5 key differences between GPT-4 and GPT-3.5

After all the speculation and claims about the capabilities of GPT-4, the much-anticipated fourth version of the GPT language model family came out on March 14, 2023.

The GPT-4 doesn't come with some of the much-rumored features. However, the latest model significantly improves on the GPT-3.5 and its predecessors. So how is GPT-4 different from GPT-3.5? Find out with TipsMake through the following article!

1. Creativity

5 key differences between GPT-4 and GPT-3.5 Picture 1

One of the most obvious advantages of the GPT-4 model over GPT-3.5 is its ability to give more creative answers to queries. However, GPT-3.5 is also very innovative. There is a long list of creative things you can do with ChatGPT using this model.

However, the GPT-4 takes it even further. While the creative advantage of GPT-4 may not be obvious when trying to solve fundamental problems, the difference in creativity between both models becomes apparent when the task becomes difficult. and requires a higher level of creativity.

For example, if you ask both versions to complete a creative task like writing a poem in both English and French, ChatGPT with the latest GPT-4 will yield better results. The GPT-4 response will use both languages ​​per line, while the GPT-3.5 will alternate between both languages, each line using one language in turn.

2. Image input

5 key differences between GPT-4 and GPT-3.5 Picture 2

GPT-3.5 can only accept text prompts, while GPT-4 is multimodal and can accept both text and image inputs. From an image of a handwritten math problem to Reddit memes, GPT-4 can understand and describe almost any image. Unlike GPT-3, GPT-4 is both a language and a visual model.

During the GPT-4 announcement live stream, an OpenAI engineer provided the model with a screenshot of the Discord server. GPT-4 can describe every detail on it, including the names of the users online at the time. A hand-drawn mockup of a website was also provided for the model with instructions for turning it into a web page, and to our surprise, GPT-4 provided a working code for a web site that works. image.

3. Feedback is safer

5 key differences between GPT-4 and GPT-3.5 Picture 3

While the GPT-4 isn't perfect, the measures it takes to ensure a more secure response are a welcome upgrade over the GPT-3.5 model. With GPT-3.5, OpenAI has taken a censorship-based secure approach. In other words, some safety measures have been taken a closer look. OpenAI tracks what users did and the questions they asked, identifying errors and trying to fix them.

With GPT-4, most of the safety measures have been incorporated into the system. To imagine the difference, it's like building a house out of solid materials from the start versus using anything later to fix things up when a fault pops up. According to OpenAI's GPT-4 technical report [PDF], GPT-4 generated only 0.73% of malicious responses, much less than GPT-3.5's 6.48% of malicious responses.

4. Realism of Feedback

5 key differences between GPT-4 and GPT-3.5 Picture 4

One of the flaws of GPT-3.5 is its tendency to confidently produce meaningless and dishonest information. This is known as the "AI illusion" and can cause distrust of AI-generated information.

In GPT-4, the AI ​​illusion is still an issue. However, according to the GPT-4 technical report, the new model has between 19% and 29% lower hallucinogenic potential than the GPT-3.5 model. The response from the GPT-4 model on ChatGPT is also significantly more realistic.

5. Context window

5 key differences between GPT-4 and GPT-3.5 Picture 5

A less talked about difference between GPT-4 and GPT-3.5 is the context window and context size. The context window is the amount of data a model can keep in its "memory" during the chat session and for how long. GPT-4 has significantly better context and window sizes than the previous version.

In practical terms, this means that GPT-4 can remember the context of the conversation, as well as the instructions given in the conversation, longer.

One problem with GPT-3.5 is that the model tends to go off topic or not follow the instructions you want to proceed with during the chat. For example, you can tell the model to call you by your first name and it will do so for a while but then stop following those instructions. While this problem persists with GPT-4, it is reduced by having a better context window.

Another problem is the limit on the amount of text you can use in the reminder at once. Summarizing long text using GPT-3 often breaks text into parts and summarizes them bit by bit. The improvement in context length in the GPT-4 model means that you can paste an entire PDF file at once and get a summary without splitting it into parts.

Without a doubt, the GPT-4 is a significant step up from its predecessor. While still experiencing some of the same limitations as in GPT-3.5, significant improvements in several areas and the addition of new functions make this model an exciting new step in the process. pursue really intelligent AI language models.

David Pac

Update 27 June 2023