Official: GPT-4 is already here – Technology Org
OpenAI has just announced GPT-4, its next-generation large multimodal AI model based on natural-language processing.
The previous GPT-3 forms the basis of the famous ChatGPT artificial intelligence chatbot, capable of generating human-like responses to queries provided by its users.
GPT-4 is said to be capable of accepting text and image inputs. This allows users to specify both vision and language tasks. For example, you can now use documents with text plus photographs, diagrams, or even screenshots – the generated output uses all initial data to generate answers based on all available data (try submitting a funny image and asking AI to tell what makes this image amusing ).
Microsoft, which currently cooperates with OpenAI in developing AI algorithms, also announced that its Bing Chat has already been running using GPT-4. This means that Microsoft has been using the new platform since November 2022, when the company conducted its first wide-scale test of its Bing Chat in India.
According to the company, GPT-4 exhibits human-level performance on various professional and academic benchmarks.
“We’ve created GPT-4, the latest milestone in OpenAI’s effort in scaling up deep learning. GPT-4 is a large multimodal model (accepting image and text inputs, emitting text outputs) that, while less capable than humans in many real-world scenarios, exhibits human-level performance on various professional and academic benchmarks,” wrote OpenAI in its announcement.
The new multimodal AI model is claimed to pass a simulated bar exam with a score approximately equal to the top 10% of participants. For comparison, an earlier major version of this algorithm (GPT-3.5) scored around the bottom 10% of all participants taking this test.
OpenAI also released a technical paper dedicated to an extensive description of GPT-4’s capabilities. Furthermore, there is also a system model card that describes the major limitations.