Monday, May 20, 2024
HomeTravelKayak launches image recognition tool to compare flight prices from a screenshot

Kayak launches image recognition tool to compare flight prices from a screenshot


Nearly one year after it became one
of the first travel brands to create a plug-in for ChatGPT
, Kayak is moving
deeper into its use of generative artificial intelligence with several new
tools including an image recognition capability that can offer price comparison
from a screenshot. 

Users of Kayak’s app can upload a screenshot of a flight itinerary from any site, and “Kayak PriceCheck” will check hundreds of sites to determine if a lower price is available.

The tool is one of a handful of new features Kayak unveiled Tuesday. Such innovations may signal what’s ahead as travel companies continue to tinker with large language models like ChatGPT to produce artificial intelligence-driven products.

“At Kayak, being quick to adapt to new technologies is part of our DNA,” Matthias Keller, chief scientist and senior vice president of technology at Kayak, told PhocusWire. “We were one of the first travel brands to launch a plugin with ChatGPT, and we will continue to explore leveraging AI technology to provide more personalized and intuitive search experiences to make it easier to plan travel.”

Kayak said its new tools, including the patent-pending price comparison with screenshots, were the result of “extensive training” of ChatGPT’s large language model on the company’s proprietary database of billions of consumer travel inquiries.

“With today’s launch of Kayak PriceCheck, we’re the only leading metasearch brand letting users compare prices using only a screenshot,” Keller said.

The product is built by and unique to Kayak, which is using third-party tools powered by AI with proprietary customizations to process the image and retrieve the parameters for the user’s flight itinerary. It then searches hundreds of sites to see if it can find a better price using those same parameters.

The four-month-long project to train ChatGPT on Kayak’s database was a balance of prompting and fine-tuning to ensure the system could distinguish between existing, missing and superfluous information on the provided screenshot.

“As with many AI tools, the more prompting and testing, the smarter it gets,” Keller said.

Another new feature is “Ask
Kayak,” which offers natural language search capabilities to provide
personalized travel planning. Rather than using Kayak’s traditional filtering
system, the new tool lets users narrow down choices by typing in specific needs,
like preferred airline, departure time and prices.

The “Ask Kayak” feature started rolling out Tuesday and will soon be available to all travelers in the United States, United Kingdom and Canada with more countries to follow.

“We’re diving deeper into solving complex travel issues more quickly and simply with the launch of ‘Ask Kayak’,” Keller said. “The addition of free form search to apply filters makes it easy for travelers to get personalized results for their next trip.”

Among the other new features Kayak announced Tuesday:

“We are very interested in AI’s ability to process text in a broad sense,” he said. “Descriptions, user-generated reviews and many other aspects of the trip planning process are text-based. Some of these data sets such as our reviews are too large for humans to analyze, and [large language models] will allow us to surface even more relevant information to our users.”

Large language models also factored into Keller’s response when asked about his predictions for this year.

“It will be interesting to see what changes result from the ongoing legal battles over what can be used to train LLMs, as LLMs become more “productized” and mainstream,” he said. “Broadly, as more multi-modal LLMs develop, there may be more natural integration of real-time information into LLMs, lowering the cost to use them and providing more visibility into why a certain prompt works or doesn’t work.”



Source link

RELATED ARTICLES

LEAVE A REPLY

Please enter your comment!
Please enter your name here

Most Popular

Recent Comments

Translate »