Personal vs. Communal Artificial Intelligence

Piggybacking on Carolina’s post this morning, I want to share my thesis on how I think about communal vs. individual machine learning.

I feel it is a flawed assumption that communal machine learning is the only value proposition for the development of AI. To understand the personal vs. public side of this, we need to understand the word “training”. Everyone doing machine learning has to train their network. To do so, they use big data sets that give the network enough information to accurately identify even the most nuanced of objects. For example, in natural language, it may take tens of millions of instances of a particular phrase in a particular language for the machine to be able to understand all the nuances of a dialect for just that word. Similarly, for a machine to accurately identify a tree, it may need tens of millions of images of a tree in order to be able to identify or distinguish it from a bush, for example. These massive data sets are the basis of large communal deep learning initiatives and they are the basis of how a network is trained.

These big data sets are useful in training a network for both specific things but also in identifying crowd patterns as well. A simple example of this in Google search is when you fill in a partial search query and the text field tries to guess or assist with your search so you don’t have to finish the word.

Screen Shot 2016-06-22 at 7.29.51 AM

It appears there are a few Bens more popular than me at the moment. 😀

These things are convenient and the machine is trained to recognize patterns. The goal is to make life easier and more efficient for the customer. Now, where I feel a schism will occur is between a large data set of crowdsourced trends, general information like a tree or dog, and data that is unique/personal/intimately related specifically to me.

Personal machine learning has yet to happen. In the same sense that large data sets will train a network based on communal trends and data, I have yet to train my own personal AI on the uniqueness that is me and my life. There is a battle looming for the “personal life assistant” consumers will hire to help them at an individual level and be smart and predictive for things unique to their lives. This is where the technology that helps you train your own personal digital assistant will be key. And this is where I think the “on device machine learning” coupled with privacy is going to keep Apple competitive if not even give Apple an edge against others who may seek to do this.

It is hard to argue that our smartphone is not the right device to begin to learn about our individual nuances. We can argue the privacy tradeoff but, come this fall, we will be able to judge how well Apple is both learning the from individual and from the community at the same time using the privacy means they have in their toolbox. But when I look at other attempts, it seems only Cortana from Microsoft is trying to become a personal assistant. The Echo and Amazon’s strategy driven by their business model motive is positioning to be more a communal assistant than an individual one. Google’s is as well. There is nothing wrong with this strategy and ultimately I can and will use them all. But what is chosen as my personal life assistant is perhaps the most important battle when it comes to what I engage with the most and build a relationship with.

Furthermore, there is another flawed assumption that these assistants will be mutually exclusive, meaning not talk to each other. I am not sure that is how this is going to play out. If I choose to hire Siri (or whatever I eventually end up naming it once Apple allows us to change its name), then it seems likely my assistant can go and talk to Google or Amazon’s AI on my behalf. Those companies’ business model does not favor an exclusive premise and they will need to make sure their data sets can be used by all.

Many companies doing the big and communal data sets are letting these be accessed by all, if not even licensed for proprietary use. This data will exist and many can benefit from it, including Apple. The harder part of this equation, in my opinion, is the learning of the individual. It is clear how Apple plans to do this, but not clear how others will. This seems of strategic importance, particularly to Apple, but not necessarily to Google or even Amazon. As the saying goes, once you hire a great personal assistant you never let them go.

Published by

Ben Bajarin

Ben Bajarin is a Principal Analyst and the head of primary research at Creative Strategies, Inc - An industry analysis, market intelligence and research firm located in Silicon Valley. His primary focus is consumer technology and market trend research and he is responsible for studying over 30 countries. Full Bio

2 thoughts on “Personal vs. Communal Artificial Intelligence”

  1. When computers can prove mathematical theorems and then come up with new ones, I’ll consider that learning. Anything short of that is training.
    The difference? Training is about finding answers to known problems, learning (education) is about asking the right questions and solving previously unknown problems.

Leave a Reply

Your email address will not be published. Required fields are marked *