the accuracy of the tests does not improve the education; setting objectives to improve the education does not improve it. Most people understand this intuitively, yet society meeps moving towards more and more standardized tests, which is really counter-productive. Tests have their application of course, but it should not be the primary method, and it should not be the goal in and of itself.
The book does not really talk about this, but I see a direct confirmation of the central idea in all the major computer science discoveries of the last 80 years. All the great innovative things came from places that invested into basic research - without thinking about the future applications too much - scientists and engineers were exploring things that looked interesting, not what would be "good for business". A lot of big companies today lose the sight of this fact. If you really want to create something radically new, you have to let people explore on their own, without a particular plan.
Overall, highly recommend the book to anyone, especially you work as an engineer, scientist, or an educator. It's a bit tedious at times, but the idea and the framework is really good.
This is the tl;dr of the book, and I would take a star off because the book could have been quite a bit shorter - it's the same argument over and over again. I also think authors did not do good enough job explaining why objectives and divide and conquer are so great.
With that caveat aside, I think the main idea of this book is profound: you can't plan greatness. The reason is rather simple: if you could see the next big thing (invention, idea, whatever) from where we as humanity stand - then the job is easy - set the objective and execute. But the thing is that of course we don't see all the possible big next things! We can only see so far. By focusing on objectives exclusively, we deprive ourselves of the opportunity to explore the unexplored. And time and time again completely accidental discoveries in one place lead to unprecedented improvements in others.
Authors demonstrate this effect more formally with the aid of a computer simulation, which is quite convincing, and is a great model for the problem.
I also really liked their application of this idea to education: standardized testing does not improve the learning; the accuracy of the tests does not improve the education; setting objectives to improve the education does not improve it. Most people understand this intuitively, yet society meeps moving towards more and more standardized tests, which is really counter-productive. Tests have their application of course, but it should not be the primary method, and it should not be the goal in and of itself.
The book does not really talk about this, but I see a direct confirmation of the central idea in all the major computer science discoveries of the last 80 years. All the great innovative things came from places that invested into basic research - without thinking about the future applications too much - scientists and engineers were exploring things that looked interesting, not what would be "good for business". A lot of big companies today lose the sight of this fact. If you really want to create something radically new, you have to let people explore on their own, without a particular plan.
Overall, highly recommend the book to anyone, especially you work as an engineer, scientist, or an educator. It's a bit tedious at times, but the idea and the framework is really good.
March 29, 2021
Stanley and Lehman explore the myth of the objective. When we focus on ambitious objectives—concrete, specific goals—we are setting ourselves up for failure. Objectives can be useful when we're looking at changes that are adjacent to where we already are. However, when our goal is major innovation, objectives actively hinder progress. The path between where we are today and the objective is often indirect and may require moving away from a seemingly direct path.
The authors refer to these intermediate discoveries as stepping stones. The myth of the objective assumes that the best way to pursue an objective is to follow the stepping stones that are nearest to the ultimate goal: the objective becomes the objective function. But this is like trying to go through a maze by always taking the path that points toward the exit. The true path is much more roundabout. The objective is a false compass.
This might not be so bad if we had good visibility of the possible paths between here and there. However, to continue with the stepping stone metaphor, these stepping stones are in a foggy landscape. The objective is tempting because we cannot see the full set of paths available to us. We want some way to navigate through the fog. However, as the authors emphasize, following a false compass is no better than not having one. Although I find their claim about objectives as a false compass compelling in the end, I did find the argument itself to be fairly weak. It was mostly argued from example and focused rather too much on the exact paths by which innovations were reached: sure, our path to computers involved vacuum tubes, but that doesn't mean that vacuum tubes were the only path to computers (especially given that we didn't stay at vacuum tubes).
If objectives provide a false compass, what are we to do? Do we just wander around randomly and hope for good luck? No. We can be more intentional than that. The authors encourage us to utilize non-objective search. Instead of measuring progress against a destination, exploration is driven by some measurement relative to the present.
This measurement has two parts. First, there are constraints or backstops. Some paths should be avoided because we cannot actually make our way through them. Constraints include physical laws (e.g., the speed of light), continued ability to participate (an organism that dies too quickly can't reproduce), and domain specific requirements (medicine should not do harm). Constraints tell you where not to go. Second, there should be something points us in a direction. The authors are particularly fond of novelty search, which follows the direction that is most interesting (for some domain relevant definition of interesting). The important part is that both constraints and guiding principles are relative to the present, not to some unrealized future state. From here, don't go a direction that will kill you. From here, go the direction which looks the most interesting.
This general idea is not completely incompatible with objectives. The authors argue convincingly that we shouldn't use objectives as objective functions, as the measure of which direction is best to go. That will cause us to miss necessary detours. However, can't we use objectives as part of a more nuanced guiding methodology? The authors encourage us to completely abandon objectives. Instead of trying to have any control over the direction we go, if we want to be innovative, we should become treasure hunters. We should always follow the path that is the most interesting. That path may lead us places that are incredibly innovative, but we won't know where we are going in advance.
I didn't find this part of the argument completely compelling.Overall, I still feel like objectives can be useful. I agree that treasure hunting is one way to innovate. It may be the only way to achieve truly groundbreaking innovations. However, there is a spectrum between objective-friendly adjacent inventions and major innovation. Even if we cannot achieve an objective in full, the interesting things we discovery on the way will likely be more aligned with the goal of the objective than if we just wander about following what is interesting or novel—even if we can't solve world hunger, trying is likely to at least reduce hunger. The problems come when we conflate objectives—places we want to go—with objective functions—measures of how close we are to being there.
However, I do agree that objectives make terrible fitness functions. If we measure progress only by measuring distance from the ideal, then we'll likely get stuck at a boring dead end. I agree with the fundamental assertion that when we are determining which next step to take, we should use an objective function that measure progress relative to where we are now, not relative to some particular place we want to be.
Maybe instead of objectives we can think of north stars. A north star is something which guides your direction. It is not a destination you can ever reach. A north star reduces the number of next steps to consider. Since we never expect to reach a north star, we don't measure success against the north star. Rather, we measure success by what we achieved while going that direction. Let's replace the objective, in the negative sense used in this book, with the north star. Let's allow ourselves to wander, and also allow ourselves a general direction to that wandering. Perhaps Stanley and Lehman might say that makes a north star more like a constraint, something which defines which directions not to go. However, I think that the positive framing of a north star is more inspiring than a constraint.
Despite the fact that this book was not as well argued as I would like, I believe that it was worth the read because it caused me to think more deeply about objectives rather than just blindly accepting them as the way things should be done.
The authors refer to these intermediate discoveries as stepping stones. The myth of the objective assumes that the best way to pursue an objective is to follow the stepping stones that are nearest to the ultimate goal: the objective becomes the objective function. But this is like trying to go through a maze by always taking the path that points toward the exit. The true path is much more roundabout. The objective is a false compass.
This might not be so bad if we had good visibility of the possible paths between here and there. However, to continue with the stepping stone metaphor, these stepping stones are in a foggy landscape. The objective is tempting because we cannot see the full set of paths available to us. We want some way to navigate through the fog. However, as the authors emphasize, following a false compass is no better than not having one. Although I find their claim about objectives as a false compass compelling in the end, I did find the argument itself to be fairly weak. It was mostly argued from example and focused rather too much on the exact paths by which innovations were reached: sure, our path to computers involved vacuum tubes, but that doesn't mean that vacuum tubes were the only path to computers (especially given that we didn't stay at vacuum tubes).
If objectives provide a false compass, what are we to do? Do we just wander around randomly and hope for good luck? No. We can be more intentional than that. The authors encourage us to utilize non-objective search. Instead of measuring progress against a destination, exploration is driven by some measurement relative to the present.
This measurement has two parts. First, there are constraints or backstops. Some paths should be avoided because we cannot actually make our way through them. Constraints include physical laws (e.g., the speed of light), continued ability to participate (an organism that dies too quickly can't reproduce), and domain specific requirements (medicine should not do harm). Constraints tell you where not to go. Second, there should be something points us in a direction. The authors are particularly fond of novelty search, which follows the direction that is most interesting (for some domain relevant definition of interesting). The important part is that both constraints and guiding principles are relative to the present, not to some unrealized future state. From here, don't go a direction that will kill you. From here, go the direction which looks the most interesting.
This general idea is not completely incompatible with objectives. The authors argue convincingly that we shouldn't use objectives as objective functions, as the measure of which direction is best to go. That will cause us to miss necessary detours. However, can't we use objectives as part of a more nuanced guiding methodology? The authors encourage us to completely abandon objectives. Instead of trying to have any control over the direction we go, if we want to be innovative, we should become treasure hunters. We should always follow the path that is the most interesting. That path may lead us places that are incredibly innovative, but we won't know where we are going in advance.
I didn't find this part of the argument completely compelling.Overall, I still feel like objectives can be useful. I agree that treasure hunting is one way to innovate. It may be the only way to achieve truly groundbreaking innovations. However, there is a spectrum between objective-friendly adjacent inventions and major innovation. Even if we cannot achieve an objective in full, the interesting things we discovery on the way will likely be more aligned with the goal of the objective than if we just wander about following what is interesting or novel—even if we can't solve world hunger, trying is likely to at least reduce hunger. The problems come when we conflate objectives—places we want to go—with objective functions—measures of how close we are to being there.
However, I do agree that objectives make terrible fitness functions. If we measure progress only by measuring distance from the ideal, then we'll likely get stuck at a boring dead end. I agree with the fundamental assertion that when we are determining which next step to take, we should use an objective function that measure progress relative to where we are now, not relative to some particular place we want to be.
Maybe instead of objectives we can think of north stars. A north star is something which guides your direction. It is not a destination you can ever reach. A north star reduces the number of next steps to consider. Since we never expect to reach a north star, we don't measure success against the north star. Rather, we measure success by what we achieved while going that direction. Let's replace the objective, in the negative sense used in this book, with the north star. Let's allow ourselves to wander, and also allow ourselves a general direction to that wandering. Perhaps Stanley and Lehman might say that makes a north star more like a constraint, something which defines which directions not to go. However, I think that the positive framing of a north star is more inspiring than a constraint.
Despite the fact that this book was not as well argued as I would like, I believe that it was worth the read because it caused me to think more deeply about objectives rather than just blindly accepting them as the way things should be done.
December 28, 2018
This book will challenge - to your core - any ideas you have about what it takes to achieve something truly great and innovative. Instead of optimizing, planning and continuously improving in a specific dimension (or set of dimensions) the book asserts that one must instead seek stepping stones based on interestingness - while ignoring a far off objective - if you have any hope of getting somewhere remarkable. Building on AI research about how algorithms perform when seeking a goal (vs when simply exploring), I'm surprised to just hear about this book. Parts are repetitive, but the "big idea" is profound and that alone makes it worth reading.
May 30, 2023
La idea del libro es de las más interesantes que he leído, pero es algo repetitivo y menos técnico de lo que pensaba. Mejor ver una conferencia o escuchar un podcast del tema https://www.youtube.com/watch?v=tZBVi...
November 25, 2023
A great intro to novelty seeking algorithms and the fallacies of objective thinking.
October 27, 2023
I began reading this book with the objective of finishing it to learn what it had to offer but realized that I was actually on a novelty search that would lead to unplanned stepping stones and my activities resembled a non-objective treasure hunt rather than what I set out to achieve in the beginning.
August 10, 2020
Interesting book! Key takeaways: while objectives are good in certain cases, more ambitious goals are less likely to be achieved by defining objectives, as many times their solutions require qualitatively different approaches. They introduce the idea of "Stepping Stones", which are ideas that lead to other ideas - the invention of the steam engine is a stepping stone to trains, the invention of computers was a stepping stone to the internet, Number Theory and Cryptography were stepping stones to Bitcoin, etc. Without the proper stepping stones, many ideas are out of reach, and defining explicit objectives towards these out-of-reach goals is likely to trick us into thinking we've made progress as we work towards them. For example, say your goal is to be a billionaire, and you treat money in your bank account as the objective. You'd believe that working harder in your current job is leading you closer to your goal, when in reality you need to do something qualitatively different to achieve your goal, likely resulting in a temporary drop in salary. So in this case, the objective is "Deceptive".
Another key point is that it's very hard to predict where stepping stones will lead us and which stepping stones are necessary for a particular goal, so instead of guiding our stepping-stone collection process towards a particular far-off goal, it's probably better to just try to find the most interesting stepping stones that are currently one hop away. They argue for the "Treasure hunter" - people not fixed on a particular far-off objective, but instead just searching for interesting stepping stones one hop away. The authors extend these ideas beyond necessarily technical topics and apply this way of thinking towards other aspects of life.
I found the ideas very interesting, and would highly recommend the book.
Another key point is that it's very hard to predict where stepping stones will lead us and which stepping stones are necessary for a particular goal, so instead of guiding our stepping-stone collection process towards a particular far-off goal, it's probably better to just try to find the most interesting stepping stones that are currently one hop away. They argue for the "Treasure hunter" - people not fixed on a particular far-off objective, but instead just searching for interesting stepping stones one hop away. The authors extend these ideas beyond necessarily technical topics and apply this way of thinking towards other aspects of life.
I found the ideas very interesting, and would highly recommend the book.
February 3, 2017
Makes an interesting case saying - at the right point in your creative / creation journey - the pursuit of objectives becomes detrimental instead of helpful. The book advocates pursuit of the journey after base objectives have been achieved (and highly emphasizes the importance of all the foundational startup work required to do anything well) and how switching to learning and searching instead of trying to find will bring you closer to your final creative breakthrough.
Wasn't a huge fan of the repeated emphasis of PicBreeder and may have missed the significance of what they were saying - didn't seem as deep an insight as the authors seemed to think it was.
Wasn't a huge fan of the repeated emphasis of PicBreeder and may have missed the significance of what they were saying - didn't seem as deep an insight as the authors seemed to think it was.
March 17, 2021
The book explores a lot of interesting topics - evolution, AI research, art, etc. The language is quite energetic but needs polishing. Overall, it's an interesting read, but in reality all of the content can be summarized by proverb: "When life gives you lemons, make lemonade". Or, in author's speak "stepping stones" should not be discarded based on (not very clever) defined global objective function but on potential they can lead to.
No practical advice on how to really separate "interesting" vs "trivial" if you don't have 3+ billion years.
No practical advice on how to really separate "interesting" vs "trivial" if you don't have 3+ billion years.
August 29, 2016
The book's core lessons read like a self-help advice on the surface, but in fact they are backed by solid experimental results. Recommended read for anyone working creatively.
August 23, 2022
I read this book at the right time, just when I needed it in my life. That sounds odd because its a mid-way techincal book that perhaps wasn't intended to be self-help. Despite that, as true to the message of this book, it did for me what self-help books probably never could, even though that wasn't its objective.
The message of this book is inspired by AI research and focuses on the myth of the objective, which is ubiquitous and taken for granted so much so that those without them are somehow made to feel like a pariah in society, or simply lost and aimless in life.
This book makes the case that, its not simply that setting an ambitious (novel, not previously done) objective takes you closer to it, even if it doesn't guarantee it when it is planned for, but that it can infact take you further away from that objective (mainly due to deception resulting from our general lack of foresight which cages our thinking to assume that the stepping stones you take to that goal somehow should resemble it). But by embracing serendipity, using clues like a treasure hunter, we can pursue things that are interesting with some chance of stumbling upon something worthwhile, even though exactly what that will be, cannot be said.
The book provides some principles which are great for tracking progress, not only in your personal life, but also for bigger organisations. Also thinking through how someone could implement the ideas for living a life in the way this book paints gives it so much meaning, without the anxiety brought upon by the thought of 'not being where you are supposed to be' in life. It's liberating.
The research covered in this book is interesting, especially if you're familiar with AI and also reinterprets evolution in an interesting way (as a type of novelty searching algorithm).
"So if you're wondering how to escape the myth of the objective, just do things because they're interesting. Not everything needs to be guided by rigid objectives. If you have a strong feeling, go for it. If you don't have a clear objective, then you can't be wrong, because whatever you end up is okay. Assessments only goes so far. A great achievement is one that leads to more great achievements. If you set out to program computers but you're now making movies, you're probably doing something right. If you wanted to create AI but you're now evolving pictures, you're probably doing something right. If you imagined yourself painting but you're now writing poetry, you're probably doing something right. If the path you're on does not resemble where you thought you'd be, you're probably doing something right. In the long run, stepping stones lead to other stepping stones and eventually to great discoveries."
I felt vindicated, because I have always gone against the narrative espoused by the myth of the objectives. I've studied across 6 different disciplines and never felt home at any one in particular. I love learning and using what I learn to create, implement or mainly just think of interesting things. At work, our organisation restricts 20% of our time on L&D. I dedicate 50% on the reg, to explore and learn new things even if they're not directly related to the areas I'm working on. Because of what i learned, i have been in positions where im the only one who has made progress in solving problems that are novel for the organisation or competency unit, doing what noone at the time who were willing, could. I think its because I spend so much time learning things - collecting the prerequisite stepping stones.
Pursue things because they are interesting. Like Moritz Schlick's philosophy on the meaning of life - the pursuit is not to have some lofty big goals (as not achieving them will leave you with anxiety while achieving them will leave you hollow and in search of a new goal), but to play with what you already have. That's also how you keep your youth.
The message of this book is inspired by AI research and focuses on the myth of the objective, which is ubiquitous and taken for granted so much so that those without them are somehow made to feel like a pariah in society, or simply lost and aimless in life.
This book makes the case that, its not simply that setting an ambitious (novel, not previously done) objective takes you closer to it, even if it doesn't guarantee it when it is planned for, but that it can infact take you further away from that objective (mainly due to deception resulting from our general lack of foresight which cages our thinking to assume that the stepping stones you take to that goal somehow should resemble it). But by embracing serendipity, using clues like a treasure hunter, we can pursue things that are interesting with some chance of stumbling upon something worthwhile, even though exactly what that will be, cannot be said.
The book provides some principles which are great for tracking progress, not only in your personal life, but also for bigger organisations. Also thinking through how someone could implement the ideas for living a life in the way this book paints gives it so much meaning, without the anxiety brought upon by the thought of 'not being where you are supposed to be' in life. It's liberating.
The research covered in this book is interesting, especially if you're familiar with AI and also reinterprets evolution in an interesting way (as a type of novelty searching algorithm).
"So if you're wondering how to escape the myth of the objective, just do things because they're interesting. Not everything needs to be guided by rigid objectives. If you have a strong feeling, go for it. If you don't have a clear objective, then you can't be wrong, because whatever you end up is okay. Assessments only goes so far. A great achievement is one that leads to more great achievements. If you set out to program computers but you're now making movies, you're probably doing something right. If you wanted to create AI but you're now evolving pictures, you're probably doing something right. If you imagined yourself painting but you're now writing poetry, you're probably doing something right. If the path you're on does not resemble where you thought you'd be, you're probably doing something right. In the long run, stepping stones lead to other stepping stones and eventually to great discoveries."
I felt vindicated, because I have always gone against the narrative espoused by the myth of the objectives. I've studied across 6 different disciplines and never felt home at any one in particular. I love learning and using what I learn to create, implement or mainly just think of interesting things. At work, our organisation restricts 20% of our time on L&D. I dedicate 50% on the reg, to explore and learn new things even if they're not directly related to the areas I'm working on. Because of what i learned, i have been in positions where im the only one who has made progress in solving problems that are novel for the organisation or competency unit, doing what noone at the time who were willing, could. I think its because I spend so much time learning things - collecting the prerequisite stepping stones.
Pursue things because they are interesting. Like Moritz Schlick's philosophy on the meaning of life - the pursuit is not to have some lofty big goals (as not achieving them will leave you with anxiety while achieving them will leave you hollow and in search of a new goal), but to play with what you already have. That's also how you keep your youth.
March 17, 2023
Explores a highly interesting and controversial notion: That, as a society and individually, we believe in the "myth of the objective", although that is often misleading. The authors understand this to be the central role that objectives play in our lives. Coming from the field of AI research, they apply another principle and explore its implications in different areas: a search for novelty with no clear goalpost.
While it sounded intriguing, I initially feared that this might be a gross oversimplification. It is indeed ambitious to connect the dots from an AI picture breeding application to the way we view and approach innovation, evolution and education.
However, while sometimes a little too repetitive, I found this model highly convincing. The central theme touches on a point I had often wondered about myself: Does it really make sense to evaluate the innovative capacity of ideas beforehand? Do we gain anything by training students to ace their tests? Personally, at least, in the case of education specifically, I have always perceived this as limiting and mostly missing the point. I used to think of this approach as overfitting to a somewhat arbitrarily defined "goal", making students miss much of the crucial exploration that can make a topic fascinating.
I also experienced this phenomenon recently when, after graduation, I started over again in quite a different field. When people ask, what I am studying now, the next question is nearly always what "one does with that". While I answer and vary my responses a little everytime, I really don't know or think much about what I am going to do afterwards and the reason I switched was just because I was curious, more interested and it seemed like the right thing to do.
This book points out how, for big questions, it is often for the best to not have a clear goal in mind, but rather to explore the full space of possibilities and just take one step at a time. Often, we tend to overestimate our capacity of foresight by quite a bit.
All in all, there were some cases where I found points to be overly emphasized, when they had already been made clear. Apart from that though, it's an inspiring read and leaves you wondering: How would the world around us change, if we permitted more exploration, just for the sake of it?
While it sounded intriguing, I initially feared that this might be a gross oversimplification. It is indeed ambitious to connect the dots from an AI picture breeding application to the way we view and approach innovation, evolution and education.
However, while sometimes a little too repetitive, I found this model highly convincing. The central theme touches on a point I had often wondered about myself: Does it really make sense to evaluate the innovative capacity of ideas beforehand? Do we gain anything by training students to ace their tests? Personally, at least, in the case of education specifically, I have always perceived this as limiting and mostly missing the point. I used to think of this approach as overfitting to a somewhat arbitrarily defined "goal", making students miss much of the crucial exploration that can make a topic fascinating.
I also experienced this phenomenon recently when, after graduation, I started over again in quite a different field. When people ask, what I am studying now, the next question is nearly always what "one does with that". While I answer and vary my responses a little everytime, I really don't know or think much about what I am going to do afterwards and the reason I switched was just because I was curious, more interested and it seemed like the right thing to do.
This book points out how, for big questions, it is often for the best to not have a clear goal in mind, but rather to explore the full space of possibilities and just take one step at a time. Often, we tend to overestimate our capacity of foresight by quite a bit.
All in all, there were some cases where I found points to be overly emphasized, when they had already been made clear. Apart from that though, it's an inspiring read and leaves you wondering: How would the world around us change, if we permitted more exploration, just for the sake of it?
November 5, 2022
There are a lot of ways to misunderstand The Myth of the Objective. I take it as a useful meditation, with aspects of The Tyranny of Metrics and Against Method, encouraging exploration. Be flexible, be curious, don't follow a plan for the sake of following a plan.
I don't recall these AI researchers ever talking about local maxima or explore/exploit trade-offs explicitly, as they seem to be trying to write for a broad audience. Natural evolution, human innovation, Picbreeder, and novelty search are their examples of unplanned greatness. They point out that evolution is about exploration as much as adaptation, and critique the dominance in the AI community of the experimentalist and theoretical heuristics.
Is a goal local (we know how to get there) or is it "great" (requiring steps into the unknown)? The authors are saying that if it's the latter, we're better off exploring via other heuristics than focusing only on a particular distant imagined goal.
They talk about "interestingness" a lot, and I think a missing note is that "interesting" shouldn't necessarily exclude a sense of what takes us closer to some distant goal.
We should also be watching for opportunities to take things in a different direction: if you find a path to something great on your way to something good, you can follow that path! (Don't let the good be the enemy of the great.)
I've sometimes been apologetic about the winding path of my life and career, as if I should have had a plan from the start and followed it without distraction. Would that have been better? I'm not sure. This book makes me think that a life or career built from a kit represents a failure to identify opportunities. What are the odds there was never a way to improve on an earlier plan? Blinders make you faster, but they limit where you go.
The book applies most to the "research" setting—where we really don't know what the future is going to hold, where we don't know the mechanics or topology of the search space. Objectives such as "hovercraft like in Star Wars" or "economy like in Star Trek." There are a lot of other things we'll find before we find these, and that's not a bad thing.
I don't recall these AI researchers ever talking about local maxima or explore/exploit trade-offs explicitly, as they seem to be trying to write for a broad audience. Natural evolution, human innovation, Picbreeder, and novelty search are their examples of unplanned greatness. They point out that evolution is about exploration as much as adaptation, and critique the dominance in the AI community of the experimentalist and theoretical heuristics.
Is a goal local (we know how to get there) or is it "great" (requiring steps into the unknown)? The authors are saying that if it's the latter, we're better off exploring via other heuristics than focusing only on a particular distant imagined goal.
They talk about "interestingness" a lot, and I think a missing note is that "interesting" shouldn't necessarily exclude a sense of what takes us closer to some distant goal.
We should also be watching for opportunities to take things in a different direction: if you find a path to something great on your way to something good, you can follow that path! (Don't let the good be the enemy of the great.)
I've sometimes been apologetic about the winding path of my life and career, as if I should have had a plan from the start and followed it without distraction. Would that have been better? I'm not sure. This book makes me think that a life or career built from a kit represents a failure to identify opportunities. What are the odds there was never a way to improve on an earlier plan? Blinders make you faster, but they limit where you go.
The book applies most to the "research" setting—where we really don't know what the future is going to hold, where we don't know the mechanics or topology of the search space. Objectives such as "hovercraft like in Star Wars" or "economy like in Star Trek." There are a lot of other things we'll find before we find these, and that's not a bad thing.
January 29, 2023
I tell you: one must still have chaos within oneself, to give birth to a dancing star. (c) Friedrich Nietzsche
The book's authors are Computer Science professionals, so it is appropriate to recall the expression: "Everything in software architecture is a trade-off."
The book is a good, high-quality set of examples for training the reader's neural network in a step-by-step planning model with a goal-setting loop after each stepping stone. The overarching image of the entire book is the exploration of a dark room. It is impossible to see the plan, but it is possible to set the goal of gaining new knowledge. Step by step, while preserving and analyzing the information obtained, the researcher becomes a greater version of himself.
Examples are given from scientific progress and inventions of household appliances, art, management, and software development. The transition to the theory of evolution is exciting. This approach is consistent with articles on the theory of evolution, for example: "Toward a theory of evolution as multilevel learning". From there: Importantly, learning here is perceived in the maximally general sense as an objective process that occurs in all evolving systems, including but not limited to biological ones. As such, the analogy between learning and selection appears obvious. Both processes involve trial and error and acceptance or rejection of the results based on some formal criteria; in other words, both are optimization processes.
The approach of setting global goals produces results. Otherwise, events such as Magellan's circumnavigation or Scott's expedition would not have occurred. However, now that I am defining my goals, I can better understand the potential side paths I am sacrificing by not seeing them in front of me. Similarly, by choosing the role of a treasure hunter, I am sacrificing the role of a new Magellan.
February 25, 2019
Superb reading. The kind of research that I would probably pursue if I would've stayed in academia. Takes the insights from AI algorithms and makes a strong case about how setting objectives is not the way to achieve true discovery. It argues that "deception" happens in problems when the distant, complex and desirable objective (such as finding human-like intelligence in AI) don't usually look like the "stepping stones" that gets them there: a worm doesn't look at all like a human, but was a necessary stepping stone for human evolution. So it argues that in complex problems, where a huge search space must be explored, one should forget about where one wants to "go", but better to focus on where one could "get" from the current position. A philosophy of life comes about: in complex problems like finding love, happiness, career, etc., just focus on gathering "stepping stones", and eventually, one will arrive somewhere great; perhaps not where one intended, but the important is for it to be great. I belive this has been my tacit way of acting most times in my adulthood, without a clear path, but only focusing in collecting interesting "skills" or "experiences" which I believe to be worthy because they might be somehow useful in some other way. Robert Greene sends a similar message in "Mastery", saying that a way of becoming a true expert is gathering skills, being the best at each specific task and hoping that at some point, those skills reveal some greater skill or job or discovery or talent or field. What a great book.
May 28, 2023
A book about searching for good outcomes without rigid expectations of what "good" means.
This book describes Artificial Intelligence research into "novelty search," and attempts to generalize that idea to many different domains, like choosing a career, scientific research, technological innovation, and evolving life. The key idea is that setting an objective up front and steadily working towards that goal often leads to getting stuck in dead ends and missing less obvious paths that might lead to better outcomes.
The ideas in this book are powerful, important, and expressed clearly in simple terms. If anything, the authors may have made this too easy to read. It spells out a lot that for me was obvious, and repeats its main points over and over. It's a short book, but it's still twice as long as it needs to be. It's also unclear what this book is meant to be. Is it a Computer Science book discussing the novelty search algorithm? A study of how the concept of "objectives" shapes our collective ontology? A manifesto on how AI research should be done? A self-help book? It seems like all of the above. That said, the meandering writing style is worth it for such an inspiring premise and the interesting examples that motivate it.
This book is a quick, easy read. It may be eye-opening or obvious, depending on your starting point. If you're interested in surprising discoveries, the chaotic path to innovation, and how to pursue something really new and interesting, definitely give this a read!
This book describes Artificial Intelligence research into "novelty search," and attempts to generalize that idea to many different domains, like choosing a career, scientific research, technological innovation, and evolving life. The key idea is that setting an objective up front and steadily working towards that goal often leads to getting stuck in dead ends and missing less obvious paths that might lead to better outcomes.
The ideas in this book are powerful, important, and expressed clearly in simple terms. If anything, the authors may have made this too easy to read. It spells out a lot that for me was obvious, and repeats its main points over and over. It's a short book, but it's still twice as long as it needs to be. It's also unclear what this book is meant to be. Is it a Computer Science book discussing the novelty search algorithm? A study of how the concept of "objectives" shapes our collective ontology? A manifesto on how AI research should be done? A self-help book? It seems like all of the above. That said, the meandering writing style is worth it for such an inspiring premise and the interesting examples that motivate it.
This book is a quick, easy read. It may be eye-opening or obvious, depending on your starting point. If you're interested in surprising discoveries, the chaotic path to innovation, and how to pursue something really new and interesting, definitely give this a read!
June 16, 2023
The central idea of the book is reasonably compelling and might be stated as follows. A sufficiently complex and ambitious goal is rarely achieved with an explicit plan that charts a course from start to finish. Rather, greatness is more often than not the product of unfettered, interest- and novelty-driven exploration which creates ample surface area for serendipitous surprises. For instance, the invention of penicillin happened by accident, and vacuum tubes weren't invented with their utility for computing in mind. The author—who invented an obscure genetic algorithm to update neural network weights—argues that we ought to give up the notion of having explicit objectives, and instead adopt a more experimental approach to engaging the world, one which gathers "stepping stones" from which we can explore new, untraversed areas. Novelty and interest ought to trump linear, unidimensional objectives.
Sure, yeah, there's something to be said about the unpredictability of scientific advances (see Thomas Kuhn) or the idea of exposing yourself to positive shocks / tail risks / Black Swans (see Nassim Taleb). But the book is an extended shower thought and pedantic exercise in shitty analogical reasoning which, by the final few chapters, devolves into utter gibberish. Let's not forget that "neuro-evolution of augmenting topologies" lost ground to backprop and reinforcement learning.
Sure, yeah, there's something to be said about the unpredictability of scientific advances (see Thomas Kuhn) or the idea of exposing yourself to positive shocks / tail risks / Black Swans (see Nassim Taleb). But the book is an extended shower thought and pedantic exercise in shitty analogical reasoning which, by the final few chapters, devolves into utter gibberish. Let's not forget that "neuro-evolution of augmenting topologies" lost ground to backprop and reinforcement learning.
November 13, 2023
They criticize pursuing objectives and instead argue for novelty search. Novelty search is when we follow the gradient of interestingness. In a classical search based algorithm it could be something like exploring a state that we haven't visited before and that is different from others, instead of using a heuristic to find the optimal path to a given goal.
For large discoveries and revolutionary innovations we should not set an objective and try to go for that. We should instead go towards the path that is most interesting. They give several examples of discoveries that have been made by people where the stepping stones to the innovation were found by others working on other things, the favorite being the Picbreeder experiment.
The problem with their approach is that it doesn’t tell you anything about how to reach a goal. Using novelty search you will find interesting new stuff but you will most likely not solve the problem that you aimed to solve. There is also a lack of comparisons in the book of how novelty search compares to objective search when one wants to solve a specific problem. As they mention, novelty search has shown a lot of problems without intending to.
The book has some interesting ideas, but as is common with these kinds of books is that they sell their idea as more revolutionary than it actually is.
For large discoveries and revolutionary innovations we should not set an objective and try to go for that. We should instead go towards the path that is most interesting. They give several examples of discoveries that have been made by people where the stepping stones to the innovation were found by others working on other things, the favorite being the Picbreeder experiment.
The problem with their approach is that it doesn’t tell you anything about how to reach a goal. Using novelty search you will find interesting new stuff but you will most likely not solve the problem that you aimed to solve. There is also a lack of comparisons in the book of how novelty search compares to objective search when one wants to solve a specific problem. As they mention, novelty search has shown a lot of problems without intending to.
The book has some interesting ideas, but as is common with these kinds of books is that they sell their idea as more revolutionary than it actually is.
Displaying 1 - 30 of 81 reviews