In traditional information systems, preservation is paramount, deletion represents data loss. MemGPT challenges this paradigm by implementing strategic forgetting through two key mechanisms: summarization and targeted deletion.
Read More
Can AI have a conscience? Of course, today’s AI isn’t a sentient being with feelings or guilt. It won’t lose sleep over a tough decision. But as artificial intelligence plays a bigger role in our lives, we do expect it to act responsibly. In essence, we want AI to follow ethical principles, a sort of programmed “conscience” so that it helps society without harming it. This is the crux of AI ethics, an increasingly important topic now that machine learning systems are making decisions that matter.
Read More
The enduring human quest to understand if we are alone in the universe has driven the development of the Search for Extraterrestrial Intelligence (SETI) programs. These programs employ a diverse toolkit of techniques aimed at detecting potential technosignatures, or signals indicative of technological activity beyond Earth. Despite these advancements, traditional methods face inherent limitations, particularly regarding the selection of appropriate models to interpret experimental data. Traditional model selection methods often rely on researcher intuition and experience, introducing subjectivity and potential biases. To address these limitations, the Finite Information Quantity (FIQ) approach is presented as an objective solution for model selection. FIQ acknowledges the inherent constraints in information processing capacity within physical systems, providing objective criteria for evaluating different models. By considering comparative uncertainty, FIQ minimizes bias and promotes a more comprehensive understanding of phenomena.
Read More
We examined the reliability of image recognition services by conducting a cross-comparison across three popular image recognition services: Google Vision AI, Microsoft Azure Computer Vision and AWS Rekognition.
Read More
AI (Artificial Intelligence) is coming at humanity fast and furious. Shalini Urs discusses the new epoch of AI with one of the fathers of the Internet, Vint Cerf.
Read More
Critical data modeling uses tools from data modeling and systems analysis to analyze information systems, connecting social critiques to the structure of digital information systems and data objects.
Read More
With data as the infrastructure, and machine learning and deep learning as the technology foundations of building AI systems combined, they largely determine the performance, fairness, robustness, reliability, and scalability of AI systems.
Read More
Have you ever wondered how Google helps you complete your search query by suggesting the next terms of your query? Large Language Models (LLMs) power this feature. But LLMs go beyond that feature. Today, LLMs are used in building AI systems and applications ranging from recognizing speech to writing poetry. They have become very powerful, but there are also pitfalls.
Read More
In this episode of InfoFire—the fireside chat series from Information Matters—a conversation with Gio Wiederhold, professor emeritus of computer science, electrical engineering, and medicine at Stanford University on “Different Flavors of AI.”
Read More
We all want fairness in our lives. We want to everyone to pay their fair share, get a fair shot at opportunities, and provide fair explanations to life-changing decisions. But what we can’t agree on is what is this fairness and how to implement it in everyday systems we use.
Read More
