Themis AI enhances AI models to detect and correct uncertainties and biases, ensuring reliability in high-stakes applications. The MIT spinout offers a solution to improve AI models and prevent devastating consequences by forecasting failures before they happen.
Demis Hassabis of Google DeepMind aims to use AI to streamline email management, tackling inbox overload with next-gen email technology. The goal is to automate mundane tasks like sorting and replying to emails, ultimately reducing the stress of inbox management.
Research team from Olivetti Group and MIT CSHub use AI to find sustainable alternatives to cement in concrete, discovering ceramics and mining byproducts as viable options. Their machine-learning framework sorts through over 1 million rock samples to identify 19 types of materials that can reduce costs and emissions in concrete production.
Leo Anthony Celi of MIT addresses bias in AI training data, highlighting flaws and proposing solutions for more accurate models. He emphasizes the importance of teaching students to thoroughly evaluate data to prevent biases in AI applications.
Crafting personalized experiences boosts engagement and loyalty. Amazon Personalize uses ML to generate tailored recommendations, streamlining the process with MLOps practices.
AI machine learning analyzes handwriting-generated electrical signals to detect Parkinson's tremors using a 3D-printed magnetic ink pen. Early diagnosis aids in accessing support for the 10 million individuals worldwide living with Parkinson's disease.
SOPs are crucial in FDA-regulated industries, like healthcare and life sciences, to ensure compliance with regulatory standards. Using Amazon Bedrock, organizations can automate the alignment of SOPs with changing regulations, streamlining processes and reducing resources.
MIT and Stanford develop SketchAgent, an AI system that creates sketches stroke-by-stroke based on natural language prompts. The tool aims to revolutionize how humans communicate with AI through a more natural and iterative drawing process.
Training linear support vector regression (SVR) poses challenges due to the non-calculus differentiable loss function. Utilizing particle swarm optimization (PSO) proved more effective than evolutionary algorithms for training linear SVR models.
Meta, owner of Facebook and Instagram, will use AI tools to aid advertisers in creating targeted campaigns by end of next year, disrupting traditional marketing industry. The move threatens advertising and media agencies by allowing Meta to directly target brands' marketing budgets.
Salman Rushdie warns at Hay festival: AI lacks humor, but if it writes a funny book, "we're screwed." Rushdie admits he avoids AI, believing authors are safe until AI can create humor.
Utah lawyer Richard Bednar sanctioned for using ChatGPT to reference nonexistent court case in filing. Utah appeals court discovers false citations, leading to apology from Bednar.
Radio host replaced by avatars & artist copied by Midjourney - the impact of being replaced by bots. Mateusz Demski shares his experience.
Amazon SageMaker Projects streamline ML lifecycle with Terraform Cloud, removing CloudFormation dependency. AWS Service Catalog engine for Terraform Cloud enables direct deployment of SageMaker Projects.
ZURU Tech's Dreamcatcher platform simplifies building design and construction with AI, allowing users to order and assemble entire buildings. Collaborating with AWS, ZURU implemented a text-to-floor plan generator using a large language model for accurate design creation.