OpenAI has unveiled its latest AI creation, GPT-4o, a more conversational and human-like chatbot capable of interpreting users' audio and video inputs and delivering real-time responses. Demonstrations released by the company showcase GPT-4 Omni assisting users in various scenarios, from interview preparation to calling customer service for an iPhone replacement.

The demos exhibit GPT-4o's diverse capabilities, including sharing jokes, providing real-time translations in bilingual conversations, adjudicating games like rock-paper-scissors, and even reacting with sarcasm when prompted. In one instance, the chatbot interacts with a user's puppy, expressing delight at the adorable pet.

OpenAI's CEO, Sam Altman, expressed his amazement at the AI's movie-like qualities, noting the significant strides made in achieving human-level response times and expressiveness. A text and image-only input version of GPT-4o was launched on May 13, with the full version slated for release in the coming weeks, accessible to both paid and free users via ChatGPT's API.

The "o" in GPT-4o signifies "omni," emphasizing its capacity for handling text, audio, and image inputs simultaneously—a notable advancement compared to previous AI iterations like ChatGPT-4. OpenAI highlighted GPT-4o's superior abilities in understanding vision and audio, including identifying user emotions and breathing patterns. Moreover, GPT-4o boasts improved speed and cost-efficiency compared to its predecessors.

According to OpenAI, GPT-4o can respond to audio inputs within as little as 2.3 seconds, with an average response time of 3.2 seconds, mirroring human conversational speeds. This innovation represents a significant leap forward in natural human-computer interactions, promising enhanced user experiences across various applications and industries.