This AI Program Was Designed to Pick Names for Craft Beers

iStock
iStock

Microbreweries no longer serve a niche market. Thanks to high demand for craft beer, bottles with quirky labels and names like Backwoods Bastard and Gose Gone Wild are easy to find wherever booze is sold. This also means that coming up with original names for new products is harder than ever. In the not too distant future, beer companies may depend on AI algorithms to do the creative work for them.

As Gizmodo reports, scientist Janelle Shane has developed a neural network whose sole purpose is naming craft beers. A neural network is an artificially intelligent computer algorithm that, after being given a list of data points, can come up with new data that fits the given theme. She’s used this method to generate Star Wars character names (like Rantar Tanter and Captain Kreet), Broadway musical titles (Hot Stans and The Wither Bean are sure to be hits), and "ancient" proverbs (including, "A good anvil does not make the most noise").

Shane fed her beer-naming neural network a dataset of hundreds of thousands of beer names pulled from BeerAdvocate.com. For the IPA category it came up with such colorful names as Dang River, Heaven Cat, and Bigly Bomb Session IPA. The strong pale ale section includes winners like The Vunker The Finger and Brother Panty Tripel.

While many entries are bizarre, there are also plenty that would fit in at the craft section of any beer seller. Names like Cherry Trout Stout and Frog Trail Ale are marketable enough. Selling a beer with the name Oarahe Momnila Day Revenge Bass Cornationn Yerve Of Aterid Ale, on the other hand, might be a tough sell. You can view Shane’s full list of names on her blog.

[h/t Gizmodo]

Google Now Lets Parents Manage Their Kids' Phone Time Remotely

iStock
iStock

Setting screen time limits on teenagers was much easier in the pre-smartphone era. Modern parents often have to choose between taking their kids' phones away or letting them text through family game night—but now Google is offering a different option. Beginning today, September 18, Android phone owners are able to set restrictions on their teens' devices, either by setting time limits, locking their phones remotely, or subjecting app downloads to parental approval, The Verge reports.

These features are rolling out through Family Link, an app Google released in 2017 that lets kids create Google accounts that their parents can access. With the new changes, minors over age 13 can create similar Google accounts, or update their old ones to enable parental controls.

As is the case with Family Link for kids, parents can use Google's software to manage how—and how much—their kids use their phones as well as track their location. The biggest difference with the Family Link apps for teens is that both account holders must consent before parents can start monitoring their kids' phones. And if teens ever decide they want to make their phone activity private, they can choose to turn off the supervision mode. The catch is that doing this will lock them out of their phones for 24 hours and send a notification to their parents.

The new features are now available on all Android phones, and will be coming to Chromebooks soon. Users in the U.S. will also be able to use their Google Assistants to manage their Family Link accounts starting next week.

[h/t The Verge]

AI Is Tackling Yet Another Creative Medium: Improv Comedy

iStock
iStock

AI-generated fan fiction, music videos, and film scripts are often so bad that they’re hilarious. Could an AI program get the same number of laughs if it attempted improv comedy in front of a live audience? As Inverse reports, artificial intelligence researcher Kory Mathewson created an algorithm to find out.

Mathewson, from Canada’s University of Alberta, teamed up with London-based researcher Piotr Mirowski to create a chatbot, A.L.Ex, which stands for Artificial Language Experiment. They fed subtitles from 100,000 films into a neural network in the hope that A.L.Ex would be able to come up with jokes and carry on a conversation with a live human performer. (They also applied a filter to the robot to stop it from saying “politically incorrect” things, and presumably to prevent a disaster akin to Tay, Microsoft’s Twitter bot.)

Once A.L.Ex was sufficiently prepared for the spotlight, a performer interacted with the chatbot (who was given a robot body) on stage in an improv scenario. Audiences were asked to participate in a Turing test: During some scenes, a human backstage was controlling the robot's responses, while in others, A.L.Ex was doing all the work. Audience members were later asked to guess whether the dialogue was coming from the bot or an actual human. The experiment was repeated in three locations: Stockholm, Sweden; London, England; and Edmonton, Canada.

The result? The bot failed to fool humans and pass the Turing test, but it still garnered a few laughs. For one thing, the system was unable to tell complete stories. “If you want to tell a story, humans tend to have to pick up the arc and carry it through, since the Cyborg rarely brings arguably important characters or plot items back,” one of the improv performers wrote, according to a paper that Mathewson and Mirowski uploaded to the preprint platform arXiv [PDF].

Mirowski told The New York Times that the bot is like a “completely drunk comedian” who is only “accidentally funny” on occasion. Fortunately for comedy lovers, machines probably won’t be taking over the stage anytime soon. “We do not think that machines will replace human actors or comedians,” Mathewson told Inverse. “We aim to build new tools and techniques for human storytellers to share their human experience. This work aims to test the development of a new form of medium.”

[h/t Inverse]

SECTIONS

arrow
LIVE SMARTER