What’s The Difference Between ChatGPT & GPT3?

ChatGPT Confusion

At the time of writing, since the launch of ChatGPT at the end of November 2022, numerous solutions have hit the market claiming to be ChatGPT-branded. However, I'm here to clarify that these solutions are not ChatGPT, but rather GPT3 solutions.

There seems to be a lot of confusion between ChatGPT and GPT3.

This is compounded by some solutions in the marketplace claiming to use ChatGPT which is either totally wrong, trying to ride the hype train, or genuinely confused with the underlying tech.

So, What's The Difference Between ChatGPT & GPT3?

If you got this far, you're probably wondering, what the differences are between ChatGPT and GPT3.

GPT3 stands for Generative Pre-trained Transformer, which is a Large Language Model (LLM) built by OpenAI and released in June 2020. The GPT3 model was later iterated into GPT3.5, also known as InstructGPT, to improve its ability to follow instructions and complete tasks.

If you are using the Davinci model from OpenAI's API, you are using InstructGPT or GPT3.5.

Many companies have already used GPT3 and GPT3.5 to enhance their existing products and create new ones, such as AI-assisted writing tools. However, these are not ChatGPT.

ChatGPT has undergone further training, including RLHF (Reinforcement Learning from Human Feedback).

The training process involved humans reviewing and rewriting responses generated by the model to make them more factually accurate and conversational.

The responses were then fed back into the model to train it on how to produce more human-like answers.

The model then underwent a reward model training process where multiple responses were generated and ranked by humans based on their quality and fit.

The data was then fed back into the model to train it on what constitutes a good response.

OpenAI also used Proximal Policy Optimization (PPO), a Reinforcement Learning approach, to create policies for the ChatGPT language model.

This process not only improved the accuracy and conversational breadth of responses but also trained the model to produce safer content by blocking racist, sexist, or inappropriate responses.

However, despite being built on the GPT3.5 model, ChatGPT produces better responses than GPT3.5, as it has been specifically trained to generate high-quality, human-like responses.

This is why people are often confused between the two models.

What Can You Do With GPT3?

GPT3 is incredibly powerful and ultimately the foundation of ChatGPT. Its applications are wide-ranging because it has an API to build plugins or tools that use Large Language Models.

It can generate text, answer questions, translate languages, summarize long articles, and even complete tasks such as coding and creating charts.

Here are a few things you can do with GPT-3:

• Text generation: GPT-3 can be used to generate human-like text in various styles, from creative writing to news articles, and even poetry
• Question answering: GPT-3 can answer a wide range of questions with high accuracy, making it a useful tool for knowledge management and customer service
• Chatbots: GPT-3 can be used to develop advanced chatbots that can handle complex conversational tasks, such as booking a flight or ordering food
• Language translation: GPT-3 can translate text from one language to another, providing near-human-level accuracy
• Content summarization: GPT-3 can summarize long articles or documents into concise summaries, making it easier to quickly understand the most important information
• Code generation: GPT-3 can write code, from simple scripts to complete applications, making it a valuable tool for software development
• Creative applications: GPT-3 can be used for creative projects, such as generating music, visual arts, or even video game design.

What Can You Do With ChatGPT?

As I've covered previously ChatGPT is built to have conversations and do it well. Its conversational capabilities (remember only its own chat interface) are better than GPT3 because it's been tuned to do it. It understands natural language input and generates human-like responses to questions and supports follow-up questions.

Some of the things that can be done with ChatGPT include:

• Question answering: ChatGPT can answer a wide range of questions, providing users with relevant and accurate information on a variety of topics
• Conversation simulation: ChatGPT can mimic human conversations, making it ideal for use in customer service, virtual assistance, and other scenarios where a human-like interaction is required.
• Text generation: ChatGPT can be used to generate text, such as product descriptions, headlines, and other types of content
• Summarization: ChatGPT can be used to summarize long articles, news reports, or other written content into shorter, more concise text
• Translation: ChatGPT can be used to translate text from one language to another, making it a useful tool for global communication.

Summary

So.... ChatGPT is built on the GPT3.5 model but is a conversational interface accessible only through a browser and does not have a publicly available API..at the time of writing. One thing we know about conversational AI is that the landscape is changing quickly and an API release is coming soon.

GPT3.5 is available via an API and a browser and it's this technology that can be used to create lots of different applications including chatbots.

It's also worth mentioning that GPT3.5 can also be leveraged to provide conversational experiences "Like" ChatGPT. Remember they are not ChatGPT, but it's possible to leverage GPT3.5 and add enhanced conversational capabilities which are similar. So memory and context coupled with a smaller domain area knowledge, not the entire internet!

This type of implementation is possible and is something we are working on at The Bot Forge so feel free to get in touch if you'd like to learn more.