The enormity of data within our modern-day society both aids and facilitates the lives of many individuals, but questions arise as a result. How can we effectively and efficiently sift through enormous amount of available information? What kind of innovations would allow the general public to derive the most useful information in the easiest manner?
On Thursday, Oct. 17, Nick Ray, a UCSB alumnus and technical program manager working on Microsoft’s Bing search engine, gave a tech talk hosted by UCSB’s Institute of Electrical and Electric Engineer (IEEE) Student Branch and Data Science at UCSB, providing us with a possible answer — visual search, the next revolution of web searching. In theory, it grants the user faster and easier researching methods. This evolutionary technology has been widely applied to multiple powerful and entertaining trends and functions, such as “explore landmarks,” “identify dogs and flowers,” “shop for items you see,” and “Celebs Like Me.”
By incorporating machine learning (a data-analyzing method achieved through a repeated process of testing and debugging), the visual search engine is trained to generalize relevant information through features identification and classification.
While this powerful artificial intelligence has generated a beneficial modern approach to efficiently search and reach results, multiple troublesome issues and challenges remain to be addressed. The search engine can only provide output based on data that is previously taught, which requires technicians to equip the engine with sufficient background knowledge of different subjects. This is often done manually, with people going through images and classifying them so the visual search can identify new images from examples.
This process of inputting and guiding the program to perform tasks accurately is both time-consuming and problematic. In the presentation, Ray demonstrated this by showing various images of flowers, and asking the audience which were roses. People called on picked one or two of the images, but all turned out to be roses, demonstrating the limits of a program trained by human classifiers.
There were also examples of how machine learning misinterprets pictures. These were not only hilarious — depicting images like dogs dressed as cats, and fail to recognize a plate of pasta as database only stored well-decorated pasta dishes — but also provided the audience with a concrete idea of how programmers train machines to reach results: through open-ended adjustment and continuous testing.
Along with the interesting introduction of Bing Visual Search, Kristin Straube, a university recruiter for Microsoft, gave a short presentation on her company and their opportunities for undergrads.
Carrying the mission of “empowering every person and every organization on the planet to achieve more,” Microsoft supports employees to pursue their own interests and elevate their personal passions, while working hard on their position.
“We encourage students to utilize the Microsoft as a platform to boldly explore what they love, and enthusiastically learn from mistakes,” Straube said.
To understand and satisfy the different needs of their global consumers, Microsoft stresses the diversity and inclusiveness of their working environment. The company welcomes global applicants with various backgrounds to make a difference.
The talk is organized to appeal to general students.
“Although students with engineering background would be preferred, all the UCSB students with interest are welcomed to participate in our event. Mainly, we hope our event to provide students with a deeper grasp of and better connection with successful companies, and find the one that suits them best,” said Brian, the president of the Data Science @ UCSB club.