ChatGPT 4: What's new, biggest changes and what we expect.

Today OpenAI released its updated model for ChatGPT: ChatGPT 4. The announcement was part of Microsoft's response to Google's announcement (PaLM API).

The release was split in two: This morning (10am PST) the official website announced via social media the releasing of the model, and three hours after (1pm PST) via YouTube Streaming Greg Brockman (President and Co-Founder of OpenAI), did a developer demo showcasing GPT-4 and some of its capabilities/limitations. There is a summary of the release and some comparisons with the previous model:

• GPT-4 is a large-scale, multimodal model that can accept image and text inputs and produce text outputs.
• GPT-4 is a Transformer-based model pre-trained to predict the next token in a document.
• The post-training alignment process results in improved performance on measures of factuality and adherence to desired behavior.
• GPT-4 is more creative and collaborative than its previous versions. It can generate, edit, and iterate with users on creative and technical writing tasks, such as composing songs, writing screenplays, or learning a user’s writing style.
• 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 (OpenAI internal evaluations). They incorporated more human feedback, including feedback submitted by ChatGPT users, to improve GPT-4’s behavior. They also worked with over 50 experts on early feedback in domains including AI safety and security.

Some of the features they shared in the streaming were:

1.?????It could be used as a code assistant, copying, and pasting the coding issues directly, and the model will adjust the code to solve the issue:

2.?????They are working on a visual model to upload, read and work with images (is not a text model, it’s multimodal processing text, images, and audio). The public version of this version will be released soon:

3.?????It could work as a personal assistant, uploading personal hand-written notes and processing them even when it comes to personal tasks (like taxes):

We tested the the new one (left) with previous model (right) via ChatGPT Plus and the results were:

• Why are you better than chatgpt3?

• Explain the plot of Cinderella in a sentence where each word has to begin with the next letter in the alphabet from A to Z, without repeating any letters.

• Andrew is free from 11 am to 3 pm, Joanne is free from noon to 2 pm and then 3:30 pm to 5 pm. Hannah is available at noon for half an hour, and then 4 pm to 6 pm. What are some options for start times for a 30 minute meeting for Andrew, Hannah, and Joanne?

• write a python script to analyze my monthly purchases

• Find x, y, and z such that x3+y3+z3=k, for each k from one to 100.

In the next months, we are going to test the API to contrast performance and outputs, and we are going to share the results with you.

As usual, feel free to share your comments/ideas about ChatGPT.