AI Wine critic creates accurate reviews without tasting, sniffing or spitting a drop


If you're looking to splash out on a fermented grape beverage, you'll probably want to read a review or two beforehand. However, with the creation of an AI wine critic, some online critiques may have been done without a sip of wine touching the reviewer's lips. (Not because they're lazy, but because they're an AI incapable of drinking.)

An AI wine critic that can’t be beat

Via scientific journal Scientific American, researchers at Dartmouth College created a cutting-edge AI wine reviewer. Built upon a dataset of over 120,000 reviews from Wine Enthusiast magazine, the artificial intelligence software was able to quickly build up its knowledge of wine.

“[I]t was just a very unique data set," said Keith Carlson, a Dartmouth College computer engineer. "People talk about wine in the same way, using the same set of words."

After learning from previous reviews, the virtual wine reviewer started to create its own pieces. With information about each product’s wineries, cost and ingredients list, the software was able to construct copy that resembled human-written articles.

In one review, the faux sommelier described a bottle as being “very dry and a little thin in blackberry fruit, which accentuates the acidity and tannins”. As someone who knows nothing about wine, that sounds pretentious enough to be a wine review!

Read More: US Senators infuriated by military for not taking aliens seriously

Readers can't tell who is the program

Of course, one of the issues with using artificial intelligence for critiques — outside of the ethical implications of reviewing a product the reviewer can't experience — is that readers will be able to tell if copy was written by a bot. Or will they?

According to the Dartmouth College study, survey members were given over 300 reviews to sample. Each review had two versions: a human-written professional opinion and one written by an AI. In almost every scenario, the AI’s presence was only felt in a handful of pieces.

It’s interesting to see how a dataset of lexicon and knowledge can help to build a critical opinion. With the Dartmouth study, it shows that a critical opinion is more based on knowledge and fact than many believe reviews to be.

This Article's Topics

Explore new topics and discover content that's right for you!

AINews