Chatbots for Researchers

1. Introduction to large language models: Intuitive explanation of the foundations of artificial
intelligence, including logistic regression, deep neural networks, vector embeddings, and
generative models. Overview of the three main training phases of large language models:
pretraining on large text corpora, supervised fine-tuning on specific tasks, and reinforcement
learning from human feedback.
2. Working with large language models through chatbots: Using the web interface to support
research tasks such as information retrieval, brainstorming, summarization of complex
concepts, and scientific writing assistance.
3. Prompt engineering techniques: Crafting effective inputs (prompts) for large language models;
using role-based (persona) prompting, structured outputs, text formatting with Markdown and
LaTeX, and tailoring responses for different audiences.
4. Advanced prompting patterns: Techniques such as flipped interaction (where the chatbot asks
questions first), template-based prompting, data extraction from texts, integration of external
information into prompts, and step-by-step question refinement strategies.
5. Information extraction and citation discovery: Using large language models to extract structured
information from unstructured text (e.g., grant calls or reports), and to identify relevant
academic references and generate citations.
6. Few-shot prompting and reasoning strategies: Prompting methods that include examples to
improve the model’s reasoning, such as chain-of-thought prompting, breaking tasks into steps,
and using examples to guide grading or explanation.
7. Limitations and ethical considerations: Understanding the risks of hallucinated or outdated
information, lack of real-world experience, and ethical concerns in using artificial intelligence
tools in academic research.
8. Demonstration of advanced use cases through the application programming interface:
Showcasing how chatbots can be used programmatically, including automated agents such as
AutoGPT.