OpenAI introduced the improved new version of ChatGPT, GPT-4o.
Last week, just days before Google's major I/O 2024 event focused on artificial intelligence, news emerged suggesting that OpenAI would announce a version of ChatGPT to rival Google Search. This alternative, designed to enhance ChatGPT's access to real-time internet data, was seen as a significant step for OpenAI and potentially appealing to users looking for alternatives to Google.
However, OpenAI CEO Sam Altman quickly refuted this rumor, stating that OpenAI would not be announcing a competitor to Google Search. Nonetheless, it was confirmed that OpenAI would hold a mysterious ChatGPT event on May 13. The anticipated event took place, and OpenAI announced significant innovations.
ChatGPT 4o (Omni) Introduced
OpenAI announced a new version of ChatGPT that can accept and produce outputs in audio, visual, and text formats. Inspired by the word 'omni,' this new version is named ChatGPT 4o.
Advanced Multimodal Capability
ChatGPT 4o can respond to user inputs as quickly as human-to-human conversations, enabling more natural human-machine interactions. While the new version matches ChatGPT-4 Turbo in English, it significantly outperforms Turbo in other languages. Additionally, it offers faster API performance and operates at 50% less cost.
Voice Processing
In previous voice communication methods, three different models were used to convert voice inputs to text, often losing important information like tone and background sounds. ChatGPT 4o overcomes these disadvantages by combining all inputs and outputs into a single model.
Security and New Filters
To enhance security, ChatGPT 4o features new filters that prevent unwanted voice outputs. Initially, only text and image inputs with text outputs and limited voice capabilities will be available. ChatGPT 4o will be accessible in both free and paid tiers, with Plus users enjoying higher message limits.
ChatGPT-Based Voice Assistant
OpenAI also introduced a ChatGPT 4o-based multimodal AI assistant capable of communicating with users through voice and text. This assistant can recognize objects and images, exhibit better logical reasoning, and even detect if users are joking based on their tone. Initially used for customer support, this assistant could eventually be utilized for tasks like reading road signs, translating languages, and helping students with math problems.
10 Examples of ChatGPT 4o (Omni) Capability
OpenAI broke the Internet with mind blowing GPT-4o demos yesterday. And people are already doing wild use cases with it.
1. Make single HTML Facebook Messenger clone
2. Data analysis from spreadsheets with single prompt in less than 30 seconds
3. Pokémon Red gameplay
4. Solve undeciphered languages
5. Generate STL files
6. Generate Matplotlib plots from scribbles
7. Transcribe old writing
8. ChatGPT-4o desktop app can read your screen in real-time
9. Answer math questions with improved reasoning
10. Improved image recognition
Comments