Hi readers! Hope you are enjoying the blogs. If you have questions pl. do ask or wait for the vlogs that will explain difficult to understand things (This is for the locals as the people in the west have generated this knowledge and do not need easy version).
Four ChatGPT versions are currently available hence, it is not always easy to know which version is to be selected and used. It is therefore, imperative to know the essentials of each version.
What is Chat GPT 1?
GPT and ChatGPT 1: both are processing tools, but they differ in their technological capabilities and pricing.
Please note that the term GPT is also being used to refer any product that uses any kind of generative pre-trained transformers, not just the versions that comes out from OpenAI.
Chat GPT 1 is trained to understand and generate human-like text based on the input it receives. It can converse so realistically that it is difficult to tell whether you are talking to a human or a machine?
Chat GPT 2 is a large language model pre-trained on a dataset comprising 8 million web pages. Unlike GPT 1, that needs specific fine tuning (a process of training a pre-trained language model on a specific dataset that helps improving chat GPT performance and adapting it to a particular task or domain). Chat GPT2 can do multitasking learning while training.
Chat GPT 3 is a neural network machine model trained to use internet data to generate any type of text.
GPT 3.5 is a refined version of GPT 3 that was released in November 2022. It is currently being offered both as a free web app and as paid Turbo API.
Chat GPT 4 is a large language model that can analyze textual data such as customer reviews, social media mentions, or internal documents, and provide valuable insight in a comprehensible manner. It was released in March 2023. It is specifically suitable for workplaces and/or organizations.
It is not free and requires monthly subscription.
It is more powerful and can select through massive data sets offering quantitative summaries and qualitative integrations as well, that are actionable and can mimic prose, art, images and voice etc. It is capable of handling over 23000 words of text allowing it for uses in content creation, extended conversations and documents search. It can follow complex instructions in natural language and solve difficult problems with accuracy specifically mathematical problem.
Currently, two versions of Chat GPT, one built on GPT-3.5 and the other on GPT-4 are being used. Both are fine-tuned for conversational applications using a combination of supervised and reinforcement learning.
GPT 3.5 can:
- Understand and generate human-like text using natural language comprehension and generation to complete various natural language-related tasks,
- Translate fluently and accurately, text from one language to another,
- Answer questions by providing relevant information, making it suitable for chatbots and virtual assistants using GPT 3.5 Turbo, which is tailored to work with the Chat Completions API,
- Generate concise summaries of longer text, such as documents and reports,
- Generate content and write projects, such as emails and codes,
GPT 3.5 Turbo can be fine-tuned by the developers on their own data thus allowing them to tailor the model to specific uses and enhance its performance on tasks. Users can access two model variants through GPT 3.5 Turbo API.
The extended capabilities of GPT 4 include:
Multimodality, which means it can accept images as input to generate captions, classifications, and analyses,
Its two sized window (one each with 8,192 and 32,768 tokens) can handle over 25,000 words of text thereby enables users for content creation, extended conversations, and document search and analysis,
Tokens are the fundamental building blocks of natural language processing (NLP) models, including ChatGPT. Tokens work as individual units of text that NLP models uses to represent textual data. When raw text is processed, it is converted into a sequence of tokens which can be words, sub word segments, characters, or symbols. For example, the word “ChatGPT” would be broken down into the tokens “Chat,” “G,” “P,” and “T.” which can be used to provide a standardized way of representing language that the model can work with. ChatGPT 4 has a token limit which can corresponds to approximately 2,731 words. This limit ensures that conversations fit within the model’s constraints.
Helping easy analyses and understanding of meanings of the inputs,
Encoding other aspects of language, such as grammar and syntax, and
Helping ChatGPT to accurately understand the structure and meaning of the text, allowing it to generate coherent and accurate responses,
supporting programming code completion and text summing up in a maximum of 10 words,
It can browse internet and is trained on data between April 2023 to December 2023, depending on the model version.
Both are currently available to the public at a range of capabilities, features, and prices (can be asked from the service providers).
That’s all for now dears. Please read these blogs carefully even if you are not using them or is not your field but it is the field of near future. The world has adopted it fully and is improving the shortcomings that they found during its usage while we are lagging to the extent that we don’t even understand the terms clearly.
See you next week with a new episode “ how does chat GPT works?”
Take care, bye.
