Skip to content
Noroff Case Studies

OpenAI GPT Plus Case Study

Author: Kevin Havn <KevinHavn>

Introduction

Chat GPT, which became a global phenomenon shortly after its global release in November 2022, received a significant upgrade released to consumers the 14th March 2023, marking the advent of GPT-4. This upgrade, the subject of this case study, was launched in a premium service named “GPT Plus”, being which included a newer model of OpenAI’s GPT model, and other additions - such as creating your own personal tailored GPT. While the differences are subtle, there are a few major differences to dive into.

Some of the largest advertised improvements between the older GPT 3.5 model and the newer GPT 4.0 model are:

  • Multimodality, capable of both text, voice and image inputs and outputs.
  • Increased capabilities, tested on both AI benchmarks and exam results.
  • Personalized user created and tailored GPTs.
  • Live access to the internet, allowing real-time information.

Brief History

  • November 30. 2022: GPT 3.5 is initially released for public usage

  • March 14. 2023: GPT 4.0 is released for consumers, under GPT Plus.

  • October 2023: Dall-E 3 is integrated into GPT 4, allowing the creation of images based on text input.

  • November 16. 2023: User tailored and customized GPTs are released for the GPT Plus subscription.

Main Features

After ChatGPT suddenly rose through the roof in popularity at its public release in 2022, the service has enjoyed a lot of popularity. When the new GPT-4 was incorporated through the GPT Plus service, it saw enough popularity to warrant that a wait list, where customers interested in a subscription had to wait for availability. In addition to this waitlist, Chat GPT came with - and currently has - a request limit of 40 messages every 3 hours.

Multimodality

GPT-4 is a multimodal model, allowing for both voice, image and text to be used as inputs. This serves a variety of purposes, ranging from the ability to send images of the contents of your fridge to have ChatGPT tell you what kind of food you can make, to sending an image of a plant to allow the AI to identify what said plant is.

Increased Capabilities

The GPT 4 model is advertised as being significantly more reliable, creative, and having a higher understanding of complex instructions. This is because of several improvements, such as a larger and more recent training dataset than GPT 3.5. OpenAI are also continually looking for more ways to refine on the GPT model, several of which they have incorporated in the GPT 4 model. All these improvements culminate in the fourth iteration of the GPT model being better than the previous in benchmarks, and excelling in several exams.

Custom GPTs

Custom GPTs allows you to set up your own chatbot based on the GPT 4.0 technology, pre-programming the chat bot by answering simple questions regarding what the user wants to use the custom GPT for. These GPTs can for example be set up to allow for extra knowledge to be installed onto the GPT, or changing its ability to browse the web, analyze data, or make images. Custom GPTs are very easy to set up, requiring no programming to be useful. Another extremely useful feature for custom GPTs is the actions feature, which allows you to define custom actions by making one or more APIs available to mthe GPT. This can allow your GPT to interact with the world, or to integrate external data.

Dall-E 3 Integration

Dall-E 3 integration allows the AI model to create images based on text descriptions. This allows for a lot of new possibilities, where users can create entire artworks based only on a few sentences. While the technology is still very young, there are a lot of possibilities for this specific technology, perhaps especially within front-end development. Creating design prototypes by only writing a couple of sentences, for instance, is one of the many big appeals one can expect from this technology.

Sora

Sora is a new piece of AI technology as of writing (february 2024), which is a text-to-video and image-to-video AI model, built on similar technology to DALL-E 3 and ChatGPT. Sora uses the same re-captioning technology that generates highly descriptive captions for visual training data, as seen within DALL-E 3, and a similar transformer architecture as within the GPT models.

Sora is still in early development, and openAI are taking several important safety steps before considering the integration of Sora as an available openAI product. These safety features will include tools to identify Sora-generated videos, and utilizing many of the same safety features as present in DALL-E 3.

Getting Started

To get started using ChatGPT itself, head on over to the OpenAI website. After logging in, chatGPT comes pre-populated with some prompts. Otherwise, simply ask whatever you’d like, and the AI will reply momentarily.

To get started with using the newer GPT-4 version, you’ll need to click the dropdown menu to at the top left of the page, and select GPT-4. This should prompt you to subscribe to GPT Plus, after which you’ll have access to all the included features.

After ensuring you’ve selected GPT-4, to access the DALL-E 3 integration you can make a prompt such as “draw an image of a central processing unit”, or “make a depiction of a server rack”.

Access to the Custom GPTs can be found in the left-hand sidebar of the website. Simply click “Explore GPTs” and “Create” in the top-right corner, where you’ll be prompted by OpenAI’s “GPT Builder” which will ask questions regarding the user’s expectations, purposes and requirements for this custom GPT. You can also toggle the code analysis, browser search, and image generation features for this custom GPT here.

Cons

  • Subscription can be steep, at $20 per month as of writing in march 2024.
  • Easy to hit limit to how many messages can use GPT 4.0
  • Many areas see a limited increase in capabilities compared to GPT 3.5
  • Can be slower in responses than GPT 3.5, seemingly when the AI model is under heavy load.

Pros

  • More capable GPT model
  • Easily created custom-tailored GPTs for personal use
  • Dall-E 3 integration for image generation
  • Multimodality, allowing user to communicate with GPT model using images and audio in addition to text.
  • Browser lookup, allowing the AI to present you with up-to date information on a wide variety of subjects present on the internet.
  • Code analysis, allowing the AI model to run code, and give detailed feedback.

Conclusion

The ChatGPT Plus subscription has a lot to offer. A better ChatGPT with things like custom GPTs, and DALL-E 3, adds a lot of useability for a lot of people - which will most likely increase as openAI add more technologies to the ecosystem. However, a negative for quite a lot of people is that the pricing can be quite steep at 20$ a month as of writing, with the Plus subscription also having a limit of how many prompts can be sent to ChatGPT 4.0 every day - the prompt is 40 messages every 3 hours. There’s also a wait list for the initial subscription to chatGPT plus, as a result of the immense popularity the service is seeing.

References