A technology reporter, Stephanie Power, assumes she is pleased with advanced AI. However, she finds the new generation of talking AI friends quite disturbing.
My spouse is an avid supporter of Liverpool FC and experiences nervousness nearly every time they play.
I’ve realized that the best strategy is to avoid him before, during, and occasionally after a game, even though it is annoying.
I heard a friendly woman’s voice speaking to him in his home office after I recently watched a match.
She acknowledged that although it was a challenging game, the Reds were playing well. The strikers for Liverpool were able to produce the results.
This woman, who was she? And why was she giving David this kind of indulgence?
Later, I realized that my spouse was experimenting with an application named Pi.ai. It was introduced by the US tech company Inflection AI last year and illustrates the emerging conversational AI trend.
The concept is that the AI can become your buddy or companion, speaking to you audibly over the speakers on your computer or phone rather than just answering any questions you ask or doing your homework. Additionally, the more you talk with the AI, the more it is thought to understand you, leading to increasingly personalized responses like having a friendly discussion. That’s the theory, anyway.
You still have to write your discussion portion into Pi.ai, but you can select from six distinct human-like voices when it responds. These include a quick-talking American man, and the elegant English woman, my husband, was speaking with.
Conversational AI aims to provide a far more natural, flowing interaction regarding the words and sentences the AI chooses and how it says them audibly. If your first thought is, “But Amazon’s Alexa has been talking out loud to me for years,” don’t worry.
“People may think that voice technology has been around for a while, and they’re probably considering Alexa,” says David Reid, an AI professor at Liverpool Hope University and, incidentally, my husband, a massive Liverpool fan.
“But in the next five years, the worldwide conversational AI market is anticipated to reach $30 billion (£24 billion). Imagine Alexa with empathy if you want an idea of what this would look like.”
Today, tech companies are racing to market their AI conversational mates. Microsoft offers Azure AI and Vertex AI Conversation from Google, and many start-ups are currently in the space. Amazon, meanwhile, intends to continue using Alexa while giving it a more human-sounding voice and conversational AI capabilities.
In a lecture last year, Alexa’s chief scientist, Rohit Prasad, gave another sports comparison to describe his hopes. “The (Boston) Red Sox are my favorite (baseball) team,” he stated. “Just think of Alexa’s happy voice answering if they won. I will be empathetic if they lose.”
A human voice recording is typically the first step in creating conversational AI’s human-like voices. But to convey the necessary tone or loudness and combine words in a natural-sounding manner, the technology must then be able to adjust this.
CEO of Eleven Labs, Mati Staniszewski, says, “We can consider the meaning of a sentence and the relationships between words.” A UK-based tech company has produced 40 artificial intelligence voices that vary in age, gender, and accent. “This means we can capture the intonation, tone, and emotion the AI speaker intends to convey.”
According to Mr. Staniszewski, tone is essential.
Getting that right is essential if you want an AI to seem artificial. Connecting a specific line of thought, emotions, and intonation must frequently expand and roll over several lines. Additionally, as tone and tempo express intent, the model maintains the appropriate flow while accounting for the surrounding context.”
According to Trevor Cox, an acoustic engineering professor at the University of Salford, conversational AI developers will probably run clear of using pronounced regional accents.
“There are still prejudices around strong regional accents,” he claims. According to studies, we are less likely to believe something expressed if it is difficult to understand.
“This is more about flow than accent. Our brains are wired for speedy information decoding. Therefore, the designers of AIs will want to guarantee that the brain can quickly decode information.
“Tone comes next after that. Much more affects the meaning of a message than just the words you use. Thus, it is helpful if an AI can express joy, excitement, or boredom.”
According to David Harley, a cyberpsychology specialist at Brighton University, there are dangers as artificial voices get closer and closer to real people.
He said I am worried people would begin to believe that AI companions and therapists can handle every issue in life. They might start to plan their lives according to the AI’s guidance, which ignores these other essential facets of what it means to be human.
He continues that people must constantly remind themselves that their AI partner needs to be more genuine.
I tried Pi.ai out for myself and felt it was a little too obedient—like a friend who agreed with whatever you said.
My other half, Prof. Reid, claims that’s how it’s designed. “What you perceive as docile, I find it friendly and encouraging. Conversational AI is a valuable tool in a place like a care facility, where residents would enjoy sharing memories of the past with an informed object. Or a call center, where the AI can sense a caller’s frustration and adjust the system appropriately.
By offering therapy to tense baseball and football enthusiasts from Liverpool to Boston and beyond, you can preserve hundreds of relationships worldwide.