Key Questions
Headlines about research generally have the same form:
• “Breakthrough in Nanotechnology Will Enable Low-Power Circuits”
• “New Medicine May Cure Cancer”
• “Algorithm Designed by Evolution Will Make Our Computers More Secure”
These headlines intrinsically encode society’s value system when it comes to new technology: either today or sometime in the future, whatever it is it, must prove useful.
Of course, there is blue-sky research and/or pure science, but funding for these is relatively scarce. In the UK, for instance, two-thirds of R & D is funded by businesses rather than by the government,1 and, even in universities and government labs, much of the work done is applied science and engineering.
This means there are two fundamental questions in technology: “What can this do?” and “What can do this?”. The first implies that you have a technology in mind and want to know how to apply it. The second implies you’re trying to build something (you have an application), have a problem
1 See Office of National Statistics, Statistical Bulletin: Gross Domestic Expenditure on Research and Development, UK: 2014 (2016), http://www.ons.gov.uk/economy/ governmentpublicsectorandtaxes/researchanddevelopmentexpenditure/bulletins/ ukgrossdomesticexpenditureonresearchanddevelopment/2014, accessed July 6, 2018.
Explaining the Future: How to Research, Analyze, and Report on Emerging Technologies. Sunny Bains © Sunny Bains 2019. Published in 2019 by Oxford University Press
DOI: 10.1093/oso/9780198822820.001.0001
to solve, and are trying to find the right technology to achieve that. These two questions define two different perspectives, both of which are important and both of which we’ll consider in the upcoming chapters.
We’ll start by considering the case of one specific technology and what it can do.
Question 1: What’s so special about this technology?
Some person, laboratory, or company has developed a new process, widget, algorithm, or material and is making great claims about its potential to change the world. Your job is to evaluate these claims, which means being skeptical. What is different about this approach compared to all the others out there (in commercial terms, what is its unique selling point?)? What can it do? What does it make possible?
Let’s make this more concrete with an example. A company has developed a program that turns your webcam into an eye tracker that can figure out what you’re looking at on the screen and the emotion you’re displaying while looking in a particular direction. What’s so special about that? There could be all sorts of answers to this question. Maybe it can be used to catalog your likes and dislikes for advertising purposes. Maybe it can help diagnose mental illness. Maybe it can add a new dimension to gaming. Whatever it is, the answer will be some kind of claim that you can then take time to research and verify. Most claims fall into one of three categories.
The most obvious, simplest assertion inventors might make is that their invention’s performance is better than the competition in some way: the algorithm is better than other similar systems that already exist on the market because it’s faster, less processer intensive, or more accurate, or it works with less-controlled input. This claim may be strong or weak depending on whether it is unlimited or comes with a lot of caveats. Saying an eye tracker is the fastest ever made is very different than saying it’s the fastest of its kind, because the latter requires that you read the small print to see what “kind” that is.
Alternatively, the company might say that the new program represents an integrated solution created using existing but state-of-the-art technologies to address a specific new application: that of finding out how people react emotionally to advertisements. None of the individual elements may be exceptional in themselves but together they form a system that is better
suited to this task than anything else ever built. Alternatively, it may simply be the first attempt to address this exact application at all. Either way, this claim is only strong if the application is really interesting. It also begs the question, is it better suited to the task than other systems that could be built? Just because something is the first of its kind doesn’t mean it’s the best.
The third claim, which is generally only made if a technology is really new and different, is that it is potentially disruptive of some technology space. This wouldn’t necessarily apply to our eye tracker but might to a new material with some unique properties, for instance. Very-hightemperature superconductors (which allow electricity to flow with almost no resistance) would fall into this category. They could allow the creation of all kinds of devices that would have been impossible to build up to now and would be qualitatively different than those we use today.
A less dramatic, but more common, claim is that a new process or system is an enabling technology: that is, it makes possible further development in a related field. For instance, if you were able to produce a more efficient, scalable purification process, you might enable the manufacture of new drugs that would otherwise still contain toxic byproducts. Enabling technologies are extremely important but, like any other, they have to be evaluated.
Finally, there are new approaches to technical fields that have the potential to completely reframe how we think about its development. In artificial intelligence, for instance, the idea of bottom-up learning based on experience (making sense of information coming directly from sensors) was for decades seen as being much less important than knowledge-based expert systems programmed by humans. Over the last ten to fifteen years, this has changed, and what we now call “deep learning” has provided us with a completely new way of looking at the subject. (We’ll be discussing deep learning a lot more in the case study.) Systematization of a field can often have this effect (think of the periodic table), as can computer design tools that allow us to create circuits, lenses, mechanical parts, and chemicals, without the skill and number crunching that was previously required.
If you are an inventor, finding out what claim(s) you can legitimately make for a new technology can be critical for getting funding, getting a job, or getting publicity. However, if your job is to evaluate the potential of research, identifying whether a claim is well founded is just the first step. The next step is to assess whether it is relevant to the task in hand: the application.
Question 2: What problem are you trying to solve?
Often, with the claims we’ve discussed, there is an application implied: an area where there is an existing problem that could be solved with the new technology. For instance, if you take a performance claim, some company might say they’ve developed the most fuel-efficient engine yet and imply that it will make cars of the future more eco-friendly.
Validating this claim doesn’t involve redoing the efficiency measurements. Although there are cases of scientific fraud that would make this necessary, it’s not usually an issue. The problem is that this one measure of performance does not tell you everything you need to know about whether a development is likely to have a positive impact.
For instance, what if the high-efficiency performance only occurs for steady highway driving, and the engine is actually less efficient than existing cars in the city? What if the engine requires new fuel additives, isn’t compatible with current car design practices, or emits particularly toxic pollutants? What if the engine design requires a lot of a very expensive and/or scarce mineral?
These are not rhetorical questions: just because the engine has some limitations or disadvantages doesn’t mean that it is useless. There may be applications where it’s by far the best option. But, without understanding the application requirements, there’s no way to answer these questions meaningfully.
Another example to consider is as follows: a group is claiming that they’ve created a computer using microelectromechanical logic gates. This kind of logic gate is much less efficient than that used in standard electronics: it is slower and heavier and takes up a lot more space on a chip. Oh, and it’s much more expensive. Useless, right? Perhaps not, if you’re trying to build some kind of failsafe device for a nuclear power station. Conventional electronics can be very easily (falsely) switched when exposed to radiation, and this makes them unreliable in any sort of disaster. Mechanical switches, thanks to the fact that they’re less efficient (require a lot more energy to switch) are more robust and therefore better for this kind of backup system.
Bugs really can be features, and features really can be bugs. You only know for sure if you get into the detail.
Sometimes the problem is that the applications of a piece of research are many and varied (a good problem to have if you are looking for work with impact). In this case, the solution is to choose an example application (or a few) to work through that will help you make a case for the likelihood of success. When you get to the reporting stage, these examples will also make it much easier for you to persuade others of the value of the new technology.
Technical requirements
We usually start with the technical requirements because, if these can’t be met, the rest don’t matter very much. To get to grips with these, you need to get to know the subject by exploring and interrogating various sources and understanding their perspectives. We’ll get to these later. For now we’ll focus on the questions to ask.
A technical requirement can be almost anything that—if the new technology doesn’t meet it—could prevent it from working, or working well enough to be at all useful. One set of requirements would come from thinking of the technology as a process with inputs and outputs: so, the webcam algorithm we discussed earlier takes in image data and produces an output that describes both where on the screen the user was looking and what their emotional state was. The purification system takes in a drug in a given form with a given proportion of impurities and then pushes it out with those impurities reduced. Understanding what some system needs to ingest and what it needs to spit out is critical to working out whether it can do the job.
Another set of technical requirements relate to performance: how quickly, efficiently, quietly, or accurately the process needs to be carried out in order for a solution to be acceptable. The number of performance measures that can be applied are as varied as the number of applications served and the number of fields from which the technologies derive. Many require deep subject knowledge to understand exactly why they are important or even what they mean. Fortunately, that’s what experts are for (we’ll discuss them at length in Chapter 2).
Physical constraints such as size, weight, and power (aka SWaP) can be important in many applications. Whether or not a technology is viable may depend on its size, its shape, its weight, its strength, or the number of degrees of freedom in which it can move, and—although this may seem
obvious—these everyday concepts mask a multitude of much more technical ideas. Strength doesn’t mean anything per se: it’s tensile strength, elasticity, hardness, and so on that matters. Which, precisely, are important for a given application, and which are not, are what you have to determine in your research.
Yet another issue to consider is what operating conditions a piece of technology will have to endure, and its related working lifetime. If a circuit board needs to be kept below room temperature to operate, then you’re unlikely to be able to install it inside a PC. If a building material dissolves in acid rain, then—whatever other great properties it has— you’re not going to want to use it in your roof tiles.
That these requirements exist may seem obvious and, when you’re in the thick of trying to design a system, you’ll be acutely aware of what each component or subsystem needs to do and/or withstand in order to fulfill its role. Engineers won’t choose products or use processes that don’t meet their exacting specifications when they are putting their own projects together.
However, this doesn’t help us when we’re trying to evaluate technologies that don’t exist on the market yet. Essentially, we have to second-guess what the requirements will be. Where one product is straightforwardly just a replacement for another, it’s easy: the requirements already exist. But some requirements are not defined explicitly. Doesn’t dissolve in the rain is a key property of many building materials: so much so that we might not have thought to write it down. However, the creators of an advanced composite with many sophisticated properties might have overlooked this issue if they were focused on indoor applications. As the one evaluating whether the material could be useful for construction, it would be up to you to make sure that waterproof was on your list.
Identifying the technical requirements for applications that don’t even exist yet is even more difficult. It is possible to do this, however, as long as you are willing to deal with some level of uncertainty. By making educated guesses (or asking others to) about how the new application is likely to work, drawing parallels with similar existing applications, and taking subsets of their requirements that seem relevant and then thinking through their potential points of failure, you can come up with a workable specification. This won’t be definitive, but what it will do is give you some idea of what is critical for success.
Ethical and legal requirements
Technology doesn’t exist in a vacuum. Any discussion of what will make an application work technically should be followed (if not preceded!) by a discussion of how it will work ethically and/or legally. Take the eye tracker that detects our emotions, again. This was a real project intended to help online platforms give useful feedback to their advertisers. The question for me would be, who would want this used on them? It’s bad enough knowing that free services have access to all the data we give them deliberately, but think of all the data we could inadvertently be giving away if they had access to our webcams, from our taste in sexual partners to our adherence to certain political philosophies.
Of course, you can think about ways to mitigate this: it’s off by default, and you have to turn it on, for instance, and maybe people are given some kind of inducement to use it. But what if this makes the whole enterprise nonviable?
Likewise, what if you produce a purification system that produces sludge so toxic that there is no way to dispose of it legally? Or, maybe you can in some places, but at great risk to those living nearby.
Ethical considerations cover a huge range of different issues. Privacy is an important one, because it is covered by legislation. Just because you are in a position to gather data does not mean you are allowed to store it, process it, and use it for your own ends. People (in some parts of the world) have the right to know what information is being held about them so that, if necessary, they can challenge it, correct it, or delete it. A technology aimed at some future application that doesn’t take this into account is—at the very least—likely to encounter some surprises in its development.
There is also a lot of law around health and safety, and this can apply to every stage of a project. For example, in most of the industrialized world, there are laws requiring your place of work to take appropriate measures to protect you from getting work-related repetitive strain injury (which is often—but not exclusively—related to typing). In addition, there are other laws related to breathing in fumes or working too long at the factory where components are made, lifting boxes of parts onto trucks, developing eye strain where the components are being assembled, and even using the product once it’s been sold. If the application is to succeed, its component technologies will have to be fabricated in accordance with these rules.
A related concept here is exploitation. If an application requires a lot of labor but is not expected to bring in much money, then the temptation may be to pay the workers as little as they will accept to get the job done. In some cases, this may be so little that either they cannot afford a reasonable standard of living or they cannot afford to live without working an unreasonable number of hours. Both of these are unethical, and, in some countries, they are also illegal.
Environmental concerns are also important to consider. There are obvious issues, like the potential toxic sludge of our purification process, but also many others that are more subtle. Life-cycle analysis is a tool that considers the impact of a product (or potential product), from the mining and shipping of the raw materials, to the energy used and pollution created during production, to the disposal of the final artifact once it has broken or become obsolete. For some industries, especially those related to nuclear or fossil-fuel-based energy, environmental concerns—and the laws put in place to address them—have had a huge impact on long-term viability.
There are fewer legal sanctions related to the impact on society of new technologies. This is partly because they can be difficult to prove, partly because society can be harmed without individuals feeling they’re being harmed, and partly because some of the responsibility for harm has to be taken by individuals themselves. Arguments are made about supermarkets causing us to waste food, video games causing us to waste time or making us violent, and bad architecture preventing us from knowing our neighbors. In the short term, such arguments may not matter very much to investors. In the long term, doing the right thing often pays off.
Unethical behavior is not always punished, but both the law and political scrutiny are having an increasing effect on how businesses run. In the UK, for instance, corruption (aka bribery) has been illegal for some time. Businesses not only have to behave ethically and conform to the law themselves but are expected to make sure all the companies in their supply chain are doing the same. The most famous examples of such issues in recent years have been related to the use of conflict minerals used in the electronics industry (mined by people enslaved by Congolese militias) and the issues surrounding the working conditions of Foxconn employees making iPhones for Apple. In the former case, specific legislation was passed in the US requiring the tracking of materials from their origin to make sure that conflict minerals did not end up in consumer products. In the latter case, Apple was forced to at least pay lip service to exhibiting better corporate responsibility.
Another critical area to consider is the regulatory framework for the application area and geographical market you are planning to go into. If you are working on medical devices, for instance, it can take years of clinical trials in the lab, in animals, and in people to get permission to sell your product. Even then, you will likely be limited in what you can sell your product for. For example, even if your new insulin injector turns out to be useful for other drugs, if your trials only tested it with insulin, then it is quite likely that you will have to do a lot more testing before you are allowed to address your potential wider market.
This points to a problem that many people have in analyzing and positioning technologies: they think too narrowly about what the particular tech in question does, can do, and will have to compete with. We’ll come back to that again later.
Related to regulatory frameworks are issues to do with security and export control. Often, these issues have to do with what are known as dualuse technologies (those that can be used both for conventional commercial purposes and for weapons, surveillance, or secure communication). For instance, there was an infamous case where the Apple G4 personal computer (released in 1999) was briefly classed as a supercomputer and therefore could not be exported to some countries that the US government deemed unfriendly (what Apple lost in sales, they seemed to more than make up in good publicity!) Another example from the 1990s was the case of three lines of Perl code (related to cryptography) classified by ITAR (International Traffic in Arms Regulations) as munitions . . . and then printed on T-shirts.
Dual-use technologies can be anything from centrifuges, chemicals, minerals, pipes, computers, code, machines that can be used to build other machines, and so on—almost anything. So, before you decide that your tech has a perfect application in another country, you need to be clear that you will legally be able to address that market.
Do all these issues matter for the application(s) that you care about? Unlikely. Do any of them matter? It would be surprising if there were not one issue among all these that could represent an important obstacle to progress if it were not specifically addressed.
Commercial requirements
If you thought that we missed some important issues in the technical section, there’s a reason for that. Some important requirements look technical but are, in fact, commercial.
Compatibility is a good example of this. An application might be technically possible without being compatible with anything (a lot of prototypes are like this). However, selling a new technology often involves it being able to work—at least in the short term—with equipment that people already have. In fact, this factor alone could help a technology beat its much-better-performing rival. What compatibility means depends on the context: it could mean making sure your widget has the right connectors, your process uses the right chemicals, or your data is in the right format. It doesn’t really matter as long as you know in advance what’s required and you understand that you may have to address your market segment by segment if the compatibility issues are different for each one.
There is a whole constellation of requirements that are connected with the issue of cost. If the technology is going to be used as part of a product or service that is going to have to meet a certain price point when it sold (whether that’s to consumers or business), this can affect everything from manufacturability (how easy and cheap it is to make), to the materials used, to overheads like server costs for cloud-based services connected to products, to any labor involved, and so on. All of these issues are completely dependent on the application, and all of them have a direct impact on the type of technology that can succeed.
And then there are the users and what they want and need. So many would-be products start as a great idea in somebody’s head—and end up in bankruptcy—because the great idea doesn’t appeal to or solve the problem of those who are expected to pay for it. Whole books are written on the subject of business models and product design, both of which are far beyond the scope of this one. But, at the very least, a reality check is warranted. That involves going out and talking to people. We’ll talk about how to do that in the Chapter 2.
Potential obstacles
With dozens of potential requirements, it may sound extremely timeconsuming to figure out whether a new technology will be derailed by one of them or not. This might be fine for an investor or CTO, but not for a researcher or student.
However, just a little research can take you a long way. Let’s say you’ve been tasked to find out (quickly!) whether a new robotic technology— developed for use in car manufacturing plants—could also be used as robotic toys, as the company claims. Part of the work is done for you
because you know the technology is already in use in industry, so you can figure out what it can successfully do by looking at that example. Now that you understand what it does, do you really believe that it will work in toys? A five-minute discussion with a toy manufacturer will probably lead you to understand that there are two things that matter in that business: safety (legal/ethical requirement) and cost (commercial requirement).
If you know enough about your application, you’ll see immediately the kinds of questions you need to ask. Will the behavior of the robot be as predictable in the chaotic environment of a home filled with children and pets as it is in a factory? If the robot malfunctions, is it strong enough to hurt anyone? How much does the processor (computer) cost that controls the robot? What about the actuators that make it move?
Where you do find problems, it may or may not be possible to work around them. It could be that an application is not feasible right now because it cannot support the cost of the computing power it needs. By waiting a couple of years for technology to naturally evolve and reduce in price, that barrier may disappear. On the other side, there could be an intrinsic technical problem that requires a major research push to fix: these issues will be important to keep track of as you do your research and analysis.
If you don’t know enough about your application, and you don’t ask the right questions, you may be so blown away by the company’s demo that you forget to see whether the technology is really fit for the new purpose.
No problem for the solution? Be creative . . .
Sometimes a technology is just cool. It isn’t particularly designed to do anything, but it seems to have some features that make it special. It’s easy to get excited about such technologies and then quickly lose interest due to a lack of any obvious application.
There are two things to do when this happens. The first is to try to look for the nonobvious application. You do this by going up the layers of abstraction. What’s special about this material? It’s expensive but really strong. Where does strength matter but not cost? Could the material be used to trade off strength for quantity (less material for the same strength?). Think of as many different scenarios as possible in which the particular features of the tech can be exploited.
Sometimes this will get you to applications, and sometimes to a set of higher-level features, that is, features that people outside your immediate technical area can understand without knowing too much about the
