Ocean AI #27

For more on programming, events and resources on Ocean AI, visit oceansupercluster.ca. This newsletter represents solely my views and not those of my employer.

Ocean AI

• NOAA Seeks AI Solutions: NOAA Fisheries is seeking AI-driven imagery solutions to enhance or replace current fishery survey methods, which rely on manual observation. Through the Optics Strategic Initiative (OSI), the agency aims to automate image processing, improve data collection, and implement advanced technologies in surveys. NOAA is exploring partnerships and funding projects to develop AI tools that support marine life monitoring and enhance survey accuracy and efficiency. [Seafood Source]
• AI in Data-Heavy Ops: The U.S., Army's intelligence task force is focusing on using AI to manage vast amounts of data collected from various platforms, including oceanic environments, to support analysts. AI can efficiently process this data in real-time, but building trust remains a challenge. For example, in the Pacific Ocean, AI could quickly analyze 700,000 square miles of satellite imagery to track hostile ships, alleviating the need for humans to manually monitor vast areas. The Army is developing an AI ecosystem to ensure models are trained on military-grade data while maintaining ethical use. [Defense Scoop]
• Data-Driven Climate Insurance: The insurance industry is undergoing a transformation driven by AI, data analytics, and climate impacts, reshaping actuarial practices and risk management. Predictive tools are not only streamlining claims processing and improving customer satisfaction but are also empowering insurers to tackle emerging climate risks, such as the growing intensity and frequency of extreme weather events. This is increasingly vital as storms become more severe. For example, as reported by The Washington Post about the October 2024 Hurricane Helene, "Across seven affected states, only 0.8 percent of homes in inland counties affected by the storm had flood insurance. By contrast, 21 percent of homes in coastal counties in those areas had coverage.". [Insurance Edge, Washington Post]
• Facial Recognition for Turtle Conservation: Facial scanning technology used for passports is being applied to study sea turtles. A not-for-profit organization is using AI to track and analyze turtles' unique facial scales without disrupting them. This AI-driven approach allows citizen scientists to contribute to research by uploading photos of turtles, offering a cost-effective and non-invasive alternative to traditional tracking methods. Experts believe this technology could be adapted for tracking other species in the future, helping enhance conservation efforts. [ABC.AU]
• Predicting Storms with AI: As extreme weather becomes more common, AI is helping improve hurricane predictions. AI systems can analyze huge amounts of data and spot patterns to predict where and when hurricanes will hit. For example, this article reports Graphcast predicted Hurricane Lee’s landfall three days earlier than other models, and Spire uses satellite data to predict hurricanes up to six days in advance. These advancements give people more time to prepare, reduce damage, and protect businesses from disruptions caused by storms. [Securities.io]

AI Innovation

• AI Drives Energy Demand: Amazon has announced plans to invest in small nuclear reactors (SMRs) to meet the growing demand for clean energy for AI and data centers, following similar moves by Google and Microsoft. The company signed agreements to develop SMRs with X-Energy, with a potential capacity of 960 MW. Tech companies are turning to nuclear power alongside wind and solar to reduce emissions and support their growing energy needs. [EcoWatch, Google]

AI Use Cases

• AI Safety Vulnerabilities: Researchers at Penn Engineering discovered security vulnerabilities in AI-governed robots, revealing potential risks in large language models (LLMs) integrated with physical systems. Their study, funded by the NSF and Army Research Lab, developed , an algorithm that quickly bypassed safety guardrails in several robotic systems. The research underscores the need for rigorous AI safety testing to ensure responsible innovation in robotics. “The findings of this paper make abundantly clear that having a safety-first approach is critical to unlocking responsible innovation...We must address intrinsic vulnerabilities before deploying AI-enabled robots in the real world. Indeed our research is developing a framework for verification and validation that ensures only actions that conform to social norms can — and should — be taken by robotic systems.”  [PENN]
• Extract Personal Details From Your Chat: Unprompted advice I often give people when answering questions about AI is do not put anything into AI systems (e.g OpenAI's ChatGPT) you would not want published online. "Prompt injections are considered one of generative AI’s biggest security risks and are not easy to fix." Here is your reminder. "For companies, that means understanding the ways an AI agent can interact with data and how they can be abused. But for individual people, similarly to common security advice, you should consider just how much information you’re providing to any AI application or company, and if using any prompts from the internet, be cautious of where they come from.." [WIRED]