| Chapter 34 | |
| Deploying Innovation | |
| Yankee ingenuity | |
| America has a wealth of valuable assets. It has, relatively, a lot of land per person (7 acres). It is rich in natural resources, and has numerous ports on 3 coasts, extensive internal pipeline, transportation, and communication networks. The financial sector is strong and American citizens are themselves a huge market. The US maintains a university system that international students, professors and researchers ßock to. | |
| Perhaps most importantly, we have a tremendous wealth of innovation. The US is not alone here. Even very poor nations innovate, often more appropriately for their own problems than an outsider could. Nonetheless, many of the most significant innovations of the last century came from the USA. | |
| Why are we stuck? | |
| Despite all those advantages, America appears, to many observers, to be stuck. The American dream seems to have unraveled. | |
| Our GDP may be up in the last couple of years, but median income is not. Job prospects for college grads are slim, and younger non-gradsÕ fate is worse. The percentage of a middle-class personÕs income that must be spent on the necessities of housing, food, and health care is going up, when technological progress should result in that percentage decreasing. | |
| In 2010, US vice-presidential candidate Sara Palin, in commenting on ObamaÕs 2008 core campaign promises, asked the rhetorical question ÒÔHowÕs that hopey, changey stuff working out?Ó [Gonyea 2010]. At the time it sounded to many on the Left as just plain stupid, but history has shown its sentiment to be an accurate prediction. We need morethan just hope and a desire for change. | |
| Two steps forward, but how many steps back? | |
| The march of science and technology progress goes on, and even accelerates. But does this translate into improvements for most people in society? | |
| Our climate is less hospitable to life, as is our politics of fear and war. Only 27% percent of Americans think the country is heading in the right direction [Rasmussen 2017]. Our national energy policy can be described either as Ònon-existentÓ, or as ÒBig OilÓ. Take your pick. Gridlock is the politest way to describe Congress. The Supreme Court that has declared Òcorporations are peopleÓ. Our police departments routinely use the Òcash and carryÓ money confiscation schemes to fund themselves and unnecessarily lethal force. A kid shooting family members in their home is so common that it is no longer a shocking news story. Our Prison Industrial Complex is second to none. The trillions of dollars the US has spent this millennium on Òthe forever warsÓ have arguably made the world, less safe. What gives? | |
| Conventional mechanisms for innovation | |
| The conventional wisdom is that our society has built-in mechanisms for fixing itself. If youÕre willing to Ògo through the proper channelsÓ, we are supposed to have self-correcting procedures that allow innovation to take place. Problem is, those mechanisms have gotten rusty. | |
| Elections are the means which democracy is supposed to fix itself. We see a decreasing percentage of eligible voters voting (36% in 2014 elections). This means it takes 1 more vote than 18% to elect our leaders. Those that do vote tend to do so against their own self-interest. More on problems with voting in the chapter The trouble with voting. | |
| Citizen protest. The latest significant ones as of this writing were Black Lives Matter of 2014-5, and Occupy Wall Street, in 2011, spreading to 951 US cities and numerous countries. Obama even said ÒWe are on their sideÓ [Halper 2011]. Then Obama signed into law an anti-protest bill that makes it easier to criminalize protests. [Lithwick 2012]. | |
| Liberal democracies allow citizen protest to take place as a harmless Òescape valveÓ. But government and corporate officials donÕt take citizen protest alone as signaling a necessity for change, unless backed up by votes, and by money. | |
| Structural change of government takes a constitutional amendment. For practical purposes, we cannot amend our constitution. It is dead. With a dynamic technical, social, economic and physical environment, a dead constitution will lead to a dead country. | |
| Startups are supposed to be the cutting edge of Capitalism. Since 90% of startups fail, we have the efforts of 90% of our innovation workforce pretty much wasted. Great for the venture capitalists, who make money off of the proceeds of the successful 10%, but for the individuals who donÕt cash out, tough luck. | |
| And tough luck for society in general, which doesnÕt get to learn very much from the failures. In science, itÕs OK for 90% of projects to ultimately fail, because the scientific community learns along the way, by publication and education activities. | |
| But most startups are secretive, so nobody outside the startup learns anything. The participants in a failed startup are usually too burned out at the end to have much more energy for further developing their ideas or for teaching others. Some go on to use what they learned in other startups, but most tend to desert innovation, looking for more stable jobs. The sale of intellectual property to buyers of failed startups is a very poor mechanism for recycling their potential for innovation. | |
| We (the authors of this book, the country, and the world) have many innovative solutions that show promise in tackling our big problems. The roadblock is deploying innovation. ItÕs time for something new. | |
| Heuristics for self-improvement | |
| LetÕs characterize these mechanisms using the language of computer science. Artificial intelligence studies which are procedures that are Òrules of thumbÓ heuristics, used for perhaps only partial solutions to problems. They donÕt always work. If you donÕt have a total solution to the problem, theyÕre better than nothing. Sometimes they can be stepping-stones to more effective solutions. | |
| When you design government or business structures, you have to design heuristics for making decisions. You canÕt make all decisions in advance, so the best you can do is design a decision-making process that makes it more likely that good decisions will result. (Our candidate in the chapter Reasonocracy.) | |
| Search algorithms are a kind of heuristic. They are methods for finding solutions to problems in a search space, a universe of possibilities. AI also studies machine learning, another kind of heuristic, by which a program can become better over time, without needing to be explicitly redesigned. | |
| In the language of AI, then, elections, protests, laws, amendments, and startups are all heuristics (processes) for improving society. They donÕt always work, but they increase the probability that the system can improve itself. Especially when compared to a more rigid system without mechanisms for change, like dictatorships or feudalism. | |
| Generate & test and hill climbing | |
| As heuristics, they mostly fall into two well-known categories: Generate & Test; and Hill Climbing. In Generate & Test, you have two processes. The first process, Generate, spits out possible solutions. The second process, Test, tries to determine whether each possibility is good or not. | |
| In evolution, mutation is the Generate process and natural selection is the Test process. In the innovation ecosystem, though a lot of planning goes into a startup, from the larger economic perspective, we could model entrepreneurs as the Generate process, and venture capitalists as the Test process. VCs weed out the implausible startups and fund the promising ones. After funding, incremental steps of success in the marketplace form another kind of Test process. In US Democracy, running for office and proposing laws are Generate processes. Voting is the Test process. | |
| Generate & Test, alone, is known to be pretty weak as a heuristic. If somethingÕs wrong with the Generate process, itÕll just keep spitting out proposals that fail the Test. If somethingÕs wrong with Test, it wonÕt be able to tell the difference between good and bad proposals. But even if we do find a decent solution, G&T alone canÕt determine why something succeeded or failed. Modern machine learning has far more sophisticated procedures for generating and testing, that use feedback to improve both processes. | |
| Hill Climbing is a way to make incremental change. Since big change is usually risky, you try a few ways of making small changes and see which one improves the situation. Then you look for another small change. Modern business and politics also use Hill Climbing. Usually, electing one more candidate, passing one more law, introducing one new product, or starting one more company doesnÕt change the whole system very much (there are notable exceptions). | |
| Entrepreneurs and political radicals may propose new and risky stuff. But the job of established businesses and politicians is actually to avoid and minimize risk, as much as possible. What they really like is Hill ClimbingÑminor tweaks to already-successful models, and models that can succeed a little bit at a time, then use that success to grow larger. ThatÕs why we get so many lookalike startups and lookalike politicians. | |
| Unfortunately, Hill Climbing is also a pretty weak heuristic. It has the nasty problem that you can get stuck in a local maximum. Hill climbing means taking incremental steps up the hill (continual, but minor improvements), and never down, until you get to the top. But once you reach the top, you can discover that there were other, higher mountains around you that you can never scale, because it requires first taking steps down before you step up again. | |
| Hill Climbing also has the problem that it has to take just one small step at a time. It is mostly good for incremental change, but not so good for innovation or more fundamental change. AI and machine learning recognized this early on, and modern techniques employ far more sophisticated strategies for making changes, large and small. | |
| Hill Climbing has the some of the same problems as Generate & Test. TheyÕre pretty blind, stupid strategies. Both can tell you what stands a better chance of working or not in particular situations if you donÕt have much else to go on. But they canÕt tell you why and what to do about it if the results arenÕt satisfactory. They donÕt have any theories. | |
| Now weÕve got our answer as to why itÕs so hard to deploy innovation in todayÕs society. Innovation isnÕt about random (generated) ideas or bit-by-bit change. An innovation usually offers a whole new theory for why things are the way they are, and requires a radically different approach. If the only mechanisms our society offers for adopting new innovations are Generate & Test, and Hill Climbing, it will be hard to adopt innovation. What weÕre doing in this book is giving you a new theory. Up to you to decide if it makes sense. | |
| Goal stacks | |
| WeÕve already talked about the idea of a goal stack (in the ÉEven Possible chapter), solving a big problem by breaking it down into smaller and smaller subgoals, and finally down to specific actions. This is the Òdivide and conquerÓ heuristic that people commonly use in everyday life. | |
| In political and economic hierarchies, this breakdown is embodied in people. Leaders are tasked with planning to solve ambitious goals by deciding what the subgoals are. Lower-level employees are tasked with carrying them out by taking small steps. Orders ßow from top to bottom, almost never in the other direction. If everything goes according to plan, this usually works. | |
| But in the real world, everything doesnÕt always go according to plan. Is the subgoal the right way of achieving the larger goal? Did each action have the result that was intended? If not, you canÕt just keep working on the current goal. You typically need to go back up the goal stack, and reconsider higher-level goals. Maybe you need to replan, abandon the current plan and break up the high level goal into a completely different set of subgoals. AI has a subfield, called partial-order planning, that deals with these issues. | |
| ThatÕs what innovation is. Innovation proposes a new solution to a high level goal, that obsoletes an existing subgoal plan. In effect, its Òredivide and conquerÓ. | |
| For example, in transportation, we can adopt the goal of trying to make a more fuel-efficient car, or a train that goes faster. Those would be incremental advances to already-existing plans. But we could also propose an innovative new transportation system like PRT (see the Transportation chapter), that is neither a car nor a conventional train, but does solve the higher-level problem of urban transportation, which is what we were trying to do in the first place. | |
| In personnel hierarchies, the problem is that people are assigned and committed to implementing incremental plans. Adoption of the innovation is a threat to them. The best chance for innovation is in those rare cases when the goal itself is entirely new, or there is no other existing plan to achieve an existing goal. | |
| Low-level people do not have the authority to go up to a higher level and change the plan. High-level people are also disinclined to adopt innovation, because it challenges their competence in executing the original plan obsoleted by the innovation. They are already highly invested in the current plan, and they perceive the innovation as being too risky. Often, they incur some costs and risks in implementing the innovation, and reap no immediate, personal benefit if the innovation succeeds. | |
| It is the nature of the hierarchy itself that presents the major obstacle to innovation. We elaborate this argument in the chapter No Leaders. | |
| Who decides about deploying innovation? | |
| LetÕs review some of the players. | |
| Big government | |
| US Democracy is supposed to be working for the people, but in The trouble with voting we argue that it is an institution supported primarily by large corporations to help them efficiently preserve the status quo. It precludes innovation. | |
| True, governments fund research, though investment in research is a trickle compared to other major expenditures. But the tax code, our laws, enforcement, our process for selecting leaders, and increasing income disparity, all support the view that government is more interested in the status quo than innovation. | |
| One of the most important kinds of innovation we can have is to drastically reduce the cost of a product. Since most of the cost of a product is in labor, decreasing production costs usually means fewer jobs required to fulfill demand. To a politician, jobs means votes. To a labor leader, jobs mean union membership. To a contractor, jobs mean billable hours. | |
| There are many ways that leaders preclude change. YouÕll hear: | |
| ÒWe need another study to make sure its safe.Ó | |
| ÒWe donÕt want to gamble with the taxpayers moneyÓ. | |
| ÒWeÕre putting out a request for proposals to make sure we get the best deal.Ó | |
| Many boil down to delaying tactics and these can literally go on forever. This is why we need complexity management tools (see Tools for Reasonocracy). | |
| Little government | |
| ÒPower tends to corrupt, and absolute power corrupts absolutelyÓ. So it stands to reason that smaller governments, i.e. local governments, should be less corrupt. However, we see a trend in national parties gaining increasing inßuence in local government. For instance, on local labor union laws, and on gerrymandering congressional districts [Fang 2014]. | |
| Big companies | |
| Companies will tell you theyÕre always open to innovation. ÒBuild a better mousetrap and the world will beat a path your door.Ó Rarely does it happen. But it does happen. | |
| More often, if they canÕt defeat a new idea in the marketplace, big companies will Òbuy it and kill itÓ. Eventually it becomes too widespread to Òbuy and killÓ, so some company will Òbuy and promoteÓ. This is a much slower process than need be, but even the most entrenched status quo gives way eventually. ÒThe average life expectancy of a multinational corporation-Fortune 500 or its equivalent-is between 40 and 50 years.Ó [Foster 2015]. | |
| Little companies | |
| If big companies are hamstrung, startups and small companies are supposed to be the vehicle for innovation. But the survival pressure on small companies is such that they are often tempted to stop starving to death and promote an incremental solution, even if it is insufficient for the customersÕ needs. Companies want innovationÑbut not primarily to help their customers. Rather, itÕs to get competitive advantage over other companies. Even a little advantage is enough to defeat another companyÑwhy risk more? | |
| If an incremental improvement is successful, it fills the niche of Òsome kind of improvementÓ. Because of the overhead of convincing the powers-that-be to consider any new solution at all, it may then become harder to consider a more radical solution that might actually be sufficient. | |
| Capitalism enforces this process. Proposers of the incremental improvement become another obstacle that our radical innovator has to overcome. | |
| Example: You might think the fuel efficiency of cars is improving. [WantToKnow 2017] says the 1908 Model T got 25MPG and the 2013 cars average 24.7MPG. Those 2013 cars are way better than the Model T in most respects, but fuel economy doesnÕt happen to be one of them. We have Òfilled the nicheÓ of Òbetter carÓ with more cupholders, air conditioning, reliability, etc. but thatÕs not helping carbon ppm. (Electric cars are promising to break this dismal trend.) | |
| HereÕs another, all too common, process under Capitalism. Two inventors have different, insufficient solutions to a problem. They work at different companies. Perhaps these two solutions donÕt preclude one another, theyÕre merely partial solutions. Capitalism discourages them from teaming up and making a sufficient solution due to the competition between their respective companies. ÒDivide and be defeated.Ó | |
| Consumer reluctance | |
| Companies say that competition between them drives innovation. But, in the Can Capitalism be Saved? chapter we show that much of the real competition is between a company and its customers. | |
| The company says ÒCustomers are our #1 priorityÓ. This is almost never the case. Making a profit is. They make a profit by getting customers to pay as much as possible for a product. To do that, marketing, to be polite, frames the product in the best possible light. To be impolite, they lie. By doing so, each such company decreases the trust a consumer has in companies in general. | |
| So when a new (or old) company comes out with what they say is an ÒinnovationÓ, customers are more than a little wary. This gives innovation a bad reputation, and prevents consumers from taking the risk of adopting innovation. | |
| Is that mousetrap really better? | |
| Of course, not every proposed innovation is really worth adopting. Before adopting an innovation, we should really do cost-benefit analysis to understand whether the benefits outweigh the costs and the risks. | |
| Often overlooked, also, is the opportunity cost, that is, the cost of doing nothing or failing to adopt complete solutions. When Republicans claim that climate measures will Òhurt the economyÓ they are perhaps correct in the short term but incorrect in the long term. | |
| Some mousetraps arenÕt better | |
| The vast majority of inventions fail. They are, by definition, unproven and therefore risky. Often the ÒinnovationÓ is good for some particular reason or situation, but other reasons or broader situations make it not worth it. We canÕt eliminate this risk, but we can be strategic with analysis to kill bad ideas before they waste too many resources. There are often ÒplateausÓ in design spaces. Given a set of parameters, we can often preclude an infinite design process with the rationale of: ÒItÕs not likely we can do much betterÓ. | |
| Bang per buck | |
| Bang-per-buck is a useful metric for comparing solutions, especially when there are various quantities of something that you can buy. Suppose product X is cheaper than product Y but it is less effective than product Y. We can buy more of X to make up for its less effectiveness. If X is better bang-per-buck wise, X is a better solution than Y. | |
| Net solutions | |
| Here, we donÕt mean ÒnetÓ as in Internet, itÕs ÒnetÓ as in Òafter everything is consideredÓ. Any complex solution is very likely to have both good and bad in it. For instance: a refrigerator that uses little energy to run, may take so much energy to produce (embodied energy) that over its lifetime, it is no better than a less efficient-to-operate refrigerator. | |
| The same kind of thing can happen with money. Using an efficient product saves on operating expenses, but the up-front costs are so high they will never be recouped by the 10 year lifetime of the product. Evaluate a solution on its net benefit, not the benefits of a given feature in isolation. | |
| Partial vs precluding solutions | |
| Say our goal is an Òoff-the-gridÓ house, one that doesnÕt require electricity from a citywide source. There is a refrigerator that uses half the energy of our old one. It will not get us off the grid by itself, so it is an insufficient solution for our goal, but it is a partial solution. The efficient refrigerator in combination with other efficient appliances and solar panels complete a sufficient solution. | |
| LetÕs say thereÕs a new maximally efficient inverter for our solar energy system. It produces 37 volts which is incompatible with all appliances. Using it would preclude the other parts needed to make it a complete solution event though the inverter is more efficient. Hill climbing gets us the more efficient inverter, but it strands us on a foothill, wherein the real mountain we want to climb is the Òwhole house efficiencyÓ. | |
| Sufficiency of solutions | |
| A very important consideration is whether the proposed solution is actually sufficient to solve the intended problem. A big problem with incremental solutions is that people will get distracted by the fact that they promise an improvement. They might then ignore the fact that that improvement may not lead to a sufficient solution. FryÕs adage: ÒAll thatÕs necessary for the bad guys to win, is that the good guys are distracted by the insufficient.Ó | |
| Nowhere is this more evident than in proposals for dealing with climate change. Consider the amount of the carbon in the atmosphere, now 400 parts per million (ppm). The sustainable ppm is 350. [350.org 2017]. ItÕs rising at 2 ppm per year. Suppose we have an array of proposals: | |
| Proposal Result | |
| 0. Do nothing Unchecked emissions. Disaster! | |
| -1. Better fuel economy Reduced acceleration of emissions | |
| -2. Paris Climate Agreement Limited emissions growth | |
| -3. No fossil fuels. PRT. Slow decline in emissions. | |
| -5. No emissions. Sequestration. Radical decline in emissions. | |
| DonÕt take this too literallyÑin this chapter weÕre trying to make a general point about solution methods, not the details of various climate plans per se (but see Transportation, about Personal Rapid Transit (PRT)). | |
| Solution 0, doing nothing, is roughly what weÕre doing now. That causes the situation to worsen by 2ppm per year. Not a good option if youÕre interested in long-term viability of the planet, but thatÕs the overall ÒplanÓ of US Democracy, the worldÕs worst polluter per capita. We humans tend to delay hard choices until true catastrophe hits. | |
| Solutions -1 and -2 are Òpolitically achievableÓ, and make activists feel good because they are Òimproving the situationÓ. But they are insufficient niche fillers. They wonÕt tip the balance to reducing atmospheric concentration of carbon, no matter how successful they are at getting adopted. | |
| Solutions -3 and -5 might be Òpolitically unacceptableÓ, but theyÕd be sufficient. They would likely constitute a means of reducing carbon ppm and halting climate change. Some methods of sequestration, such as geoengineering, might involve a lot of risk, and itÕs not clear weÕd have to go that far, but it would be worth considering. | |
| Solution -5 is the least likely to be chosen even though it could theoretically get us to a healthy state faster. If we were deciding between -3 and -5, weÕd hear ÒThe perfect is the enemy of the good.Ó and that argument would have merit. But, applied to comparing -1 to -3, it would be incorrect. -3 is more perfect than -1, and riskier to deploy, but -1 is just plain old insufficient. | |
| Search by design | |
| With our complex world, how can you find the best thatÕs out there? Even with todayÕs advanced web search engines, you have to know the right words to enter to search for. | |
| One strategy is to search by design. First design what you think the ideal solution would look like, not in detail but enough to clearly articulate high level product features. | |
| Say you are looking for efficient cars in the USA. If you understand car usage, you know that average people in a moving car is about 1.2. So your Òideal designÓ might have 2 seats. If they have any more, thatÕs extra weight and volume that wastes energy. Once you realize that, it becomes easy to recognize inefficient carsÑthey have more than 2 seats! | |
| Most products will have numerous criteria. Search first by the most restrictive one, and filter the remainder with the other criteria. | |
| Recap: The real InnovatorÕs Dilemma | |
| The creative spark of innovators is alive and well all over the planet. This chapter argues that the hard part of innovation is moving that Òa-haÓ from idea to reality at scale. Clayton ChristensenÕs InnovatorÕs Dilemma [Christensen 1997] posits that companies favor incremental changes over disruptive ones, even though that strategy may kill the company. The memes of ÒToo little, too lateÓ and ÒThe road to hell is paved with good intentionsÓ surround insufficient solutions. | |
| Nobody will admit to being anti-innovation. But most big decision makers act that way. They have toÑthey are embedded in command-and-control hierarchies, where they have incentive only for incremental change and not innovation. They work in business and government structures that have only Generate & Test and Hill Climbing as improvement heuristics. | |
| Science doesnÕt limit itself that way. The reason that science is so much better at generating and adopting innovation is that it is willing to go up its own goal stack when it needs to. It considers not just incremental improvement, but theories about why things are the way they are, and what they could be. If a new theory requires non-incremental change, well, so be it. Redivide and conquer is embraced, not excluded. We can look particularly to the science of AI for inspiration, which explicitly studies goals, planning, action and change. | |
| In The Structure of Scientific Revolutions [Kuhn 1962], Thomas Kuhn dissects how this process works (and the obstacles that science, too, faces, when an innovation is too disruptive). This is why we advocating making economic and political structures work more like the social processes of the scientific community (see The process of Science and subsequent chapters). Maybe then weÕll get our better mousetraps. |