On Monday, OpenAI debuted a recent chatbot with impressive – or bizarre, depending on your perspective – capabilities. With the ability to see, hear and speak like a real person, OpenAI managers demonstrated how the latest ChatGPT can translate live conversations and tell bedtime stories in a variety of voices.
OpenAI’s head of pioneering research, Mark Chen, asked the chatbot to tell a “bedtime story about robots and love.”
“Oh, a bedtime story about robots and love? I will cover you!” – an enthusiastic female voice answered immediately. “Once upon a time, in a world not much different from ours, there lived a robot named Byte. Byte was an intriguing robot. I’m always discovering.”
Chen paused, “I want there to be a little more emotion in your voice, a little more drama.”
“I see. Let’s add some drama,” ChatGPT-4o said, starting his story over again, but this time with a deeper voice and more intrigue.
Chen interceded again: “No, no, no, ChatGPT, I really want maximum emotion, maximum expression, much more than you did before.”
“Understood. Let’s amp up the drama,” he replied, repeating his opening line again in a much louder, even more dramatic voice. OpenAI’s chief technology officer, Mira Murati, asked the chatbot to switch to the story in the robot’s voice; it obeyed. Fellow researcher Barret Zoph then chimed in , asking for a singing voice. ChatGPT sighed and began to sing.
Murati Chen and Zoph showed how ChatGPT-4o can translate a live conversation between Italian and French, and how it can listen to your breathing and guide you through breathing exercises. The chatbot is twice as quick as GPT-4 Turbo and is available for free users.
“We’ve been really focused on improving the intelligence of these models over the last few years, and they’ve done quite well,” Murati said. “But for the first time, we are really taking a huge step forward in terms of ease of utilize.”
Perhaps even more impressive than its real-time conversation skills are ChatGPT’s visual capabilities. A chatbot can look at someone’s face through their phone’s camera and tell them how they’re feeling based on their facial expression. It can view a user’s computer screen through screen sharing and facilitate them solve a math problem.
Safety considerations
Murati said the chatbot’s capabilities, while a significant improvement, pose “recent challenges for us when it comes to security.”
“Our team has been difficult at work figuring out how to implement fraud protections,” she said, adding that OpenAI works with governments and companies from “all industries” to achieve its goals. She did not provide details about security issues, OpenAI’s plan or the nature of its partnerships.
