Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

The Use of AI-Robotic Systems for Scientific Discovery

Alexander H. Gower, Konstantin Korovin, Daniel Brunnsåker, Filip Kronström, Gabriel K. Reder, Ievgeniia A. Tiukova, Ronald S. Reiserer, John P. Wikswo, Ross D. King

TL;DR

Automating scientific discovery through robot scientists couples AI with laboratory robotics to test hypotheses in real experiments. The authors map scientific reasoning to machine learning paradigms, arguing that active learning, not reinforcement learning, best captures discovery dynamics. They illustrate the approach with biology-focused case studies, culminating in Genesis, a next-generation system that uses 1000 micro-bioreactors and a first-order-logic model LGEM+ to generate and test hypotheses. The chapter discusses foundation models, their benefits and risks, and outlines future integration of computational models with hardware for closed-loop discovery.

Abstract

The process of developing theories and models and testing them with experiments is fundamental to the scientific method. Automating the entire scientific method then requires not only automation of the induction of theories from data, but also experimentation from design to implementation. This is the idea behind a robot scientist -- a coupled system of AI and laboratory robotics that has agency to test hypotheses with real-world experiments. In this chapter we explore some of the fundamentals of robot scientists in the philosophy of science. We also map the activities of a robot scientist to machine learning paradigms, and argue that the scientific method shares an analogy with active learning. We demonstrate these concepts using examples from previous robot scientists, and also from Genesis: a next generation robot scientist designed for research in systems biology, comprising a micro-fluidic system with 1000 computer-controlled micro-bioreactors and interpretable models based in controlled vocabularies and logic.

The Use of AI-Robotic Systems for Scientific Discovery

TL;DR

Automating scientific discovery through robot scientists couples AI with laboratory robotics to test hypotheses in real experiments. The authors map scientific reasoning to machine learning paradigms, arguing that active learning, not reinforcement learning, best captures discovery dynamics. They illustrate the approach with biology-focused case studies, culminating in Genesis, a next-generation system that uses 1000 micro-bioreactors and a first-order-logic model LGEM+ to generate and test hypotheses. The chapter discusses foundation models, their benefits and risks, and outlines future integration of computational models with hardware for closed-loop discovery.

Abstract

The process of developing theories and models and testing them with experiments is fundamental to the scientific method. Automating the entire scientific method then requires not only automation of the induction of theories from data, but also experimentation from design to implementation. This is the idea behind a robot scientist -- a coupled system of AI and laboratory robotics that has agency to test hypotheses with real-world experiments. In this chapter we explore some of the fundamentals of robot scientists in the philosophy of science. We also map the activities of a robot scientist to machine learning paradigms, and argue that the scientific method shares an analogy with active learning. We demonstrate these concepts using examples from previous robot scientists, and also from Genesis: a next generation robot scientist designed for research in systems biology, comprising a micro-fluidic system with 1000 computer-controlled micro-bioreactors and interpretable models based in controlled vocabularies and logic.

Paper Structure

This paper contains 20 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. The Philosophy of Robot Scientists
  3. Elements of scientific method
  4. Logical inference
  5. Statistical inference and probability
  6. Parsimony
  7. Comparing scientific models
  8. Scientific Discovery as Machine Learning
  9. Reinforcement learning is an unsatisfactory paradigm for robot scientists
  10. Supervised learning is a more useful paradigm
  11. Semi-supervised learning
  12. Active learning integrates agency into supervised learning
  13. Foundation models and their use in scientific discovery
  14. Biological Systems are a Good Target for Scientific Discovery Automation
  15. Discovery through superhuman logic and probabilistic reasoning
  16. ...and 5 more sections

Figures (2)

  • Figure 1: Flowcharts representing high level design for: generic machine learning; supervised learning; unsupervised learning; and reinforcement learning. Unitalicised text represents inputs and outputs; italicised labels for connectors represent processes.
  • Figure 2: Flowcharts representing the scientific discovery process of Genesis. The robot scientist starts with a systems biology theory, contructed from community knowledge. After using a computational model (e.g. LGEM+) to form hypotheses, Genesis will design and run lab experiments using its hardware, and will take measurements of phenotype using automated procedures, e.g. for metabolomics. Simulated phenotype will be compared against the observed phenotype to generate information used to refine the theory, and the cycle will begin again with the improved theory.