Artificial Intelligence in Laboratory Automation
AutoPilot is Artificial Intelligence software that processes scientific assay requirements and tells the available automated systems exactly all of the actions that should be taken to achieve good results.
The enormous amount of knowledge generated by laboratory automation researchers has created the means to perform science using automated systems. To realize this in the laboratory requires the application of this extensive knowledge combined with the intricate expertise of the scientists. Additional layers of complexity are added by the ever increasing number of available automated systems as well as the evolving field of computer science. It can be said then that what allows groups to achieve good scientific results using automated systems is that they know a lot of things about a lot of things and are able to apply that vast knowledgebase appropriately to their specific environment.
Handing Off the Knowledge to Autonomous Systems
An area of Artificial Intelligence (AI) known as Knowledge Representation and Reasoning (KRR) is concerned with how knowledge can be represented in a form that an Agent (e.g. computer system) can use it to perform a task intelligently. Instead of trying to understand or build brains from the bottom up, its goal is to understand and build intelligent behavior from the top down, putting the focus on what an agent needs to know in order to behave intelligently, how this knowledge can be represented symbolically, and how automated reasoning procedures can make this knowledge available as needed.
The knowledgebase required to automate science is vast but finite. Therefore, the AI used to process it does not need to create new knowledge but instead only correctly access and process the existing data. KRR provides the architectural foundation for AutoPilot to autonomously shoulder the task of applying the correct data to achieve good science using automation.
Autonomously Controlling Lab Automation
The decision rules for applying the high-quality knowledge are readily codified by automation experts but are frequently unavailable to end users in the laboratory. Therefore, the conditions are ripe for automating the decisions. At AutoPilot’s core is an automated decision making system named VAST (the Virtual Automated Sciences Technology). VAST provides the means for generating conclusions from available knowledge, using propriety logical techniques, and communicates those conclusions to AutoPilot for implementation on laboratory automation systems.
The knowledge available to VAST is extensive but incomplete. VAST is governed by Negation By Failure rules to ensures that any gaps in AutoPilot’s knowledgebase do not result in incorrect automation instructions. In other words, if VAST doesn’t know something it assumes that it is wrong. This approach provides the framework to reason about dynamic processes while maintaining appropriate responsiveness and control. The result is that AutoPilot only directs actions that are known to produce good results.