Pinterest allows users to “pin” photos and videos onto boards, helping them to “discover ideas through images,” especially those pinned by people or companies that they follow. It uses neural networks, which make millions of calculations quickly, to surface and suggest the images that people will like. According to Pinterest senior vice president of engineering Jeremy King, this tool is responsible for “nearly 100 percent” of the company’s growth. In Q1, Pinterest’s AI-powered formula drew in almost 480 million people.
The Wall Street Journal reports that this number is “up 30 percent from last year’s first quarter, while ad sales have more than doubled since 2018.” According to eMarketer, that compares to an 80 percent jump in ad sales at YouTube and 56 percent at Facebook, two companies often grouped with Pinterest.
The use of neural networks is improving AI’s ability to tackle so-called unstructured data, which includes images, speech and text. Forrester Research and IDC both reported that, “about a third of companies that use AI in some form have adopted neural networks or will do so within 12 to 18 months.”
Microsoft’s LinkedIn is another company that uses neural networks “to match users and ads.” Twitter said it does not, and Facebook and YouTube declined to reveal what technology they use. Gartner research vice president Nader Henein noted that such networks are “incredibly good at finding patterns.”
That worries some privacy advocates who point out that, “with a large enough dataset, [neural networks could] theoretically make enough connections to identify people whose data is in a repository.” AI company Collective[i] co-founder and vice chair Stephen Messer noted that, “organizations need to take responsibility for their data and technology and carefully manage their uses.”
Pinterest said it “takes steps to protect that data but declined to provide specifics.”
Neural networks ingest a lot of data to learn about a specific object’s features and use artificial neurons (a combo of processors and software algorithms) to “mimic the way human neurons transmit and process information.” In other words, their learning process is quite similar to how humans learn, a “trial-and-error method” to start recognizing patterns.
Deep neural networks, with many layers of neurons, is “able to make finer distinctions.” Pinterest “searches, the boards of people they follow and what pins they click on and save … [as well as] ad data on users.” WSJ reports that, “a single neural-network model focused on Pinterest ads may make more than 30 million predictions per second.”
IDC program vice president for AI Ritu Jyoti said that, “considering the complexity and speed needed to instantly match the right ads with the right users, it would be impossible for a company such as Pinterest to narrowly target ads without neural networks.”
Pinterest Rolls Out Idea Pins, Like Stories But Not Ephemeral, ETCentric, 5/20/21