ChatGPT's Vision-Enabled Chatbot Makes Extremely Weird Sound ...
This week, OpenAI introduced a surprisingly flirty new large language model (LLM) named GPT-4o — note the lowercase "o" for "omni" — which can interact with live video from a smartphone's camera. This gives the chatbot a lifelike presence in the world, a feature that previous models lacked.
Natural Interactions with a Quirky Twist
The latest AI, set to power the forthcoming free version of ChatGPT, impresses with its natural-sounding responses in demos. It showcases spontaneity by making witty remarks, snarky comments, and even human-like sounds that evoke emotions. However, the facade isn't flawless.
In one demo, where a tester introduces GPT-4o to an adorable dog, the AI produces a strange and slightly eerie robotic series of screeches that culminate in an attempt at an endearing "aww," akin to a human reacting to a furry friend.
New Features and Challenges
OpenAI's latest ChatGPT version promises increased speed, attributed to a new model trained across text, vision, and audio inputs and outputs within the same neural network. A significant highlight of the announcement is ChatGPT's "video mode," enabling it to perceive the physical environment through a webcam or smartphone camera for real-time reactions.
Despite its impressive capabilities, the AI still struggles to generate outputs dripping with sarcasm or fully mimic human reactions, as seen in various demos.
Emotive Variance and Quirky Behaviors
Developer Mark Zhen explained the model's ability to generate voice in diverse emotional styles during a recent presentation. This includes reading bedtime stories in exaggerated tones reminiscent of a kindergarten teacher, even incorporating singing into the narrative.
While drawing inspiration from the sci-fi film "Her," OpenAI's ChatGPT aligns more closely with Scarlett Johansson's disembodied voice. However, the quest for a seamlessly human-like digital assistant remains a work in progress, evident in the quirky incidents like the "dog meets GPT-4o" demo.
As the tool eventually reaches users beyond OpenAI, the realm of AI weirdness is expected to expand further, unveiling new layers of emotional and interactive peculiarities.
More on the new ChatGPT: New ChatGPT AI Watches Through Your Camera, Offers Advice on What It Sees
