Conversational AI Like ChatGPT May Soon Have a Face That Looks Human

Chatbots that look and sound like you

  • New software allows conversational AI to look and act more like humans.
  • Some observers warn that human-like AI could be deceptive.
  • Users report falling in love with one type of chatbot companion.
An artificial intelligence made of binary code that has a human-like face.

Yuichiro Chino / Getty Images

Chatbots might soon look and act more like people, but the technology could open the door to deceptions. 

A new human-like interface for conversational AI (artificial intelligence)—chatbots—is intended to give the wildly popular ChatGPT a face and speaking voice. The software gives developers the ability to create photorealistic digital assistants. Experts say chatbots masquerading with people could be a problem.  

"Once it becomes difficult to distinguish a Zoom call with a human from a chat session with an AI bot, lots of things will change," Jed Macosko, a physics professor at Wake Forest University, told Lifewire in an email interview. "Scams will become very compelling. People will enjoy the AI' friends' they make online and will use their free time to indulge in long conversations."

Conversational AI That Looks Human

The Israeli company D-ID claims its software is the first platform to enable face-to-face conversations with conversational AI chatbots in a natural way. 

"Large language models like GPT-3 and LaMDA are changing the way we relate to and interact with technology, and we are not far off from all of us having our own personalized AI assistants and companions," Gil Perry, the CEO of D-ID said in the news release. "We are making tech more human by giving it a face and making the interaction more natural."

We are making tech more human by giving it a face and making the interaction more natural.

Putting a human face on, or humanizing objects or animals, is far from a new concept, Bob Rogers, the CEO of Oii.ai, a data science company, told Lifewire via email. He said it's only natural that humans want to 'humanize' a chatbot. 

"ChatGPT creators aimed for a chatbot that could mimic how people speak to each other," Rogers added. "With its success, generating a 'face' for the chatbot could potentially help people build an even more lifelike relationship with it."

The new software from D-ID is part of a push to make computer-generated animated faces more realistic. For example, the Replika AI companion is a conversational AI that uses animated figures to give the software a human interface. The app is so realistic that some users even report falling in love with their Replika companions. 

Increasing loneliness is part of the draw of AI chatbots like Replika, Rob Brooks, a professor of evolutionary ecology at the University of New South Wales Sydney in Australia, recently wrote in an article on The Conversation. He said that humans are developing real emotional connections with AI. 

Giving AI a human appearance might fool users into thinking that chatbots have actual emotions. Tech analyst Bob Bilbruck, the CEO of the consulting firm Captjur, said in an email that it's best to think of AI chatbots as a tool like a phone or a laptop. 

"A human face is not necessary, literally or figuratively," Bilbruck said. "Some day this technology will reside in a pretty complex robotic human-like form factor, but this would be mostly geared toward the job or function that this application is geared for; for instance, a robot that checks you into your hotel or a gate agent at an airport, or a TSA checkpoint robot."

The Future of Human-Like Chatbots

Chatbots will likely continue to get more human-like, and some experts say that the move could have unforeseen consequences. Scientists will continue to train AI to be aware of 'feelings' or emotional impact, Rogers said. For example, if conversational AI can learn to detect emotions accurately, it can be a tool in healthcare for initial diagnostic purposes. Or the AI can start distinguishing between a sarcastic human response and a genuine one, leading to improved interactions between humans and chatbots.

"At the moment, ChatGPT has a very bland personality," Rogers added. "Sure, it's nice. But it sticks to a very average, non-controversial perspective on almost everything it writes. That is proper since it was designed to be a 'safe,' human-friendly technology. I suspect that humanizing AI will include deviating more from the average and developing AI systems that have a specific perspective on things so that their interactions are warmer and more human."

Was this page helpful?