Tackling Emergent Science Questions
Artificial Intelligence (AI) systems are computer programs with the (limited) ability to reason like humans, learning tasks from a collection of data. Some are quite general, like the GPT5 large language model, others are designed to tackle a particular problem, such as DeepMind’s GraphCast weather model.
There are many opportunities for AI development and application in ocean sciences. NOC is beginning to engage with AI, emphasising infrastructure development and staff upskilling with several early-stage projects tackling emergent science questions.
How is NOC Harnessing Artificial Intelligence?
Modern AI models are tuned on large pools of data to learn a mathematical mapping from a set of inputs to a desired output. This process relies on well-managed, easily accessible data and GPU-accelerated compute. NOC is supporting AI projects by building digital infrastructure to manage data for AI, provide on-site compute resources, and lower barriers to entry.
Key AI Projects
The Data Science Platform (DSP) provides easy to use and scalable compute to NOC staff. The centrally managed service maintains preconfigured compute environments, including for AI, and provides seamless access to on-site and cloud resources.
CRAB (Centralised Repository for Annotations and BLOBs) is a framework for collection of biodiversity observations and annotations for use in machine learning. It can process arbitrary data, such as images, audio, and even 3d holography reconstructions. It's designed for seamless integration with cloud computing, implementing the MLCroissant metadata format for automatic discovery of dataset structure. CRAB aims to streamline the data processing pipeline for biodiversity observations of all kinds, and enable the use of ML techniques previously hindered by the sheer quantity of data to process.
BRAIN is a hardware and software stack for providing ML processing capability on-board Autosub Long Range. Running off a modified Jetson Orin AGX on-board ALR, we have demonstrated this capability in shoreside tests, and are soon to test this capability at sea. Below is an image straight off the device, running in near real-time with old ALR benthic imagery fed in as input. We hope that alongside computer vision tasks, this tool could be used for autonomous mission planning, adaptive sampling and multi-vehicle autonomous missions.
Dive Deeper: Into the Blue Podcast
The emergence of Artificial Intelligence (AI) in society in recent years has given us a glimpse into the future - but how does it help us monitor and learn more about our ocean? Professor Sara Bernardini and Dr Fatma Jebri tell us more about how we use AI in NOC's research and how they see it progressing in the future.
News Splash: The New Genetic Testing Approach to Predict Toxic Algae Using Artificial Intelligence
NOC scientists have developed a genetic testing approach to be used alongside artificial Intelligence (AI) to test for toxic algae that can potentially result in severe and fatal sickness in humans. Watch our News Splash video from 2023 to learn more.