GPT4 released 😍 a comprehensive guide

🎓 OpenAI releases the much smarter GPT-4

🖼️ Visual image prompts now possible

🤯 GPT-4 accepts 25k words

As you may have heard, GPT-4 launched yesterday. Below, we break down what’s different, cool, and useful in OpenAI’s new model.

GPT-4 to accept ~25,000 words in the prompt

This improvement opens up a lot of new opportunity with this model. Previously, with GPT-3, the size of the prompt was limited to about 3,000 words. With the new GPT-4 model, the word limit is ~25K words (roughly the size of a small ebook).

This opens up a lot of use cases. Below are a few:

• Extended conversations
• Document search
• Academic research papers
• eBook summaries
• Longer coding problems and outputs
• Long form content creation
• Legal document analysis
• and more.

GPT-4 understands images and expresses logical things about them

Though not fully released yet, GPT-4 will be able to receive images as prompts. When prompted, the model will have the capability to understand the physics of the image. For example, when presented with an image of hot air balloons with flames blowing into the balloon, the model can logically determine whether the balloon is likely to be rising or falling.

The ability for GPT-4 to accept images as inputs opens up a variety of use cases, including:

• Image captioning and description
• Image search and retrieval
• Visual question answering
• Object recognition and classification
• Image generation and manipulation
• Medical image analysis
• Autonomous vehicle navigation and perception
• Robotics and automation

Those are a few use cases listed above, but I am sure there are so many.

One interesting example is “Be My Eyes.” This technology has been developed since 2012 to assist more than 250 million people who are blind or have low vision. The Danish startup connects volunteers with individuals who are blind or have low vision to assist with daily tasks, such as identifying products or navigating airports. Be My Eyes began developing a GPT-4 powered Virtual Volunteer™ within the app that can generate the same level of context and understanding as a human volunteer.

GPT-4 can analyze and make sense of images

In OpenAI’s research paper, they have a couple great examples

In this first example, the model was provided an image with 3 separate panels on it showing a product that is a faux VGA cable allowing you to put a lightning cable into it and plug it into your iPhone. Obviously, a VGA cable doesn’t get plugged into an iPhone, so let’s see what the large language model thinks of this image.

User Prompt: What is funny about this image? Describe it panel by panel.

GPT-4 Response:

The image shows a package for a “Lightning Cable” adapter with three panels.

Panel 1: A smartphone with a VGA connector (a large, blue, 15-pin connector typically used for computer monitors) plugged into its charging port.

Panel 2: The package for the “Lightning Cable” adapter with a picture of a VGA connector on it.

Panel 3: A close-up of the VGA connector with a small Lightning connector (used for charging iPhones and other Apple devices) at the end.

The humor in this image comes from the absurdity of plugging a large, outdated VGA connector into a small, modern smartphone charging port.

The next example in the research paper shows an image of a man strapped to the top of a taxi cab ironing his clothing. Obviously, this is not normal and should throw the model off. Let’s see what happened in their research when the model was asked what was unusual about the image.

User Prompt: What is unusual about this image?

GPT-4 Response:

The unusual thing about this image is that a man is ironing clothes on an ironing board attached to the roof of a moving taxi.

As you can see, even this release of the GPT-4 model is very sophisticated.

GPT-4 can teach a variety of subjects personalized to your skill level

As mentioned earlier, GPT-4 has significantly increased its memory capacity, allowing it to store up to 50 pages of text during a conversation. This is a significant improvement from its predecessor, GPT-3, which had a more limited capacity.

In addition, GPT-4 has been extensively trained on over 100 trillion parameters from various sources, including academic articles and web pages. This allows it to generate more accurate and contextually relevant responses.

GPT-4’s responses are also remarkably human-like, making it an excellent tool for educational purposes. In fact, GPT-4 can serve as a virtual tutor for students, offering them personalized and comprehensive learning experiences. With its advanced features and capabilities, GPT-4 is a valuable tool for anyone looking to improve their language skills or expand their knowledge base.

GPT-4 has added serious expertise

OpenAI tested GPT-4’s abilities by simulating exams designed for humans, including the most recent publicly-available tests and 2022-2023 practice exams. They did not do any specific training for the exams, and only a minority of the problems were seen by the model during training. However, OpenAI believes the results to be representative. See below for more details on the test scores:

GPT-4’s visual prompting can code websites based on hand-drawn wireframes

On Tuesday, Greg Brockman, President and Co-Founder of OpenAI led a demo showcasing GPT-4’s capabilities and limitations.

During the livestream, Greg took a hand-drawn wireframe and asked GPT-4 to turn the mockup into a colorful website where the jokes are replaced by two real jokes.

Below is the output – a working website with HTML and JavaScript.

ICYMI: Full GPT-4 Developer Livestream

Check out the full livestream above. Greg Brockman, President & Co-Founder of OpenAI demo’s GPT-4 and walks through some interesting use cases.

How can you try GPT-4 yourself?

At the time of writing this, to take advantage of GPT-4, you have to be a ChatGPT Plus user which costs $20/mo. This may change in the future, but this is the current way to use this new model.

OpenAI’s GPT-4 has launched with many new and exciting features. With the ability to accept larger word prompts, understand images, and teach a variety of subjects, GPT-4 is a valuable tool for researchers, educators, and anyone looking to expand their knowledge base. Additionally, GPT-4 has been extensively trained on over 100 trillion parameters, making its responses remarkably human-like. While currently only available to ChatGPT Plus users, GPT-4’s advanced features and capabilities are sure to be a game-changer in the field of Artificial Intelligence.

ai language models generating random ai art

Daily, we instruct the AI to generate a random prompt to give to the MidJourney AI image generator. The results are beautiful.

This week, we are focusing on using a filter called “infrared” which gives the images a really unique and dreamy feel.

Prompt: A city skyline at night: Take a picture of a city skyline at night, with the lights reflecting off the water. The infrared filter will add an otherworldly glow to the image.

MidJourney’s Image: