GPT-4: A Closer Look at the Highly Anticipated Language Model
GPT-4 was just released. Explore everything you can do with it.
On March 15th, 2023, the world witnessed a monumental event in the history of artificial intelligence as OpenAI released its highly anticipated GPT-4 language model. The hype surrounding this new language model is well-deserved, as it boasts some impressive features and capabilities.
OpenAI has released a powerful new image- and text-understanding AI model, GPT-4, that the company calls “the latest milestone in its effort in scaling up deep learning.”
According to a recent summary by a speaker, GPT-4 can understand both images and text, making it a powerful tool for a variety of applications. It is available for OpenAI’s paying users through chatGPT Plus and developers can sign up on a waitlist to access the API at a reasonable price:
GPT-4 -8K context window (about 13 pages of text) will cost $0.03 per 1K prompt tokens and $0.06 per 1K completion tokens.
GPT-4–32k with a 32K context window (about 52 pages of text) will cost $0.06 per 1K prompt tokens and $0.12 per 1K completion tokens.
Priority Access: Developers can get prioritized API access to GPT-4 for contributing model evaluations to OpenAI Evals that get merged, which will help them improve the model for everyone.
One of the most exciting aspects of GPT-4 is its ability to generate text and understand images. In fact, it has a new capability to analyze the content of images and provide instructions for the AI’s next interactions. This opens up a whole new world of possibilities for image-based applications and AI-powered systems.
However, GPT-4 can accept images as inputs and generate captions, classifications, and analyses. But this functionality is not available yet as they are working on this.
Despite its impressive capabilities, the speaker notes that GPT-4 is not perfect, and improvements can still be made. However, they believe that the model has a lot of potentials and can improve people’s lives by empowering a wide range of applications.
Longer Text:
GPT-4 is capable of handling over 25,000 words of text, allowing for use cases like long-form content creation, extended conversations, and document search and analysis.
GPT-4 is 82% less likely to respond to requests for disallowed content and 40% more likely to produce factual responses than GPT-3.5 on our internal evaluations.
Partner Organizations:
1-Duolingo
GPT-4 deepens the conversation on Duolingo.
2-Be My Eyes
Be My Eyes uses GPT-4 to transform visual accessibility.
3-Stripe
Stripe leverages GPT-4 to streamline the user experience and combat fraud.
4-Khan Academy
Khan Academy explores the potential for GPT-4 in a limited pilot program.
Infrastructure:
GPT-4 was trained on Microsoft Azure AI supercomputers. Azure’s AI-optimized infrastructure also allows OpenAI to deliver GPT-4 to users around the world.
Limitations:
GPT-4 still has many known limitations as it also has data cut off till 2021 that OpenAI is working to address, such as social biases, hallucinations, and adversarial prompts. OpenAI encourages and facilitates transparency, user education, and wider AI literacy as society adopts these models. OpenAI also aims to expand the avenues of input people have in shaping our models.
GPT-4 significantly reduces hallucinations relative to previous GPT-3.5 models (which have themselves been improving with continued iteration). GPT-4 scores 19 percentage points higher than our latest GPT-3.5 on their internal, adversarially-designed factuality evaluations.
Performance in various Exams:
OpenAI used the post-trained RLHF model for these exams.
Conclusion:
OpenAI has introduced GPT-4, a large multimodal model that performs at a human level on challenging professional and academic benchmarks. GPT-4 has shown superior performance compared to other existing large language models in a range of NLP tasks, without requiring task-specific fine-tuning, and this improved capability has been demonstrated across multiple languages. The predictable scaling of GPT-4 has allowed for accurate predictions of its loss and capabilities. However, the increased capabilities of GPT-4 also present new risks, and OpenAI has taken steps to understand and improve its safety and alignment. While more work needs to be done, GPT-4 is a significant step toward the development of useful and safe AI systems.
Note: You can read the OpenAI research paper GPT-4 Technical Report.
