Bradley Horowitz, responsible for novel technology development at search giant Yahoo, looks ahead to the "internet of things".
Bradley Horowitz wants to merge the real and virtual worlds
Imagine this scenario: I am in a supermarket and I pick up a can of tomatoes and I place it in the shopping trolley. Immediately my mobile phone flashes green to indicate to me that it is a good buy. I go down the aisle and choose a bottle of wine but this time my phone flashes red to suggest I reconsider.
This is only possible when we have a universal resolver for every entity in the world.
What do I mean by universal resolver?
On the internet we have something called DNS - the Domain Name System.
When I type in yahoo.com there's a service set up in multiple distributed servers around the world which helps "resolve" the mnemonic "yahoo.com" (easy to remember!) to a numerical IP address (hard to remember!) which machines can understand.
The service translates yahoo.com into a specific IP address so I don't get mistakenly sent to another website.
We do that very well for resolving domain names but we don't do it very well in the real world for resolving entities.
What do we mean by entity? Frankly almost anything qualifies: a person, a place or a thing, real world and digital objects, even concepts or ideas.
Amazon has done a very good job on this. We know every book gets issued an International Standard Book Number (ISBN) number. Amazon additionally associates an Amazon Standard Identification Number (ASIN) with every product in their system.
Through the magic of RSS, feed readers can aggregate my digital life as it exists across many silos on the internet
But it shouldn't just be products in Amazon's catalogue (vast though it is). It should be all entities - everything from the particular chair I am sitting on to objects like the Lincoln Memorial monument should have a unique digital identifier.
As an example - let's start with people.
I don't know if email@example.com is the same as Darren_waters@gmail.com.
There is a problem of managing identity across the internet, so when I say Darren Waters I mean this person and all of the manifestations and representations and personas of that person. The ability to knit those together is a huge challenge and opportunity for us as an industry.
That's what I mean by resolving people - I mean this person and not the likely thousands of other people who share your name.
As a tool for disambiguation, names are notoriously weak. Governments have known this for a while, and most governments around the world issue a unique tracking number at birth, so that they can ensure each and every one of us, at a minimum, is taxed appropriately.
The bigger opportunity is across walled gardens and networks - for Yahoo, AOL, Bebo, MSN, Facebook and all these other isolated islands of ID to be united in a way much like DNS, where it's not owned by a single company. It's in our industry's mutual interest.
Time is already used as a universal identifier
There has been a lot of activity and progress on this front.
Through the magic of RSS, feed readers can aggregate my digital life as it exists across many silos on the internet (or at least my public-facing digital life.)
Better yet, systems like Flickr or Facebook's F8 Platform allow for "trusted applets" to act as my authenticated agents and access my deeply personal data.
Systems like onxiam.com attempt to help others (or even me) keep track of who I am across the many services I use. (Get it? "On x I am" as in "On Yahoo I am darren45, on MySpace I am...")
This level of openness is both a technical and business challenge.
Openness exposes deep issues around security and trust (are you sure all of those fun, viral Facebook applets you authorized are benign?).
The saying "A chain is only as strong as its weakest link" applies here. On the business side, companies consider their registered user-base as a dearly treasured, proprietary asset. But this notion of user lock-in is an increasingly antiquated notion.
At Yahoo we've taken steps such as providing browser-based authentication, which provides even a one-person start-up with access to our hundreds of millions of registered users.
It's an early step and short of being completely open, but its a step in the right direction nonetheless.
So how do we move to a world where everything has a digital identifier?
My colleague Marc Davis has framed this as the "W4" problem, the four "W's" being who, when, what and where.
Street addresses are little good as a universal identifier
We do a great job as a culture of "when". Using GMT I can say this particular moment in time and we have a great consensus about what that means - although as a frequent international traveller I must admit I've incorrectly translated time zones and called home at 4am more than once.
We also do a very good job of "where" - with GPS we have latitude and longitude and can specify a precise location on the planet.
Latitude and longitude are great for machines, but most of us aren't fluent in lat-long.
We convey place using conventions like "addresses" which contain elements like streets, counties, states, countries, etc. Unlike the pristine mathematical grid lat-long provides, these entities are often "sloppy" and arbitrary.
A country's border is sometimes defined by a natural occurrence like a shoreline, but equally as often by a long-forgotten war with a neighbour.
Translating these sloppy notions of place into precise mathematical definitions is more complicated than might be expected, and is big business; companies like TeleAtlas and Navteq collect, clean up and broker this data to a huge array of customers.
The remaining two Ws - we are not doing a great job of.
I've already mentioned some of the challenges around "who" - identity - but tackling "what" is perhaps even more difficult.
Thinking about how we make sense, coordinate and create resolvers for who and what represent big opportunities. Once we do we have a four-dimensional "W" space, this structure is hugely valuable and allows for some incredible killer apps.
Mobile phones could automatically collect a wealth of data
Here's an example. Today, at many high-tech conferences the organiser will implore the attendees to "tag your photos and blog posts with londonconference07".
Typically the tag the organiser proposes (e.g. londonconference07) is a made-up word designed to be both easy to remember, but also distinct and unique from the universe of tags that already exists.
If everyone sticks to this convention, it allows for easy access to all of the photos, videos, blog posts, etc. that refer to the event using a search engine. Nice!
Except not. People often forget the tag, invent one of their own, or worse don't tag the media they create at all.
And of course most events in the "real world" (outside of the high-tech bubble) don't come with a "use this tag:" directive on the ticket stub. So at Yahoo, we've made efforts to automate this process.
Mobile phones do a great job of providing some of our "W's" for free. My phone knows what time it is. It knows approximately where I am, either through GPS or through the mobile phone network.
Knowing these two key bits of data, we can key into Yahoo's giant event database (known as upcoming.org) and determine: "Aha! Alexandra Palace, 15th of May at 7pm, you must be at this particular event".
Picture sites like Flickr display data about the cameras used
So now when I snap a photo, the system automates the process of tagging. The organiser doesn't have to invent a tag and communicate it to attendees, and I don't have to remember anything. The machines do all this for us.
Of course, I can over-ride this behaviour and point out that I'm not actually at the event, merely in a coffee shop across the street. But the system does the tedious work of introducing that metadata to my media object automatically.
That's what we're heading towards - a world where the what, who, where and when can be generated, read and resolved automatically by machines.
But who decides on the digital ID for real world objects? Do we need a standards body like ICANN to decide on these universal resolvers, or arbitrate on them? To what degree can we lean on efforts like Amazon's ASIN?
Personally, I'm pretty pragmatic. I think there's a lot of hope that a crowd sourcing, tagging approach might work.
Where we find people codifying big blocks of entities - whether in a movie database or books or restaurants, or business entities - I am comfortable taking a pragmatic approach so long as the companies contributing their respective intellectual property are committed to open standards and strategies.
It will happen through small pieces loosely joined, and it is emerging already. Different domain specialists will grab different domain patches.
Once we begin to have this information we can then put it in microformats on the web, which are machine-readable. So then in an automated fashion crawlers can take advantage of that structure.
Right now we don't have much to go on apart from the text.
The web itself is sloppy, loose and unstructured - and these, by the way, are virtues!
Tags could become common place for real world objects
But in making it easy to add microformats, which are just machine-readable, coded bits of structure, we let machines talk to machines and ambiguity over which restaurant I am blogging about, or which film, or which person, will end.
This represents a huge step toward the vision of the semantic web, and will not only create entirely new applications, but will also solve problems that users today have come to accept as part of "life on the web."
This structure should be optional, not imposed. The onus is on us, the builders of the tools, to make it brain dead simple to add this structure.
Often a lot of this crucial information already exists but is discarded or ignored. Take photos, for example.
At Flickr, whenever possible, we've been saving information about the camera that was used to take a digital photo - remembering things like camera make and model, shutter speed, focal length, etc.
Our users can then search the millions of photos in the Flickr database using queries like, "Show me all other photos taken with a Canon PowerShot SD800 IS". This is a fantastic research tool if you're in the market for a new camera.
All of this data is relatively easy to acquire and store but almost impossible to reverse engineer.
Storage is cheap and getting cheaper. We should attach metadata to objects early in the production process and then create formats and structures that allow that to hang out with data itself so we can put it to work later on.
It's beginning to happen.
In the academic community there is lot of work going on with citations databases. The patent office has a highly structured database. Companies like Navteq (for geo-data) or Gracenote (for music data) have done this for their respective domains (see also non-profit, community-driven analogues MusicBrainz.)
As an industry we're picking off these domains one-by-one, but we will need increased cooperation and standards for solving problems like identity.
So, here I am back in the supermarket taking advantage of the universal resolver and the four-dimensional W space.
On my phone I see prices for that can of soup in my neighbourhood. It resolves not only that particular can of soup but knows who I am, where I am and where I live and helps me make an intelligent decision about whether or not it is a fair price.
It has to be transparent and it has to be easy because I am not going to invest a lot of effort or time to save 13 cents.