AI, Interaction and Business Models
How disruptive is AI going to be?
Bill Gates once said ‘we always overestimate the change that will occur in the next two years and underestimate the change that will occur in the next ten’. This quote rings particularly true in the case of AI technology.
The moment when mainstream technology consumers opened their eyes to AI technology was the moment Apple announced Siri, in conjunction with the iPhone 4S. All of a sudden, the fantastical notion of talking to a computer was not just one that appeared real, but was one consumers could buy. The history of AI, like flying cars is one filled with false promises and misguided optimism, Siri appeared to break the mould. But then she didn’t.
Since Siri’s mainstream release five years ago, our lives have changed little, our interactions with our computing devices is very much the same as it has been since the first iPhone came out - done primarily through the graphical interfaces that our operating systems have come so close to perfecting. Google’s Material Design philosophy that pervades Android is insanely great, a three syllable remark more often reserved for Apple. As Ben Evans astutely notes, technology on its deathbed is technology at its very best — the best is the last. It’s hard to conceive a rendition of the graphical user interface that is significantly better than it is today, we’ve reached the point of diminishing returns — the point of ‘good enough’. Our computers are fast enough, our phones are easy to use enough, our animations are fluid enough. The time might be right to bet on AI making a move.
And sure enough, the AI conversation in 2016 is easily the loudest in the room. AI is particularly interesting because it augurs a completely new form of interaction, and if you’re not afraid of perhaps prematurely inflating its potential impact — AI mobilises the third big wave in computing interaction. We started off with the CLI (Command-Line Interface), we’re currently head and shoulders deep in the GUI (Graphical User Interface) paradigm, and with the help of AI, we are slowly moving into the CUI (Conversational User Interface).
These three modes of interaction all have their own definitions for what’s important. What are the characteristics that make a great graphical user interface? What matters to consumers? It’s important to ask these questions for each of the interaction modes because not only does it help build great experiences, but also is an accurate determinant of which businesses have the right bundle of competencies to succeed in the respective spaces. To first examine the GUI, the characteristics that make for compelling experiences is fluidity and ease of use. It’s no surprise that Apple’s religious design focus has enabled it to succeed in the era of GUI, particularly in the case of mobile in which ease of use is paramount on such miniaturised computing devices. Obviously, Apple’s success owes itself to a host of other factors: network effects and ecosystem dependence the pilotis holding the Apple empire aloft. However, Apple’s tasteful and fluid design are the factors from which its competitive advantages stem.
So what about the CUI? The Conversational User Interface is a wholly different beast to the GUI and CLI in the sense that it elevates the computer out of the passive realm and transforms computers into active companions. Instead of users navigating devices in order to complete tasks and extract information, the CUI will be more pre-emptive and assertive, providing suggestions and giving answers. Computers won’t just be communicated to, they will be communicated with.
The CUI manifests itself in many ways. One could perhaps discount Google’s ‘Google Now’ as a strict example of a Conversational UI. But it’s competence in context-awareness and conveying information pre-emptively allows it to fit squarely in this paradigm. In fact Google Now remains a deeply disciplined exercise in self-restraint. For all of the company’s proficiencies in machine learning, the company understands that the technology is simply too raw to be truly useful in natural conversation. When using Siri, Cortana or Alexa the gaps are clear, particularly when the assistants simply revert to generic web searches or dodge queries through witty pre-determined responses. As such, Google Now only gives you what it can, eliminating the possibility of humiliating itself by attempting requests it can’t perform. The CUI therefore can be a much more passive incarnation as Google has proven, or a pulled and active voice assistant such as Siri. It can also be implemented as chat bots like the one recently announced for Facebook Messenger, or the ones already available on apps such as WeChat and Line. Whatever it is, they all share one thing in common — they’re heavily personalised and reverse decades of preconceptions around technology use. We no longer have to learn how to use our technology, the technology learns us.
So in the same way that ease of use remains the most important factor in a great GUI, intelligence is easily the most important factor in a great CUI. Intelligence is how much an AI knows about its users and the topics that users are querying. Communicating with an AI has to be useful, or users will simply revert to traditional methods of device interaction — the graphical interface, using apps and browsers in the same way that we always have. In fact these very contingencies are built into current personal assistant technologies. When Siri can’t solve a problem, she pushes you back into the browser, and AIs still aren’t capable of recommending exactly what users need, instead providing users with lists of options – for example a list of Japanese restaurants downtown, instead of the Japanese restaurant perfectly suited to the context, with a table for 2 that just opened up.
With this example, it’s easy to see how far AI technology has to go in order to be the be-all and end-all arbiter of our requests. Development of AI technology is heavily dependent on access and interpretation of all kinds of information, allowing it to extrapolate conclusions and reach levels of ‘intelligence’ that have otherwise been the stuff of science fiction. In the context of today’s technology landscape, the shift in priority – from UI fluidity to intelligence – is incredibly disruptive, and plays to the advantage of two internet behemoths: Google and Facebook.
In the era of data, none are better placed than Google and Facebook. Unsurprisingly, these two companies also happen to be the two most successful examples of the dot-com advertising business model. They’ve done it better than any one else: one through harvesting enormous masses of public knowledge, the other through the deep learning of people. In the context of AI, what’s more important, information or identities? It’ll be fascinating to see the way in which this plays out. Other players such as Microsoft and Apple are also making incredible strides in aspects of machine learning as well, positioning them well to participate in the AI transition.
Probably the biggest business model impact of AI is how it undermines the utility of individual apps, and this is easily the most disruptive factor, particularly from Apple’s personal vantage. Despite the revolutionary nature of mobile apps, they are still often a very stunted way of interacting with data. For many activities apps are entirely necessary, for example in social media which involve constant and high levels of interaction. But for more fringe applications such as event ticketing or airline bookings which are used sparingly, the friction associated with needing to install an app is arguably too great. There are better ways for businesses to communicate, something that AI and the CUI have enabled us to realise.
The moniker ‘smartphone’ is contentious, given most are designed as a simple grid of functions with very poor interoperability and zero contextual awareness — a limitation propelled by the need for ease of use in UI design. Both Google and Microsoft have attempted to overcome these flaws through implementation of widgets and live tiles in their respective platforms, allowing certain functions to operate in the fore without being proactively summoned. AI is poised to take full advantage of the limitations inherent in the traditional pull-based graphical interface, with AI apps such as Google Now aggregating the functions of multiple minimal-interaction apps — such as public transport, traffic, weather, news, sport, post— into a single place, whilst constantly learning and surfacing data that matters to us. It’s AI assistants such as Google Now, that put the ‘smart’ back in smartphone.
Therefore, in undermining the utility of many existing applications, AI cuts deep into Apple’s ecosystem competitive advantage. AI offers way for businesses to reach customers without forcing them to undergo the cumbersome processing of installing an app, let alone the costs on the business’s end of developing such an app. AI interfaces and chat bots are a much more efficient avenue for minimum-interaction type applications. As described earlier, as many businesses move away from traditional app development, not only does it weaken the need in many scenarios to develop good user interfaces that many businesses and ecosystems have spent years developing, but it also lowers ecosystem barriers that have limited user mobility in recent years. Delivering a service through a cross-platform AI mechanism or chat bot is inherently less costly for a business than delivering that service through a multitude of different platform-specific apps. For consumers, being able to access services in a greater variety of places significantly reduces switching costs between platforms. If AI technology does develop in this manner, then it will reinstate the democratisation of the web back into mobile. We should be expecting a lot of change in the next 10 years, maybe more than we can imagine.
