Scientists at MIT and the MIT-IBM Watson AI Lab have developed a new approach to teach computers to pinpoint actions in videos using only transcripts. This method, called spatio-temporal grounding, improves accuracy in identifying actions in longer videos and could have applications in online learning and healthcare.
US tech startup OpenAI establishes safety and security committee for critical decisions. New AI model in development to replace ChatGPT system.
MIT CSAIL and Google Research introduce Alchemist, a system that can alter material properties in images with a unique interface. The system could enhance video game models, AI visual effects, and robotic training data, offering precise control over attributes like roughness and transparency.
Google uses entity resolution to match products across platforms, helping e-commerce companies with competitor analysis and price comparison. Entity Resolution (ER) framework aids in detecting duplicate listings and setting competitive prices in the retail space.
The new TunedThresholdClassifierCV in scikit-learn 1.5 optimizes decision thresholds for better model performance in binary classification tasks. It helps data scientists enhance models and align with business objectives by fine-tuning thresholds based on metrics like F1 score.
UK promotes 'Bletchley effect' of safety institutes at AI summit. Top scientist warns big tech distracts from AI risks.
Elon Musk's xAI secures $6bn investment, valued at $18bn pre-investment, challenging OpenAI with cutting-edge LLM technology. Startup advances generative AI for human-like text, pictures, video, and voices in just one year.
Experts divided on human-machine relationships: Embrace benefits vs. heed Hollywood warnings about AI relationships. Chatbots offer comfort, despite cautionary tales from movie 'Her'.
AI tools like LLMs make complex HR analytics more accessible, enhancing organizational understanding and predicting network dynamics. By combining network analysis and psychology, organizations can gain deeper insights into leadership, turnover, and team performance.
Scarlett Johansson upset over ChatGPT update mimicking her voice, sparking tech ethics debate. Hollywood tensions rise over rapidly advancing AI technology.
The 96th Scripps National Spelling Bee showcases discipline and focus in a world reliant on technology. Despite AI advancements, the beloved Bee continues to captivate globally, raising questions about the value of human skills in an automated world.
LangChain's built-in metrics for AI output correlate helpfulness with coherence and controversiality with criminality. The study suggests users prefer concise over detailed responses in certain contexts.
OpenAI unveils GPT-4o, a more versatile and user-friendly large language model, showcasing its ability to interact in voice, text, and vision. The live event highlighted features like mid-sentence interruptions, low latency, and emotional sensitivity, with amusing interactions between tech bros and the machine.
Outlier detector method supports categorical data, provides explanations for flagged outliers, emphasizing need for interpretability in outlier detection. Identifying errors, fraud, unusual records in various datasets crucial for practical applications in business, scientific discovery.
Data quality management is evolving rapidly in the AI era, with monitoring, testing, and observability as key elements. As data environments grow, traditional methods may not be enough to ensure data trust in 2024.