Why A.I. assistants need to stay neutral

With the release of Google Allo last week, we’ve officially entered the era of the assistant. Every company that owns a major tech platform is now betting that assistants will be an important interface in the post-mobile world. 

You might think assistants today are trivial or stupid, and you wouldn’t be wrong. They often don’t understand what we say and can’t hold real conversations yet. Mostly they’re relegated to simple tasks like playing music, sending texts, or setting timers.

The people building the assistants of a not-too-distant tomorrow see it differently. The next generation, of which the Google assistant that ships with Allo is one, have a grander ambition: to replace search and mobile apps as we know them. They are designed to provide information and guidance to everyday questions, to hold longer conversations, and even make complex decisions on our behalf.

Looking only at how assistants work today ignores the nature of how A.I. works. It improves at an exponential scale. Look at self-driving cars: A technology that seemed impossible for many years is already shuttling passengers in Pittsburgh. Assistants that influence our everyday decision-making may seem far off, but they will come upon us suddenly.

It’s critical that we start to think about how next-generation assistants decide what information they share with us now, before it’s decided for us. How will they determine which sources to consult when choosing recommended flights, hospitals, medical treatments, TV shows, or Chinese food delivery options? Who owns those data sources and decision processes?

When I ask my assistant a question or give it a task, I want to know it’s going to give me the best answer it possibly can, no matter where it comes from. I don’t want it to be biased towards some other corporate goal. If we wait until assistants are performing daily tasks and functions for the mass consumer, it will be too late. Our information sources will have been pre-determined, without our consent.

As of today, the situation doesn’t look good. The assistants offered by major companies today come with significant bias because they exist primarily to promote another business interest. For example, when you ask the Google Assistant questions, it always leads you to one of Google’s properties for answers. If Google thinks one of their products, like Google Places or Knowledge Graph, can provide an answer, you get that. Otherwise, you get a single blue link from web search (the ultimate “I’m feeling lucky!”). There are no third parties directly informing Google’s assistant.

Yelp and other important sources of reviews and information have complained for years of Google crowding out web results in favor of their own products on web search. The assistant exacerbates this problem, eliminating other voices altogether. It makes sense if you are building an assistant to drive traffic to your own products, but it’s terrible for consumers, who want (and deserve!) the best answer, and developers, who need (and deserve!) non-competitive distribution partners. 

What about the independents? Independent assistants aren’t focused on driving another business through their advice and recommendations; they are focused leveraging technology to provide the best possible results and guidance to the consumer. They live or die based on this promise. But being a true multi-domain knowledge-based assistant is a highly specialized and expensive task.  Few teams are poised to pull it off — and many of them are being gobbled up. Just last week, Api.ai, one of the granddaddies in the assistant space, was bought (by Google).

A neutral assistant has the potential to be an incredibly valuable business for consumers and publishers/developers alike. Far more valuable than the limited source versions we see today. It’s perhaps inevitable, with developers desperately seeking an efficient discovery platform, that an independent assistant interface will become the search engine of the future. Access to unbiased, quality content may well depend on it.

The era of intelligent assistants is upon us. As the A.I. rapidly improves, assistants will increasingly replace search and apps as the starting point for any decisions or tasks we want to perform. That’s why we need to start talking about assistant neutrality now.

Humans make the best decisions when they can take into account many different points to view. We need our assistants to do the same.