The Best Use of AI Ever: A 'Grandma' Built to Waste Telescammers’ Time
cereal_burpist writes:
I know many Soylentils are sick of AI, but I found this to be clever and actually doing something useful:
The Best Use of AI Ever: A 'Grandma' Built To Waste Telescammers' Time
https://decrypt.co/291711/the-best-use-of-ai-ever-a-grandma-built-to-waste-telescammers-time
British telecom giant O2 announced what could be the first deployment of conversational AI specifically designed to counter telephone fraud in the UK telecommunications sector.
The hero: a clueless grandma who endlessly and frustratingly digresses while idiot scammers attempt to hoodwink her out of money.
The system, dubbed "Daisy," uses AI models to snare fraudsters in meandering conversations lasting up to 40 minutes, potentially preventing millions in losses across the network's 24 million mobile customers. The tool operates autonomously around the clock, requiring no human intervention to maintain conversations with suspected scammers.
"Stop calling me dear, you stupid [expletive deleted]!" one exasperated tele-scammer shouted in a video O2 aired to show off its new tool.
"Got it, dear!" the kindly AI grandma replied.
Here's how it works. When a user receives a scam call, they forward the call to the number 7726, which had been used as a hotline to report fraud. From there, Daisy employs a custom large language model with a "personality layer" that generates real-time responses to the scammers.
The system transcribes incoming voice to text, processes it through the AI model so responses are context-relevant, and converts the text output back to speech, creating natural-sounding conversations that can include fictional personal details and bank information.
It's designed to exploit fraudsters' targeting of elderly victims by presenting as a chatty grandmother, complete with meandering stories about family and knitting.
Read more of this story at SoylentNews.