Google‘s Hearts and Minds campaign
At Google I/O 2017, Sundar Pichai’s opening statement laid out an ambitious plan for Google’s future, and the future of computing more broadly:
We’re excited by the shift from a mobile-first to an AI-first world. It is not just about applying machine learning in our products, but it’s radically re-thinking how computing should work… We’re very confident about our approach here because we’re at the forefront of driving the shifts to AI.
It makes sense that Google would bet on AI: Even in its infancy in late 90s, its founders conceptualized it as more than just a search engine. In the two decades of the company’s existence, it has invested a significant part of its immense resources into building a powerful AI ecosystem spanning its entire suite of products. As a result, Google today has a significant lead over its competitors when it comes to AI: its Maps, Translate and Assistant services are widely considered to be much better than the competition.
This incredible ecosystem was monetized through advertising. In 2000, Google launched its AdWords program to allow advertisers to feature their ads among the links returned by a search. AdWords was a self-service platform, which meant Google never needed to invest in a salesforce and could scale massively to match the explosive growth in internet usage at the time. AdWords has since expanded to include video ads on YouTube and display ads on other websites (including Google’s own Gmail). However, search advertising remains the most valuable business because of the nature of search: it’s easy to serve up highly relevant ads when users are volunteering information about what they’re interested in.
Google leverages its AI capabilities and data to accurately target customers with ads they’re likely to click. Since advertisers effectively only pay for conversions (i.e. when the ads gets clicks or impressions), Google is incentivized to provide ads which actually interest the user. In return, advertisers set a budget which determines the extent to which Google promotes their ads. Search page advertising is an auction: advertisers bid for search keywords, and the highest bidders get featured most often and to the most number of users, increasing the likelihood of a successful conversion.
However, it is not guaranteed prosperity in an AI-first world, simply by virtue of being its premier evangelist. In fact, it isn’t really clear how their current ad-based search model translates to this new paradigm. Search in an AI-first world will be radically different and won’t present the advertising opportunities that Google currently leverages with AdWords.
Search in an AI-first world
Here is what a typical Google search looks like:
A search for a product turns up ads for relevant results as images at the top, and as the first few links in the search results. Ads are clearly marked and users that aren’t interested in any of them can scroll down and find the list of links that Google has found organically.
The model works because the amount of real estate Google has on your screen is effectively infinite. It cordons off a part of that real estate (the most valuable bit, right at the top) to show you ads, but you can ignore those ads and scroll down page after page until you find the result you’re looking for. It’s built for a desktop/mobile UI, where most of the interaction with Google is done through a screen. But as Sundar Pichai said, we’re moving away from that paradigm and towards an AI-first UI.
We already have an idea of what that looks like:
AI-first interfaces are essentially conversations, where you ask an AI assistant something and it responds with an answer. In an AI-first UI, Google no longer has the luxury of infinite real estate. Especially with the advent of voice-based UI using devices like Google Home, the space for answering queries is tiny. Nobody wants their voice assistant to spend hours reeling off a list of results. If you ask a question, you are expecting an answer, and moreover the right answer. A list of links is no longer good enough. And if there’s no list of links, there’s no space for Google to insert its ad links. Any ads they try to push to users will have to fit into the conversational paradigm. Ads will be more like suggestions given in a conversation.
Word-of-(artificial)-mouth advertising
Picture a conversation with a friend. You ask them for a good running shoe. They tell you to get the Adidike MegaRunner 6000. Would you buy it? The answer to that probably depends on how much you trust your friend’s judgement. If you knew they’re knowledgeable about running shoes, if they had a track record of suggesting good buys to you, or if you just knew that what works for them works for you, you’d take their suggestion a lot more seriously.
Trust is a valuable commodity in the AI-first world.
In simple terms, trust is a multiplier of the probability of a user accepting search results offered to them: The more users trust a service, the higher the probability of them accepting suggestions from it. Trust is the reason why Google won in the search engine wars of the 90s: users preferred Google over the alternatives because they trusted it to get the best results. However, the importance of trust is magnified in the AI-first world because Google can no longer offer you an endless list of links to scroll through. They can only offer a few answers (and ideally only one) for every query, which means they need to make sure the results they do give you have a high probability of being clicked.
Trust will be the highest barrier to advertising revenue in an AI-first world: getting users to accept product suggestions from an AI assistant is unlikely, and only a small percentage of Google’s total users will probably ever respond positively to any suggestion from an AI. However, once a user does accept a suggestion, there is potential to create a powerful feedback loop if Google gets its suggestions right: users who have had a good experience with an AI assistant’s suggestions will be more likely to trust it for their next search. Over time, Google can leverage the data they gather (actively and passively) to give users an extremely personalized experience.
In other words, if users can be induced to accept suggestions from an AI assistant in the first place, Google finds itself back in familiar territory: using data offered by users and their AI expertise to provide them with the best possible answers to their questions.
A lesson from South Korean loudspeaker propaganda
How do you induce users to take that first leap and accept suggestions? In other words, how can build up a basic level of trust with people so that they will even consider listening to their assistant at all? This excellent comment on Reddit talks about how South Korea uses loudspeakers on the DMZ to carry out a propaganda campaign against North Korea. They start with broadcasting simple information to make life convenient: the weather, or when it’s going to rain. It eventually ramps up to forbidden music, world news, and finally information about the North Korean regime.
What’s interesting is that this is essentially an exercise in building trust using better quality, actionable and verifiable information. Weather updates, in particular, are a great way to build trust: everyone is affected by it, it’s freely available and easily verifiable information, and it is easy to create habits and build relationships with users by telling them about it.
Google has already started taking its first steps towards building relationships with users with its Now/Feed/Assistant feature. Introduced in 2012, Google Now was designed to be a proactive personal assistant, offering users information without them needing to ask for it. It gives users useful notifications like how long their daily commute will take, what the weather is like, or when to leave to make appointments on time.
Google Assistant doesn’t offer this kind of proactive reporting yet, but the potential here is obvious: a voice notification telling people about the weather, or reminding about their appointments is an easy way to build a relationship and establish a level of trust with them. When this basic level of trust is in place, Google can start using Assistant notifications to serve ads to users. Their AI capabilities will be the keystone of this entire process: targeting the right ads at the right users is critical to keep the feedback loop of increasing trust going.
Google certainly has the technological capabilities to adapt its advertising model going for the AI-first world. They may very well also have a strategy in place to manage this transition. Ads as Assistant notifications aren’t going to eclipse AdWords anytime soon, but as the world moves to smarter devices and AI becomes a more established UI paradigm, Google seems to be prepared for the shift and set to prosper when it comes.
