Pages

Tuesday, November 11, 2014

The Era of Artificial Intelligent Assistants: What can we learn?

When we look in the area of artificial intelligent assistants, a few big names come to mind: Apple Siri, Microsoft Cortana, Google Now, and even IBM Watson. Each service is very different from each of its peers and aims to solve different types of problems. While all of the aforementioned names, as well as the smaller ones that are still trying to make it big, offer their unique selling proposition, the aim of this article is to determine which problem offers the most opportunity. Said differently, I inspect what an ideal intelligent assistant should be capable of and then reverse engineer the problem to see which service is closest (at the moment) to solving this need.

What Do The People Want?



It is true that everyone is going to want something different. Some folks want app integration, so they can speak a command into their phone and have already downloaded apps perform their wishes. Other people want information access beyond just a web search. There will always be a group of individuals, albeit a small one, that just want an assistant that can do something 'unexpected', such as show some humor in a response.

App Integration: Life Made Easy

While the majority of people will fall into the category of wanting an assistant who can help make their daily life easier, I suspect that this offers the lowest amount of opportunity. Sure, if you speak to an intelligent assistant and that assistant can either put something on your calendar, type a text message out for you, or read an email, then you saved some key strokes. 



However, that is not far from the best that this kind of service will be capable of. It might be convenient in times when the eyes are busy, but the ears are not (i.e. driving), but the matter of the fact is that people are currently open to putting up with pressing a few extra buttons on their phone to get something right versus talking to an assistant who has a track record of messing it up. Intelligent assistants can offer more than saving us a few seconds or making life mildly more convenient.

Decision Making Backed by Information

Another popular use case for intelligent assistants is to find information. In this case, speed is critical because in most cases, people want to fact check another party and must do this quickly to avoid the conversation stalling or moving on a tangent. The current search process either involves a lot of steps and physical typing if performed manually, or the intelligent assistants can help, but are limited. There is some real opportunity here. 

In an ideal world, when person A states a fact to uphold their opinion, person B should be able to fact check this statement quickly. Current solutions involve accurate self web searches but require more time or leveraging intelligent assistants which means that unsophisticated techniques are used to key word search. 

However, a perfect solution would be one that has the following attributes:
1. Everyone has access to it
2. Returns a single answer if there is one
3. If not, returns a confidence level with a backed source for answers that might have more than one correct answer
4. Additional guidance to make an informed decision



We must let the facts speak. Not the ego. In the end, users want an intelligent assistant that can help them make decisions backed by facts.

The Golden Solution: User Awareness

The perfect solution would consist of a combination of the above into a single entity. This 'perfect' assistant should be capable of administrative tasks via app integration as well as reliable information access with a confidence rating. However, one difference would be to make the intelligent assistant aware of the user's environment

Some apps have already started to make steps in this direction. An example is to, "remind me to wish my dad happy birthday next time I call him". The system has to be aware when you call your dad next. This ideal system would be familiar with the user's past and their current surroundings. 

Another great example of this would be to ask "where can I get some food?" to the assistant. The assistant could know that you don't have a car and must walk, what you ate previously to avoid those recommendations, and even what you normally like to eat. The examples are endless. 



The perfect intelligent assistant will take care of the 'situational awareness' for us. In the end, the system that will win will be one that builds on the app integration and information access assistants that already exist. The true winner will be one that knows the user down to the last DNA strand to help the user make the best decision at the right time.

Please add your thoughts, opinions, comments, and any other topics for me to look into below.