AI granny
Spammers and scammers will now have to deal with grandmothers. At least, that’s what a British mobile company O2 wants to do with them. She appears like just another stereotypical granny who does not know the internet and is technologically inept, loves to knit and talk about her family and her cat named Fluffy. What, however, becomes unknown to the scammers is the fact that Daisy is an AI granny and has all the time in the world to talk to them with some tech-enabled tricks in her box. The company said in its statement, “O2 has created a human-like AI ‘Granny’ to answer calls in real time from fraudsters, keeping them on the phone and away from customers for as long as possible. Trained using cutting-edge technology and real scambaiter content, lifelike ‘Daisy’ is indistinguishable from a real person, fooling scammers into thinking they’ve found a perfect target when really, she’s beating them at their own horrible game.”
Debuted last month, the company says this custom-built AI scambaiter can answer calls and hold autonomous conversations from scammers. It combines various AI models that work together to first listen to the caller and transcribe their voice into text. Appropriate responses are then generated through a custom large language model complete with a character ‘personality’ layer, and then fed back through a custom AI text-to-speech model to generate a voice answer which takes place in real time, allowing the tool to hold a human-like conversation with a caller. Reports say, Daisy does not intercept any calls and has multiple phone numbers of her own that O2 has worked to get into circulation online.
A sound box
Nvidia, the biggest supplier of chips and software used to create AI systems, has announced that it will come up with the “world’s most flexible sound machine”. Named Fugatto (short for Foundational Generative Audio Transformer Opus 1), this new generative AI model can create any combination of music, voices and sounds with text and audio inputs. It shared one such video of the output recently on X and said that some music, sounds, and the voice of its CEO Jensen Huang used were AI-generated. Calling it the equivalent of a Swiss Army knife, it claims that none of the existing AI models that can compose a song or modify a voice have the dexterity of this new offering. The model claims to create a music snippet based on a text prompt, remove or add instruments from an existing song, change the accent or emotion in a voice and even let people produce sounds never heard before. The technology is aimed at the producers of music, films and video games, it wrote in its blog. The company is mulling releasing this new model trained on open-source data publicly. But it can be an ideal tool for online scams as well, where voices can be faked.
Fizz out of Xmas spirit
Coca-Cola recently released AI-generated Christmas ads that triggered a huge debate on social media. Many users said the ads were devoid of the emotional element which is intrinsic to the holiday spirit. The AI-generated ads, intended to pay homage to their iconic 1995 Christmas ads ‘Holidays are coming’, failed to strike a chord with the audience. Experts pointed out that it did not focus so much on storytelling than on the mere execution of AI video tools like Leonardo, Luma, Runway and new model Kling. The ads show decked up red delivery trucks with Christmas lights and photos of Santa Claus with two smiling human faces, one holding a bottle of Coke. The criticism is rooted in the view that these are more cautious and less complex and have ended up being quick montages of short clips merely showing vehicles and close-ups of smiling faces. Many say that the ad appears to be in some kind of a hurry to showcase frames but people cannot seem to understand what exactly they want to convey.
Voice search
AI-powered search company Perplexity is planning a hardware debut by building an AI-enabled, voice-based device priced under $50 for users to ask questions and get answers. Sharing the news, Aravind Srinivas, founder and CEO of the AI-powered search engine, recently wrote on X, “Considering making a simple, under $50 hardware device, that will reliably answer your questions voice to voice. Just do this, but do it very well.”