GPT-4, the next major update to the technology that powers ChatGPT and Microsoft Bing, was announced by OpenAI.
What exactly is GPT-4?
- GPT-4 is an advanced version of GPT-3 and GPT-3.5 that is a large multimodal model created by OpenAI that accepts images as input.
- It outperforms humans on a variety of professional and academic benchmarks, and it can solve difficult problems with greater accuracy.
What distinguishes GPT-4 from GPT-3?
- Unlike GPT-3 and GPT-3.5, which were limited to textual input and output, GPT-4 is multimodal, allowing it to understand more than one modality of information.
- It is more difficult to fool than previous models, and it can process a lot more information at once, making it more suitable for long-form conversations and content generation.
- It has improved in accuracy and understanding of languages other than English.
The capabilities of GPT-4
- GPT-4 can generate captions and analyses from images, as well as answer tax-related questions, schedule meetings, and learn a user’s creative writing style.
- It can handle more than 25,000 words of text, allowing for more use cases such as long-form content creation, document search and analysis, and extended conversations.
- It significantly reduces hallucinations while producing fewer undesirable outcomes such as hate speech and misinformation.
GPT-4’s multilingual abilities
- GPT-4 is more multilingual, and it can correctly answer thousands of multiple-choice questions in 26 different languages.
- It handles English the best, with an accuracy of 85.5%, but Indian languages like Telugu aren’t far behind, at 71.4%.
GPT-4 availability
- GPT-4 has already been integrated for a variety of purposes into products such as Duolingo, Stripe, and Khan Academy.
- Image inputs are still in the early stages of development and are not yet publicly available.
Source: https://indianexpress.com/article/explained/explained-sci-tech/gpt-4-new-different-chatgpt-openai-bing-explained-8500024/