Predictions on the Evolution of Intelligent Assistants

NEural networkPeter Sweeney, entrepreneur, technologist, and founder of Primal.com, recently published an article on Medium called Siri’s Descendants: How Intelligent Assistants Will Evolve. In the article, Sweeney does a great job at contrasting how intelligent assistants are being positioned today versus what types of use cases he believes IAs are better suited for in the future.

Sweeney uses some very imaginative and compelling graphics to depict his predicted evolution of IAs from today to tomorrow. The key areas that Sweeney sees as ripe for evolution are:

  • Natural language interfaces vs. contextual / associative communication
  • IAs as generalists vs. IAs as specialists and domain experts
  • IAs as isolated with limited autonomy vs. IAs as networked with expansive autonomy

On the question of how we interface with IAs, Sweeney believes the preponderance of today’s IAs require the user to interact with them via natural language. But Sweeney points out that you can learn much more about people, their habits and needs, by silently observing their behaviors in context. Though he doesn’t provide these examples, Sweeney’s projections for the future would give the edge to context aware assistants that work in the background, such as Google Now and EasilyDo, over voice-prompted assistants like Siri and SpeakToit.

In the area of what the IA knows and can do, Sweeney’s graphic shows that today’s IAs are overwhelmingly focused on trying to cover very broad topic areas. You can ask Siri anything and she can give you some sort of answer, even if it’s only pulling up distantly connected search hits. But Sweeney thinks the sweet spot for IAs in the future will be specialization, with IAs having in depth knowledge of targeted domains. This specific knowledge will enable assistants to execute the tasks that we want them to perform for us. Sweeney calls today’s IAs “trivia buffs.” He sees a huge opportunity for IAs that have enough insight to carry out tasks, such as helping us navigate to work. Another example that I’ve been thinking of lately: helping us plan our meals out for a week, including putting together recipes that we’ll like and that meet our dietary goals, and automatically drafting the associated shopping list. That sort of help sure would save time.

Lastly, Sweeney sees today’s IAs as being too solitary. They are locked inside our smartphones. In the future, specialized IAs could exist within websites or connected devices and they could communicate with each other to perform higher level services for us. Since each IA is focused on a narrow domain, Sweeney suggests that we might be comfortable allowing them more autonomy. An example I have might be our willingness to permit a book shopping IA to choose and purchase (up to a dollar threshold) books it has a high confidence level we’ll enjoy. And it could even know to make the purchase only after we’ve run out of good things to read.

Sweeney makes some insightful points in his article and his supporting graphics are great. As always, it’ll be interesting to watch how the marketplace evolves. We’ll see if IAs move in the direction that Sweeney predicts.

One thought on “Predictions on the Evolution of Intelligent Assistants

  1. Thx for the summary, Amy. My intent was really just to highlight some underserved areas within the market landscape, where new product forms might emerge. The only things I can predict relate to the product I’m collaborating on at Primal. Existing assistants are far from trivial and will continue to develop in fascinating ways within their own turf.

Share your thoughts on this topic

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s