1. Introduction to Claude 3 AI Chatbots

   - The recently announced Claude 3 AI models by artificial intelligence start-up Anthropic have been identified as ground-breaking in industry standards for numerous cognitive tasks.

   - Anthropic, a competitor of OpenAI, was originally established by preceding leaders at ChatGPT.

   - This AI start-up has closely associated itself with Amazon being its primary cloud computing partner whereas OpenAI has a significant partnership with Microsoft.

2. Claude 3: A Trio of Artificial Intelligence

   - The Claude 3 family comprises three advanced AI models in consecutive order of abilities- Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus.

   - Claude is a set of Large Language Models (LLMs) created by Anthropic.

   - LLMs is a special group of generative AI models designed to comprehend and generate human-like text.

   - The Claude chatbot is skilled in handling text, voice messages, and documents and can generate faster, contextual responses compared to its contemporaries.

3. Training Mechanism of Claude Chatbots

   - Training for these chatbots involves Supervised Learning (SL) and Reinforcement Learning (RL).

   - The process under SL involves the LLM producing and later revising responses based on certain principles. The aim for this is to reduce any damaging effects of the AI's outputs.

   - The RL phase involves training the model based on feedback generated by the AI itself, in which the responses are evaluated based on constitutional principles.

4. Unveiling Claude 3: Haiku, Sonnet, and Opus

   - Claude 3 Opus is heralded as the most powerful model, Claude 3 Sonnet is the mid-tier model balancing capability and cost, and Claude 3 Haiku is suited for use cases requiring instantaneous responses.

   - Currently, Claude Sonnet powers the Claude.ai chatbot for free with only an email sign-in required.

   - However, Opus is only available through Anthropic’s web chat interface and Claude Pro subscription service.

5. Situational Limitations of Claude 3 Models

   - While Claude 3 excels in tasks such as answering factual questions and OCR - optical character recognition (the ability to extract text from images), and creative tasks like writing Shakespearean sonnets, it sometimes struggles with complex reasoning and numerical problems.

   - The models also showcased biases in their responses, favouring certain racial groups over others.

6. Unveiling Large Language Models (LLMs)

   - LLMs are capable of decoding common language problems like text classification, answering questions, and text generation.

   - These models undergo training on extensive datasets to comprehend patterns, structures, and relationships within human language.

   - They can be divided into Autoregressive Models, Transformer-based Models, and Encoder-decoder Models based on their architecture.

   - Also, these models can be Pretrained and Fine-tuned Models, Multilingual Models, and Domain-specific Models based on their training data.

   - The size and availability of LLMs also identify them as large high-performance requiring models, open-source readily available models, and closed-source proprietary models. Some notable examples of these models are LLaMA2, Bloom, Google BERT, OPT-175 B for open-source models and GPT 3.5 by OpenAI, and Gemini by Google for closed-source models.