Built with GPT-4

??Etienne Pretorius 2023 rights reserved

Director, Business Analyst and Legal Consultant at PBC Group

It seems Chat GPT-3 has not been handed a formal announcement of Chat GPT-4 launch. That is even though the recent version of Chat GPT Mar 14 Version exists. I ran a query today via Chat-3 and got a response "As of my knowledge cutoff of September 2021, there was no official announcement or information available about Chat GPT-4. However, I can provide information on Chat GPT-3 and how it compares to other AI available." ChatGPT Mar 14 Version.

The most recent success in OpenAI's attempts to scale deep learning. The large multimodal model GPT-4 emits text but also takes image and text inputs. GPT-4 works on a range of academic and professional standards at a human-level, despite being less successful than humans in many real-world situations.

According to reports by OpenAI, Chat GPT-4 can be applied to a variety of contexts, including machine translation, text summarization, and dialogue systems. As a result, they have generated a lot of attention and advancement in recent years. Enhancing these models' capacity to comprehend and produce natural language text, especially in more intricate and nuanced contexts, is one of the main objectives of their development. GPT-4 was assessed using a range of tests that were initially created for humans in order to measure its abilities in such situations.?It apparently performs admirably in these tests and frequently outperforms the overwhelming majority of human test takers. For instance, GPT-4 receives a mark on a mock bar exam that places it in the top 10% of test takers. GPT-3.5, on the other hand, receives a number that is in the bottom 10%.

Greg Brockman, the president and co-founder of Open AI, gave users a sneak preview of the image-recognition capabilities of the most recent iteration of the system on Tuesday. This version is not yet accessible to the general public and is only being tested by a business called Be My Eyes. The feature will enable GPT-4 to react to images that are provided with prompts, analyze them, and answer questions or carry out tasks based on them. According to Brockman, "GPT-4 is not only a language model; it is also a vision model. It can flexibly accept inputs that arbitrarily intersperse images and text, kind of like a document."

During the demonstration, GPT-4 was asked to explain why a picture of a squirrel holding a camera was amusing. We "don't expect them to use a camera or act like humans,". Brockman later uploaded to GPT-4 a picture of a crudely drawn website design that he had made by hand, and the program used the image to build a functioning website.

References:

ChatGPT Mar 14 Version at https://chat.openai.com/chat