Democratic secretaries of state seek social media companies' plans to moderate inflammatory content and AI during elections. Seven officials, including those from Maine and Washington, urge Google, X, and Meta to address the issue.
Summary: Introducing a new GraphRAG approach for efficient commercial contract data extraction and Q&A agent building. Focus on targeted information extraction and knowledge graph organization enhances accuracy and performance, making it suitable for handling complex legal questions.
AI vs human in cryptic crossword challenge. Times' competition sets gold standard. AI named Ross competes on Crossword Genius app.
Es Devlin wins $100,000 Eugene McDermott Award at MIT for innovative art exploring biodiversity, AI poetry, and more. Devlin's residency includes a public lecture and collaboration with MIT's creative community.
MIT researchers have developed 3D transistors using ultrathin materials to surpass silicon's energy efficiency limits. These nanowire devices offer high performance at lower voltages, potentially revolutionizing electronics.
W. E. B. Du Bois' legacy on systemic racism research at MIT continues with the ICSR Data Hub, providing vital criminal justice datasets for analysis and solutions. Computational technologies uncover racial bias in American society, offering hope for a more equitable future.
AI models, like LLaMA 3.1, require large GPU memory, hindering accessibility on consumer devices. Research on quantization offers a solution to reduce model size and enable local AI model running.
Dynamic execution in AI tasks can optimize performance by distinguishing between Hard and Easy problems. By identifying and addressing the complexity of data points, accuracy can be maintained while saving computational resources.
Generative AI is reshaping the art world, with humans still playing a crucial role in creating AI art. Rachel Ossip explores the intersection of technology and creativity in the evolving art landscape.
Tech platforms like Meta and Google could be legally responsible for AI-generated content. Google Maps introduces new features with Gemini, raising concerns about defamation risks from user comments used in AI responses.
AI-generated summaries by Google and Microsoft are being used in search engines, but the accuracy and authority of the information are questioned. The debate over the carcinogenic properties of aspartame highlights the potential pitfalls of relying on AI for contentious topics.
Thorsten Kleppe created a stunning visualization of the Pollen Dataset, showcasing its density and data distribution. The dataset contains 3,848 rows with predictors like "ridge" and "crack" to predict "density".
Enterprises face challenges managing costs of generative AI. Amazon Bedrock introduces tagging for on-demand models to enhance cost management and control.
LLMs can now be run locally for enhanced privacy and control over model settings, with various sizes available. Quantization reduces memory usage, while local implementations prove cost-effective compared to cloud-based solutions.
Evaluating AI-generated outputs of large language models is crucial for building robust applications, with supervised and unsupervised methods available. Self-evaluation and iterative self-reflection can improve the quality of generative models, reducing the need for human involvement in evaluations.