NEED DIRECTIONS? YAHOO CHOOSES THE MOST SCENIC ROUTES
Be prepared to surrender to computerdom more of that vanishing territory known as the uniquely human sensibility.
Researchers at Yahoo Labs in Barcelona havedeveloped software that can determine the most beautiful route you should take.
Calculating the fastest route has been a key trick of computing for some time. But now a team of researchers from Yahoo and the University of Torino in Italy has developed an algorithm that chooses the most “emotionally pleasant” way home.
Photos along London routes from Google Street View and Geograph were tested two at a time on UrbanGems.org, with humans choosing the image that was more beautiful, happy, or peaceful/quiet.
The researchers then plotted the images, with scores representing their pleasantness, on a map. Their algorithm chose the most pleasant route among several alternatives based on beauty scores. Beauty has its price, of course, with the more pleasant routes taking an average of 12 percent longer in travel time.
Then, the team picked 30 Londoners to evaluate the paths the computer thought were nicest. The Londoners agreed with the computer’s choices.
Using fickle humans to evaluate beauty one by one could take an eternity. So the research team mined metadata from about five million photos on Flickr taken of the same places along the London routes. The number of photos taken at a given spot, positive-sounding comments, and other kinds of data were treated as endorsements.
They then carried out the London process in reverse for Boston routes. The software determined the most beautiful locations along Boston paths in Flickr photos and plotted the most pleasant paths via software. Several dozen humans were asked to test the choices. Again, the humans, apparently yielding their territory without a fight, agreed with the software’s choices of the nicest routes.
Mapping beauty by hand
“I just took a family road trip from Seattle to the Northern California redwoods and wanted to pass through various parts of the Oregon shore,” IDC‘s director of software development research Al Hilwa told VentureBeat. “Neither my car’s nav system nor my phones could handle the automated routing to optimize for scenery. I ended up mapping where I wanted to go by hand with Google Maps and user forums of best sites to visit and such.
“Had I had such software, it would have made the job a lot easier.”
Taking this into account, how close are we to having computers that can appreciate beauty?
“Probably far,” Hilwa assured us. “But an algorithmic assessment of beauty according to programmed human values will happen for sure, much earlier than the singularity.”
Forrester Research VP James McQuivey is sanguine about our new taste-makers.
This kind of “route personalization,” he told us, “is good practice for the holy grail of route personalization which will someday occur.”
On that day — all of which is possible once the data collection is automated, McQuivey said — “your GPS system can intelligently tailor your route based on dozens of competing factors, offering you two to three routes that provide whatever mix of routing you have demonstrated you prefer in the past.”
The route will also “incorporate crime statistics by time of day, traffic reports, panhandler sightings, tourist trap avoidance, shopping opportunities, chocolatier locations, and more.
“That’s why it’s vital that the [Yahoo] team move from human assessments of beauty to computer-based assessments of beauty, so it can quickly ingest more and more information to make the experience all the more personal to you and your preferences.”
And don’t forget to add the randomizer so the app can show us the best route for walking around aimlessly.