16th March 2023 – (New York) OpenAI, a startup backed by Microsoft Corp, has released a powerful and updated version of ChatGPT called GPT-4. This latest technology has the ability to adopt a Dialogic conversational stance and provide better responses to questions using factual data and analysis. Compared to its predecessor, GPT-4’s tone and delivery style can be modified by its users, making it a more customisable chatbot.
One of GPT-4’s most remarkable features is its “multimodal” capability, which enables it to respond to both image and text commands. This advanced version of ChatGPT offers a vast scope for collaboration and creativity, generating, editing, and refining writing projects, including screenplays, songs, and even imitating a user’s writing style. Furthermore, the text generator has strong analytical skills, enabling it to act as an expert in various fields, including law, chemistry, and more.
Developed with input from over 50 experts, GPT-4 is the latest step in OpenAI’s effort to scale up deep learning, focusing on AI safety and security. According to internal evaluations, GPT-4 is 40% more likely to produce factual responses than its predecessor GPT-3.5 and 82% less likely to respond to requests for prohibited content. The technology’s broader general knowledge and problem-solving abilities enable it to resolve complex issues more effectively.
As the most sophisticated system in OpenAI, GPT-4 generates safer and more insightful responses. It will be continuously monitored and updated with its use in the real world. The technology’s advanced reasoning and command-following abilities have accelerated safety work, aiding in the generation of training data for classifier iterations across training, evaluations, and monitoring.
Trained on Microsoft Azure AI supercomputers, GPT-4’s infrastructure enables it to be distributed worldwide, exhibiting human-level performance on various professional and academic benchmarks. For example, it has passed a simulated bar exam with a score around the top 10% of test-takers, whereas GPT-3.5’s score was around the bottom 10%.
GPT-4’s advantages over ChatGPT are clear. Unlike ChatGPT, GPT-4 is a multimodal tool that can generate information from both texts and images, making it highly difficult to deceive. Additionally, GPT-4’s language understanding capabilities are much more extensive, as it has been designed to understand various languages. While ChatGPT can process 4,096 tokens or roughly 8,000 words, GPT-4 can handle up to 32,768 tokens or over 64,000 words.
GPT-4 is available on ChatGPT Plus and as an API for programmers to create software and services. However, only the text-input feature will be accessible to ChatGPT Plus members and software developers, with a waitlist. With a $20 monthly subscription fee, users can upgrade their account model from the GPT-3.5 default to GPT-4, which offers quicker response times, priority access to new features, and more frequent updates.
OpenAI acknowledges that GPT-4 still has limitations, such as social biases, hallucinations, and adversarial prompts, which are still under development. As more users adopt the technology, OpenAI is working hard to promote transparency, user education, and broader AI literacy to improve and shape the GPT model.