|
v If you want to have a good invention, have a lot of them. Attributed to T. A. Edison
p.1 The history of life is a history of innovations... are there principles behind them? Is there a property that facilitates innovations, regardless of their physical manifestation? I argue here that the answer is yes, and I characterize this property - I will call it innovability... an evolutionary innovation... is usually easy to recognize: a new feature that endows its bearer with qualitatively new, often game-changing abilities... they may... create broad platforms for future innovations
p.2 While Darwin's theory rightly emphasized the role of natural selection in preserving useful variation, it left untouched the question how new and useful variation originated.
p.2 genotypes... the genetic material (DNA or RNA) of organisms... all observable characteristics of organisms, their phenotype.
p.2 a theory of innovation may have little to say about any one specific innovation. Instead, it may provide an explanatory framework for innovations in general... it can be powerful in its generality, without trivializing the individual innovation and its marvelous uniqueness.
p.2-3 Here is a minimal list of what a theory of innovation should accomplish.
- (The paramount problem.) It should explain how biological systems can preserve existing, well adapted phenotypes while exploring myriad new phenotypes...
- It should unify innovations that involve different levels of biological organization...
- It should be able to capture the combinatorial nature of innovation...
- It should be able to capture that the same problem can be solved by different innovations. Innovations can be viewed as solutions to a problem an organism faces...
- It should enable us to study how environmental change influences innovability. The environment determines whether any one novel phenotype is an innovation...
- It should be applicable to technological systems. A theory of innovation... should also apply to non-biological systems. [JLJ - yes, even to game theory]
p.4 Innovations originate with a genotypic change whose effects translate into a phenotypic change. Thus, if we do not understand how exactly genotypic change maps into a phenotypic change, we cannot hope to develop a comprehensive explanatory framework for innovation.
[JLJ - Yes, and this is where (in playing a complex game of strategy) in my mind we need to be able to form a diagnostic test of the adaptive capacity to mobilize coercion. We use intelligently constructed estimates of what typically develops in the phenotypes (future game positions) from changes in genotype (equivalently, a move made in our present game position).]
p.5 I view a system as a set of elements or parts that cooperate to perform a task.
p.13 Some innovations may arise in a single large step, but many arise more gradually, through a series of changes with individually modest effects.
p.14 A genotype network is a set of genotypes that have the same phenotype.
[JLJ - This wins the 2016 John L Jerz award for the most useless definition ever constructed. Gee, thanks for that Andreas. What a smart and helpful man you are. How lucky we are to have you define terms like that for us. What would we do without you.]
p.14 All human understanding requires abstraction from the unfathomable complexity of the world around us... In my view, the concepts of genotype spaces and genotype networks are the right level of abstraction to understand evolutionary innovation comprehensively and systematically.
[JLJ - Ok, so if that is true then why is no one reading your book? Kind of an academic fail if you ask me. Looks like you tried to fix that by writing Arrival of the Fittest in 2014. But that didn't work either, because you insisted on talking about genotypes and phenotypes and genotype networks. That's ok. You can still try again. In my view, the concepts of genotype spaces and genotype networks are *not* the right level of abstraction to understand evolutionary innovation comprehensively and systematically.]
p.15 Most phenomena I will discuss do not require that the genotypes on the same genotype network have exactly the same fitness. For example, many mutations in proteins of well-studied organisms are deleterious, but weakly so [227, 676]. Such weakly deleterious mutations can rise to high frequency in a population by chance events (Chapter 7), or they can persist until other mutations arise that compensate for their deleterious effects and thus preserve them [393, 428]. They are no strong impediment to evolutionary change on one genotype network. Conversely, many mutations that increase fitness do so only very slightly, and their fate can be determined by the same forces that determine the fate of neutral mutations [310, 676].
p.15 A few words are now necessary to motivate my use of the neologism "innovability." Perhaps a more popular word, such as "evolvability" might be a better choice? ...When studying innovation, however, qualitative variation becomes important. The approaches I use below to analyze different phenotypes all aim to distinguish such qualitative differences. We currently do not have a good word to refer to such qualitative differences. This is the main motivation for using a new word, innovability.
p.79-80 Chapter 1 listed several minimal requirements for a theory of innovation. I will briefly revisit them... The first requirement is to explain how the old can be preserved while the new is being explored... The second requirement is to unify different kinds of innovations... The third requirement is to capture the combinatorial nature of innovation. The framework I use is also ideal in this regard, because it explicitly captures innovation as new combinations of system parts... that give rise to novel genotypes... The fourth requirement is to capture that the same problem can be solved through different innovations... The fifth requirement regarded environment change. I will discuss it in Chapter 11. The sixth and last requirement regarded applicability to technology; it is the focus of chapter 15.
[JLJ - So Wagner 'proves' his concept by using 'rules' he created earlier in the book. The reader can now verify that Wagner is not violating any of Wagner's laws - he must therefore be correct...?]
p.83 The previous chapter showed that extended genotype networks with phenotypically diverse neighborhoods are necessary for innovability.
p.91 Genotype networks are examples of self-organizing structures.
p.92 In Chapter 1, I quoted the geneticist de Vries who stated that natural selection cannot explain the origin of novel phenotypes [170]. More than 100 years later we can say this: Genotype networks help explain the arrival of the fittest, and natural selection permits their survival.
p.123 robustness qualitatively facilitates evolutionary innovation.
p.157 The more environments a system needs to function in, the more complex it tends to be, and the more robust it will appear in any one environment. Thus, the robustness that brings forth genotype networks - and thus innovability - is ultimately a result of life's need to cope with changing environments.
[JLJ - What drives innovation is that life must out of necessity function in multiple environments, and must function well. But an organism does not just function, it continuously produces a trick from a bag of tricks that works, and often well. All life struggles to exist in a predicament, often of its own choosing, and innovation emerges from the variation of successful copers.]
p.198 innovability comes at the price of system complexity.
p.205 Engineering implications for fault-tolerant circuitry
[JLJ - First, is anyone willing to pay for the luxury of 'circuitry' being fault tolerant? Wagner also forgets or never understood that electronics have a high reliability to begin with. Who will pay for a redundant computer in their car, ready and willing to take over if the main one fails? Likely no one. In the unlikely case this happens you have the car towed to the dealership and they will fix it for you. But in a rocket or airplane, with lives at risk, there is possibly more of a need.]
p.211 observations from the last two sections have implications for the design of technological systems, especially of adaptive systems that can change their configuration and behavior in response to new environments... In such systems, hardware that can be reconfigured on-the-fly, while the system performs a task is highly desirable.
p.211-212 Engineers tend to value simplicity and elegance in system design. The 16-gate circuit from Figure 15.7 clearly violates this principle. Its apparently unnecessarily complex design, however, has other advantages. As Figure 15.4b showed, larger circuits that compute a specific function tend to be more robust to wiring changes... Figure 15.8a shows that this increase of robustness is a generic property that occurs for many different functions.
[JLJ - Figure 15.7 is a useless example. Wagner apparently does not understand electrical circuits. Many of his extra gates do nothing and can be deleted - they will be optimized away by the FPGA when the design is compiled. His OR gate with split input is really just a bare wire with a delay to it. Wagner should consider a workable example with a purpose, not this academic construction which does nothing - except possibly reveal his ignorance of the subject.]
212 higher circuit complexity facilitates access to a larger number of novel functions.
In sum, circuit complexity increases robustness and functional versatility. A lack of simplicity (and elegance) in circuit design may be the price to pay for these features. Evolvable systems may be neither simple nor elegant, but complex and messy.
p.213 Robustness and other features of such circuitry tend to increase with circuit complexity. Circuit complexity may thus be a price to pay for both robustness and functional versatility. In systems like these, the existence of fault-tolerant circuitry is a generic property of circuit space. In addition, circuit spaces with this organization are well suited for the design of autonomous adaptive systems that can dynamically reconfigure themselves. The reason is that such systems should be able to explore many new behaviors while reconfiguring themselves only minimally.
[JLJ - Yes, but you might not need a fault tolerant design if you have an easy way to swap in a back-up, or perhaps two fault-prone circuits in parallel and an automatic failover switch. Modern computer servers have RAID configurations of hard drives and hot-swap power supplies. Critical data is often backed up to "cloud" storage. There is no need for an exceptionally high-reliability component design in each case, and you get effectively the same effect.]
p.215 The greater a system's ability to adapt to new or changing environments, the more complex it needs to be - the more parts it must contain. In other words, environmental change is one key driver of biological complexity. This complexity indirectly leads to robustness in any one environment, where small genetic changes in individual genotype do not affect the phenotype... Evolutionary adaptation or innovations that organisms acquire in order to cope with novel environments will tend to increase system complexity. Such increased system complexity implies robustness, which brings forth genotype networks. These networks facilitate innovations that allow the conquest of new environments or persistence in changing environments. The ascending spiral of the figure hints that this process may well be self-accelerating.
|
