Is AI getting any closer to reaching a technological singularity and embarking on a quest for world domination and human extinction? Probably not, but it is getting better at having conversations with itself without human input or interaction.

OpenAI – one of the front-running artificial intelligence research companies in the world – has just launched GPT-4o, its new flagship version of the ChatGPT large language model (LLM) which can “reason across audio, vision, and text in real time”.

• READ MORE: New earbuds by Teenage Engineering-founded firm Nothing feature ChatGPT integration

So what do you do to test one of the most technologically advanced iterations of artificial intelligence right now? Make it sing, obviously.

In a new demonstration video posted on the OpenAI YouTube channel, two smartphones are placed side by side, one with its camera enabled and one without. The phone without its camera enabled is instructed to ask questions of the one with its camera enabled, in order to make sense of the surrounding world without seeing it.

After some AI-to-AI pleasantries, the first AI begins asking questions of the second AI in order to try to paint a picture of the surrounding space.

“What are we looking at today?” it asks, to which the second replies, accurately: “We’re looking at a person wearing a black leather jacket and a light-coloured shirt. They’re in a room with a modern industrial feel, with exposed concrete or plaster on the ceiling and some interesting lighting.”

The AIs are even able to notice when a new person walks into the camera’s view and places bunny ears over the head of the host.

After interacting with one another for two minutes, both AIs are instructed to create and sing a song about what just happened, during which they once again reveal their penchant for interior design.

“In a room where modern lights peak,” belts out the first AI. “A surprise guest with a playful streak,” replies the second.

Now, would GPT-4o make it through the audition stage of The X Factor in its current form? Probably not. To be honest, it feels like it has a way to go before it can write authentic-sounding original music that rivals creative human minds. 

That said, what artificial intelligence is able to do is accelerating at a tremendous rate, and the powers that be are clearly teaching LLM’s how to be creative, so without government intervention and legislation, the future may very well spell trouble for the value of human creativity.

Last year, Black Eyed Peas star will.i.am detailed his predictions for the future of artificial intelligence in music, suggesting it’s likely AI will be good enough to replace some human talent at some point.

“You know it’s gonna make better songs than you. It’s Pac-Man right now, we ain’t even got to Halo. We’re in freakin’ Super Mario Bros., we ain’t even got to Call of Duty yet. This thing’s gonna make better songs than you soon, bro.”

Learn more about GPT-4o at OpenAI.

The post Watch two AIs sing a duet with each other in real time appeared first on MusicTech.