Unless we specifically disable them, crawlers constantly watch us move around the web, forming an image of our interests and prejudices. Then, the algorithms reflect these opinions and present us with news, articles and answers that support what we already think.
We are more likely to click on things that fit our existing thoughts and interests, but it would not be objectivity. Best?
Jordi Ribas, corporate vice president of AI products at Bing, thinks that Ribas manages Microsoft's search engine from its US headquarters. UU., But lived in the United Kingdom for three and a half years while establishing the Bing team in Europe. .
"Obviously, as a search engine, our mission is to provide results that are as complete, objective and reliable as possible," Ribas told TechRadar. "In any case, in a world of false news and misinformation on the web, I believe that objectivity in the search could not be more important".
Identification of questions with multiple perspectives
To that end, Bing launched a new feature called Intelligent Answers. When you enter a question with several valid answers, the search engine summarizes them on a carousel to provide a balanced overview.
"A Sometimes there is a unique answer for a query, but sometimes we can understand and identify that there are multiple perspectives, "said Ribas. "We use advanced AI techniques based on deep learning that basically read the entire web, then we try to find which passage or series of passages are most relevant to that question.With machine reading comprehension, or MRC, we can sometimes identify multiple perspectives, where multiple sources converge on the same answer. "
Identifying questions with multiple answers involves several techniques, including the analysis of feelings, which identifies the opinions expressed in a text: positive, negative or neutral.
"Answer a simple question like 'Is coffee good for you? & # 39; "said Ribas. "There are many reliable sources that tell you that there are good reasons to drink coffee, but there are also some very reputable ones that say the opposite: deep learning allows us to project multiple queries in the passages to what we call semantic space and find coincidences
"Then we find that there are documents that are grouped separately when applying the sentiment analysis technique. There is a set of documents that are grouped around positive reasons for coffee, and others that are grouped around the negative reasons for coffee. If we find that there are authoritative sources in both, then we realize that this question really deserves a response from multiple perspectives. And that's what we call it. "
Bursting the bubble
Although Smart Answers could challenge our expectations, Ribas says the response so far has been very positive.
" I think what's happening today, Due to a lot of custom feeds on the web, social networks, trying to reinforce some of the same articles and the same information that users click on, people end up living in a kind of bubble.
And if you have certain political views, or have certain biases, you interact with technology in a certain way, and then the Aloritmos learn it, and end up reinforcing the same biases you have. That's what makes society a little more distant these days, and it's helping to polarize society. I feel that search engines have a responsibility to be more objective, and ultimately, our goal is to provide the most reliable and objective information we can. "
Ribas says that industry professionals are also satisfied with the results " the comments we received from analysts in the US UU They were "Aha," finally someone assumes responsibility and takes a step forward, and not only says that the answer is negative because that is what the algorithm tells us & # 39;
"No, we have to work harder and invest in these more advanced algorithms that help us understand that a given question has multiple perspectives, we feel it is our responsibility to provide those perspectives and get people out of their bubble."
Intelligent Answers aren also influenced by your browsing history, and does not contain any ads or "sponsored" articles.
"The ads follow a different process," Ribas said. "In fact, even our ad team is separate from what we call the algorithmic team, and we have a specific location for ads, it's usually at the top of the page, as you can see, sometimes in the right lane , and we label them as advertisements.This part has no advertising signal whatsoever. "
Comments and future
Smart Answers only form a small percentage of the search results at this time, but Ribas and his team plan to build it – Although not too fast.
"We are still learning a lot, and we are still trying to improve it, and we also want to be careful not to exaggerate," he said. "We want to make sure that the accuracy is high, since the danger of any algorithm that uses AI, since it is an automatic learning algorithm, is that sometimes it will make mistakes.
" We want to make sure that users have a solution quick way to tell us. We can take a look at what happened and how we can improve the algorithm. And that's why we started small, but you'll see more coverage as time goes by. "
You can offer comments on Smart Answers (and any other aspect of Bing) by using the link at the bottom of the page. Results, and the option could become more prominent in the future, appear along with the answers themselves.
You may soon see Smart Answers elsewhere as well, including Cortana. "If you ask Cortana if coffee is good for you, I think Cortana probably will not do it today "I have an answer because there is not one," Ribas said. "But every
time you have a unique answer on top of Bing, that really flows through Cortana, and then we're working now for Cortana to say," Actually, there are different perspectives on this. There are some things for which coffee is good, but according to this other source, if you drink too much coffee it can be harmful for you. "And that is definitely in process."
Hopefully the slow but steady approach means that the team You will not need much caffeine to see them in the afternoon shifts.