1.1 A Brief Guide to AI History and References#
We are not experts in AI history, but we think it is important to have context around how AI agents were developed. This page collects references we’ve found useful in understanding key developments and how we got to the AI agents we use today.
If you’re looking to learn more about the evolution of AI, large language models (LLMs), and their applications in science, the links below may help. Otherwise, we suggest skipping forward to Chapter 2 to the practical applications of AI agents in biology.
Foundational Papers#
These papers mark key developments in modern AI, from the invention of the Transformer to reasoning and agent capabilities.
Attention Is All You Need — Vaswani et al. (2017). The Transformer architecture that replaced recurrence with self-attention, becoming the basis for all modern large language models.
Scaling Laws for Neural Language Models — Kaplan et al. (2020). Showed that model performance improves predictably with more parameters, data, and compute. This gave a map for how to improve intelligence of models.
Language Models are Few-Shot Learners — Brown et al. (2020). Demonstrated that large language models can perform a wide range of tasks with only a few examples, without task-specific fine-tuning. Showed transformers have general intelligence.
Chain-of-Thought Prompting — Wei et al. (2022). Showed that prompting models to produce intermediate reasoning steps substantially improves performance on complex tasks. Showed that expending more inference compute and give better answers.
ReAct: Reasoning and Acting in Language Models — Yao et al. (2022). Interleaved reasoning traces with tool use, establishing a pattern for building autonomous agents from LLMs. The foundational agents paper.
DeepSeek-R1 — DeepSeek-AI (2025). Demonstrated the reinforcement learning can break out of the scaling laws and LLMs can improve simply be repeatedly attempting tasks rather than requiring worked-out examples.
AI Agents in Science#
Papers on building and using AI agents for scientific research, including from FutureHouse co-founders Andrew White and Sam Rodriques.
ChemCrow: Augmenting large-language models with chemistry tools – Bran, Cox, …, White, Schwaller (2023). Showed that LLMs equipped with chemistry-specific tools can autonomously plan and execute multi-step chemical tasks, establishing a template for tool-augmented scientific agents. (FutureHouse)
PaperQA2: Language agents achieve superhuman synthesis of scientific knowledge – Skarlinski, Cox, …, Rodriques, White (2024). Showed that a language agent optimized for factuality matches or exceeds human subject matter experts on literature research tasks. (FutureHouse)
Aviary: Training Language Agents on Challenging Scientific Tasks — Narayanan, Braza, …, Rodriques, White (2024). Introduced Aviary, a framework for multi-step reasoning and tool use across scientific tasks for language agents. Showed that language agents backed by open-source, non-frontier LLMs can match and exceed both frontier LLM agents and human experts on domain-specific scientific tasks. (FutureHouse)
A Review of Large Language Models and Autonomous Agents in Chemistry — Ramos, Collison, White (2025). A comprehensive review of how LLMs and autonomous agents are being applied across chemistry. (FutureHouse)
ether0: Training a Scientific Reasoning Model for Chemistry – Narayanan, …, Rodriques, White (2025). Demonstrated that reasoning capabilities can be post-trained into LLMs for chemistry without domain-specific pretraining, achieving strong performance with substantially less data than contemporary domain models. (FutureHouse)
Kosmos: An AI Scientist for Autonomous Discovery — Mitchener, Yiu, …, Rodriques, White (2025). Showed that an AI scientist can autonomously conduct data-driven discovery through iterative cycles of literature search, hypothesis generation, and data analysis, maintaining coherence over extended research trajectories. (FutureHouse)
Robin: A multi-agent system for automating scientific discovery — Ghareeb, Chang, …, White, Hinks, Rodriques (2026). A multi-agent system that automates the full cycle of scientific discovery — from hypothesis generation through experimental design to analysis — demonstrated with real biological experiments. (FutureHouse; arxiv)
Accelerating scientific discovery with Co-Scientist – Gottweis et al. (2026). Introduced a multi-agent system built on Gemini that generates and iteratively refines scientific hypotheses using tournament evolution and scaled test-time compute, with experimental validation in drug repurposing for acute myeloid leukemia. (Google; arxiv)
Autonomous biomedical research with an artificial intelligence agent — Huang et al. (2025). Built a general-purpose biomedical agent capable of performing diverse tasks across clinical, genomic, and biomedical domains within a unified framework. (Stanford/Biomni; arxiv)
Perspectives on AI History#
We’re not the only ones to summarize history of AI Agents, check out these other introductions to the history and landmark papers:
Ilya Sutskever’s Reading List — A curated set of papers recommended by the OpenAI co-founder for understanding deep learning foundations.
An Interview with Alex Krizhevsky — Krizhevsky discusses AlexNet and the deep learning revolution in computer vision. (video)
Geoffrey Hinton: The 60 Minutes Interview — The “godfather of deep learning” on the promise and risks of AI. (video)
Andrej Karpathy’s Intro to LLMs — A one-hour talk covering how LLMs work, from tokenization to training to inference. (video)
Additional Resources on LLMs and Deep Learning#
Here are AI for science topics that are good to know, but not directly related to LLMs and their use in AI agents:
7 AI Terms You Need to Know — Video primer on common AI terminology. (IBM)
Hugging Face LLM Course — Highly recommended. A hands-on course covering LLM fundamentals, fine-tuning, and deployment. (Hugging Face)
LLM Pre-Training and Custom LLMs — How LLMs are pre-trained from scratch and adapted for specific domains. (Databricks)
Aligning Language Models to Follow Instructions — OpenAI’s blog post on instruction-tuning with RLHF (InstructGPT). (OpenAI)
A Comprehensive Introduction to Fine-Tuning LLMs — Overview of fine-tuning strategies including LoRA and full fine-tuning. (Medium)
What is fine-tuning? — IBM’s explanation of fine-tuning techniques and when to use them. (IBM)
Explaining the Transformer Model — Video with a visual walkthrough of the Transformer architecture. (Grant Sanderson)
Attention Mechanism in Transformers — Video of a detailed explanation of how attention works inside Transformers. (Grant Sanderson)
Deep Learning for Molecules & Materials — Andrew White’s open textbook on ML for chemistry and materials. (FutureHouse)
Providing Feedback#
We’d love to hear from you! Whether you run into issues, have ideas for improving the tutorials, or want to suggest new topics, feel free to reach out.
Email us at: tutorials@futurehouse.org