Google Now is great. But how do you make it better?
I believe Google is building a software-based artificial human brain, capable of interacting with people convincingly as if it were a real friend.
And Google isn't alone in this quest. Here's what's going on.
The Wavii app, which I use on my iPhone, takes an article and reduces it to a single sentence. That sounds easy, because it's an easy thing for humans to do. But that's really hard for software to do.
For starters, how can software scan words in a story and choose which of those words are the important ones and which are filler or supporting context rather than the main idea?
Second, how can software concisely summarize the main idea of a story in language that sounds natural, rather than clunky computer-speak?
These are computing problems researchers have been working on for decades, and Wavii is an impressive, usable example of that.
Wavii also does some impressive work around content categorization and social sharing.
Google's acquisition should be compared to Yahoo's recent acquisition of Summly, a news summary app that uses something called a "genetic algorithm" to extract meaning from blather. Not only is the technology conceptually similar, the price was identical -- $30 million.
Summly takes articles and summarizes them into 400 characters.
When that acquisition was reported recently, everybody focused on the age of the founder -- 17-year-old Nick D’Aloisio, who started the company and launched the app when he was 15.
The most impressive technology in Summly wasn't created by a kid with acne, but by Inderjeet Mani, a man who has been researching natural language processing for significantly longer than D'Aloisio has been alive, and who has published 92 research papers on the subject.
Summly has some amazing technology, and I expect Yahoo to apply it broadly as an interface for using many of the company's services.
I told you about this product back in September.
The MindMeld app is fundamentally different from Wavii and Summly in what you use it for. Instead of summarizing news and other content into concise, meaningful, natural-language nuggets, MindMeld eavesdrops on your conversations and shows you related information.
But in concept, the technology is doing something similar behind the scenes. It's scanning content, separating out the meaning from the fluff, and using the important information to take action in the same way a person would.
MindMeld does something every human does -- it listens to and "understands" people's conversations, then adds to the conversation relevant information based on what it knows. You and I do this every day.
The difference is that what MindMeld "knows" is: all knowledge on the Internet.
MindMeld offers up search results, for lack of a better term, on topics discussed. The idea is to add context and information about your conversations while you're still talking about it. As you talk about a movie, information about the movie pops up (no more "what-else-was-that-actress-in?" madness -- MindMeld will tell you).