Exploring the Cutting-Edge Features of ChatGPT-4.0: What’s New and Exciting

Users of the free version of ChatGPT are in for an upgrade. OpenAI has introduced a new flagship model, GPT-4o, alongside significant modifications to its user interface.

‍

At its Spring Update event, OpenAI unveiled the latest GPT-4o model, now available in both the free and premium versions of ChatGPT. This update introduces a more natural and emotionally resonant voice assistant, alongside enhanced vision capabilities.

‍

‍

First Look 

This advanced model offers real-time reasoning across text, audio, and visual inputs, marking a significant leap toward more natural human-computer interactions. GPT-4o can process any combination of text, audio, image, and video inputs, generating corresponding outputs seamlessly. 

‍

The response time to audio inputs with GPT-4.0 rivals that of human reactions, delivering replies in just 232 milliseconds and averaging around 320 milliseconds. While it matches GPT-4 Turbo's efficiency in handling English text and code, it notably outperforms in processing non-English languages. Moreover, it is significantly faster and costs 50% less via the API. GPT-4.0 also boasts advanced capabilities in vision and audio comprehension, surpassing previous models in these domains.

‍

‍

Features 

1. Free 

GPT-4o breaks away from the exclusivity of the $20 per month premium Plus service by offering unrestricted access to all users. OpenAI's decision ensures that GPT-4o's advancements are available to a wider audience. Alongside the inherent features and enhancements of GPT-4o, users can utilize additional resources without any cost. These resources include personalized chatbots and entry to the ChatGPT store, where users can explore various models and tools crafted by the community.

‍

Free users can also, harness advanced data analysis tools, image recognition capabilities, and Memory, a function that enables ChatGPT to retain past interactions.

‍

‍

2. Conversational Speech 

OpenAI's live demos showcased ChatGPT's interactive vocal exchanges, highlighting its real-time conversational abilities. Users can interject, change tone, and receive personalized responses. Presenters engaged ChatGPT in tasks like storytelling, demonstrating its adaptability across natural, dramatic, and robotic tones, as well as its versatility in singing and intensifying narratives.

‍

3. Live Translations 

The showcased functionalities of the voice assistant left a lasting impression, particularly the live translation tool which truly stood out.

‍

Throughout the demonstrations, Mira Murati, chief technology officer of OpenAI communicated in Italian with the voice assistant, while Mark Chen tested its ability to translate between English and Italian seamlessly, yielding promising results.

‍

‍

4. Video and Screenshot Uploads

In addition to voice inputs, ChatGPT now offers expanded modes of engagement, such as video sharing. This means you can livestream a math problem you're grappling with and seek assistance directly. Whether you're seeking a solution or guidance to work through it independently, ChatGPT is there to assist.

‍

Furthermore, users can easily exchange screenshots, images, and documents containing both text and visuals. Additionally, ChatGPT facilitates seamless access to past conversations, and real-time information retrieval within ongoing discussions, and even supports advanced data analysis tasks. You can upload charts or code and seamlessly integrate them into your queries for enhanced insights.

‍

‍

Safety Measures 

GPT-4o prioritizes safety across its various modalities by implementing measures such as data filtering and post-training behaviour refinement. New safety systems have been developed to regulate voice outputs, ensuring user security. Evaluations conducted according to a Preparedness Framework and voluntary commitments indicate that GPT-4o poses no higher than a Medium risk across cybersecurity, CBRN, persuasion, and model autonomy. Extensive external red teaming involving over 70 experts has been conducted to identify and address risks associated with newly added modalities. Although audio modalities pose novel risks, steps are being taken to enhance technical infrastructure, usability, and safety measures before their release. Despite ongoing improvements, limitations exist across all modalities, which are being addressed through testing and iteration.

‍

What to Expect Next?

Keep an eye out for further announcements as ChatGPT-5 is slated for release this year, promising the next generation of AI integration.

‍