Main

June 28, 2017

AirBnB personalises, tunes search results

Airbnb learned over time that machine learning could be used to offer this personalization, Mike Curtis said. Airbnb introduced its machine learned search ranking model toward the end of 2014 and has been continuously developing it since. Today Airbnb personalizes all search results.

Airbnb factors in signals about the guests themselves, as well as guests similar to them, when offering up results.

For example, guests provide explicit signals in their search -- the length of stay, the number of bedrooms they need. But as they examine their search results, they may show interest in similar, desirable attributes that the guests themselves might not even notice.

"There's a bunch of other signals that you're giving us based on just which listings you click on," Curtis says. "For example, what kind of setting is it in? What kind of decor is in the house? These are things Airbnb can use to feed into the model to come up with a better prediction of which listings to show you first."

The company pulls well over a hundred signals into the search rank model, Curtis says, and then the machine learning algorithm figures out how all the signals interact, to produce personalized search rankings

March 29, 2017

Adsense for the masses

Major publishers admit to 'advertiser-friendly' skew:

If you want to take something good and make it less good, there's no more reliable method than to chop it up into tiny bits and then recombine them. A door made of particleboard isn't as strong as one made of solid pine. An MP3 of a song lacks the sonic richness of a high-fidelity record. A hamburger may or may not be as delicious as a rib-eye, depending on your personal taste, but it's definitely likelier to contain fecal bacteria and pink slime.

The global advertising industry is currently experiencing its own version of food poisoning from tainted ground beef. Johnson & Johnson, Verizon, and AT&T are among the giant marketers that have stopped buying ad space on Google's ad network and on YouTube in response to reports of ads appearing alongside hate speech, ISIS recruiting propaganda, and other objectionable content. Racing to contain the boycott, Google issued an apology on Tuesday and said it is taking steps to ensure greater "brand safety" in the future. Those steps include "taking a tougher stance on hateful, offensive and derogatory content," changing the default settings for ad campaigns, and giving marketers new controls allowing them to exclude specific websites or types of content from their campaigns.

Continue reading "Adsense for the masses" »

February 1, 2017

Flatiron Institute

Computers have been a fixture for decades in astrophysics and many other fields of science. But typically, the computer programs are written by graduate students, often abandoned after they finish their programs. "Those people aren't great coders, for the most part," Mr. Simons said.

At the Flatiron Institute, a good fraction of the staff will be professional computer programmers, producing software not only for the in-house scientists but also available for anyone else who needs it.

"These are really interesting questions, and we can think longer than the three-year grant cycle. They can tackle tough questions and put the time in that's necessary."


-- Marilyn Simons.

January 30, 2017

Propublica: breaking the black box what Facebook knows about you

Propublica's breaking the black box what Facebook knows about you.

January 26, 2017

Cambridge Analytica's psychographic profiling for behavioral microtargeting for election processes

Understand personality, not just demographics. OCEAN model: Openness, Conscientiousness, Extroversion, Agreeableness, Neuroticism.

In a 10 minute presentation at the 2016 Concordia Summit, Mr. Alexander Nix discusses the power of big data in global elections. Cambridge Analytica's revolutionary approach to audience targeting, data modeling, and psychographic profiling has made them a leader in behavioral micro-targeting for election processes around the world.

Cambridge's voter data innovations are built from a traditional five-factor model for gauging personality traits. The company uses ongoing nationwide survey data to evaluate voters in specific regions according to the OCEAN or CANOE factors of openness, conscientiousness, extroversion, agreeableness and neuroticism. The ultimate political application of the modeling system is to craft specific ad messages tailored to voter segments based on how they fall on the five-factor spectrum.

The number-crunching and analytics for Mr. Trump felt more like a "data experiment," said Matthew Oczkowski, head of product at Cambridge Analytica, who led the team for nearly six months.

Continue reading "Cambridge Analytica's psychographic profiling for behavioral microtargeting for election processes" »

August 3, 2016

Apple acquired Turi, a machine learning software startup

Apple has acquired Turi (the former GraphLab and Dato), a machine learning software startup. The startup formerly went by the names GraphLab and Dato.

Carlos Guestrin, cofounder and chief executive of the startup, serves as Amazon professor of machine learning at the University of Washington. Amazon has brought on machine learning talent in the years since 2012, when Guestrin got the position. The startup employed some former Microsoft employees, so it's fascinating to see Apple acquire it.

Apple has acquired several other machine learning startups recently, including Emotient, Perceptio, and VocalIQ.

Turi's competitors included Databricks, H2O, and Neo4j, among others. But because Turi has added artificial intelligence into its technology, it faced competition in that area as well.

July 31, 2016

Future is mining cloud data

The next big competition in cloud computing also involves artificial intelligence, fed by loads of data. Soon, Mr. Kurian said, Oracle will offer applications that draw from what it knows about the people whose actions are recorded in Oracle databases. The company has anonymized data from 1,500 companies, including three billion consumers and 400 million business profiles, representing $3 trillion in consumer purchases.

"Most of the world's data is already inside Oracle databases," said Thomas Kurian, , Oracle's president of product development

That's the kind of hold on people's information that perhaps only Facebook can match. But Mark Zuckerberg doesn't sell business software. At least, not yet.

July 8, 2016

Culture Digitally Facebook trending its made of people but we should-have-already-known-that/

Culturedigitally: facebook trending its made of people but we should have already known that.

July 2, 2016

Facebook makes the news

According to a statement from Tom Stocky, who is in charge of the trending topics list, Facebook has policies "for the review team to ensure consistency and neutrality" of the items that appear in the trending list.

But Facebook declined to discuss whether any editorial guidelines governed its algorithms, including the system that determines what people see in News Feed. Those algorithms could have profound implications for society. For instance, one persistent worry about algorithmic-selected news is that it might reinforce people's previously held points of view. If News Feed shows news that we're each likely to Like, it could trap us into echo chambers and contribute to rising political polarization. In a study last year, Facebook's scientists asserted the echo chamber effect was muted.

But when Facebook changes its algorithm -- which it does routinely -- does it have guidelines to make sure the changes aren't furthering an echo chamber? Or that the changes aren't inadvertently favoring one candidate or ideology over another? In other words, are Facebook's engineering decisions subject to ethical review? Nobody knows.

The other reason to be wary of Facebook's bias has to do with sheer size. Ms. Caplan notes that when studying bias in traditional media, scholars try to make comparisons across different news outlets. To determine if The Times is ignoring a certain story unfairly, look at competitors like The Washington Post and The Wall Street Journal. If those outlets are covering a story and The Times isn't, there could be something amiss about The Times's news judgment.

Such comparative studies are nearly impossible for Facebook. Facebook is personalized, in that what you see on your News Feed is different from what I see on mine, so the only entity in a position to look for systemic bias across all of Facebook is Facebook itself. Even if you could determine the spread of stories across all of Facebook's readers, what would you compare it to?

"Facebook has achieved saturation," Ms. Caplan said. No other social network is as large, popular, or used in the same way, so there's really no good rival for comparing Facebook's algorithmic output in order to look for bias.

What we're left with is a very powerful black box. In a 2010 study, Facebook's data scientists proved that simply by showing some users that their friends had voted, Facebook could encourage people to go to the polls. That study was randomized -- Facebook wasn't selectively showing messages to supporters of a particular candidate.

Facebook tinkered with users emotions in 2014 news feed experiment

NY Times Technology on Facebook's tinkering with users emotions in 2014 news feed experiment: outcry stirred.

http://www.nytimes.com/2014/06/30/technology/facebook-tinkers-with-users-emotions-in-news-feed-experiment-stirring-outcry.html.