Join Transform 2021 for the most important themes in enterprise AI & Data. Learn more.
Using the Google Home on a daily basis makes you appreciate how helpful it is. You can ask for directions and find out about the weather. After a while, you realize the Assistant that answers questions can provide a wealth of information, but it’s essentially a duplicate of Google Search. Just imagine how much more valuable the device would become if it could also give advice.
Here’s an example. Today, most searches are fairly linear in nature. You can ask for facts, but even then you have to ask the right questions. Recently, I started asking questions to prepare for a bike trip in my state. I could ask about the population of a town, the distance between two towns, and even what the weather might be like that weekend.
All of this works fine from a search box as well. When I asked questions that were a bit vague, like which bike trail is the best in my state, the bot started to balk.
This is something that also doesn’t work by search, even though it could. There’s a massive treasure trove of collected opinions about bike trails all over the web, including articles by bike enthusiasts and bloggers. There isn’t an AI that is trying to parse this information. For one thing, a bike trail is not an official location on a map, and it’s not a business that users have rated. Asking about the best pie shops in my state did work — that’s based on ratings. Asking which one has the best apple pie didn’t work. That’s based on user opinion.
The bot can’t really parse any complex question, either. When I asked which bike would work best on a gravel trail, the Assistant didn’t know. None of these questions worked, either:
What’s the most scenic bike trail?
What’s the longest?
Which bike trail has the fewest hills?
Which bike trail is closest to me?
The problem with the bot not answering these questions is that many of them are based on opinion. Scenery is a personal preference, but then again, it’s obvious a trail with many lakes and vistas is better than the one that runs along an old rail line. An opinion is possible for a bot to generate. If you read ten articles about bike trails in my state, one particular trail emerges as the most obvious pick for scenery — and a low number of hills. It is fairly obvious what it is considered the best. Why doesn’t the bot know that?
One reason is that the bot on Google Home is not that intelligent yet. It doesn’t really know me, and it doesn’t really know how to give advice. It can tell me to bring a jacket on the trip because of a weather report, but doesn’t go a level deeper and know that the trail I’ve picked is known for inclement weather and wind — especially 20 miles from my origin point. As a voice-enabled version of search, it is helpful. But a true bot needs to parse complex information and provide better advice. It needs to go a few steps further and understand what I’m trying to do, become more proactive, and engage in a discussion with me that is helpful in a way that goes beyond the facts.
I realize advice is hard. Or is it? An AI could “research” topics like bike trails or movies, asking me a series of questions to learn about my tastes and analyzing a vast number of opinions online to develop an adequate viewpoint. “John, I know you will like the movie Baby Driver because I’ve read your car reviews” is what I consider to be helpful AI. In some ways, it does this already with businesses that are rated by users, like the pie shops. It made me wonder, beyond my bike trail questions, if a bot will be capable of giving advice on more topics, like where to spend a vacations or the best places to go stargazing. This is when bots become exceptionally helpful — and indispensable. It’s the tipping point.
I’d like an AI assistant that gives me advice about safe driving routes, the most interesting movies (not just the ones that are rated the best), where to find the best deals on the clothes I like, and where to catch a walleye on the lake in front of my house. In other words, I want relevancy. A true AI assistant would know about me and my tastes, and know how to match the data already out on the web with my individual preferences. It would know how to give advice by correlating various inputs — exactly like the human brain.
So how did I end up finding a bike trail? Really, the steps were a precursor to what a bot will do one day. I read several articles, looked at Google Maps, watched a few videos — I educated myself about the opinions from other users and matched them to my tastes. The bot was totally in the dark on that one.
VentureBeatVentureBeat's mission is to be a digital town square for technical decision-makers to gain knowledge about transformative technology and transact. Our site delivers essential information on data technologies and strategies to guide you as you lead your organizations. We invite you to become a member of our community, to access:
- up-to-date information on the subjects of interest to you
- our newsletters
- gated thought-leader content and discounted access to our prized events, such as Transform
- networking features, and more