Kakao unveils AI model that can interpret pictures and text

Jeong Hyeon-hwan & Kim Tae-gyu, UPI News Korea
·2 min read
UPI
Korean tech company Kakao unveils an AI model that interprets words and pictures at the same time. Photo courtesy of Kakao

SEOUL, Jan. 22 (UPI) -- South Korean tech giant Kakao has unveiled an open-source artificial intelligence program named Honeybee that can interpret pictures as well as words.

After introducing the program Friday during the Korea Ministry of Science and ICT-hosted AI strategy meeting, Kakao made the large-language app available online.

Honeybee, with its ability to comprehend image and text concurrently, can answer questions involving varied content types, according to the company.

For example, when prompted to create a conversation between two animals in a photo with a cat sitting on the back of a dog, Honeybee produced a dialog in which the cat says, "Hey there, big guy. Thanks for letting me ride you," and the dog responds, "You're welcome, feline friend. I love having you on my back."

In another example, a user who is shown a photo of two basketball players asks, "How many times did the player on the left win a championship?" and receives an answer from Honeybee.

Kakao has uploaded Honeybee's source code on GitHub, an online software developer platform that enables users to create, store and manage code.

The company said it was sharing the Honeybee program with global developers to contribute to the improvement of multi-modal large-language models.

"In the AI age, the collaboration between corporations and establishment of eco-systems are critical," Kakao CEO nominee Chung Shin-a said during the meeting.

Kakao said the large-language app could prove to become an effective educational tool.

U.S. consultancy Allied Market Research expects the AI-powered education market to grow more than 40% annually over the next decade, from $2.5 billion in 2022 to $88.2 billion by 2032.