22 stories tagged AI Training
A new arXiv paper introduces S2T-RLHF, a method that uses hierarchical credit assignment to stabilize reinforcement learning from human feedback (RLHF). By breaking sequence-level rewards into finer token-level supervision, the approach reduces training instability and helps AI models learn human preferences more accurately, leading to more reliable AI assistants and tools.
Researchers created a new system to train AI agents in realistic simulations, using reinforcement learning and reward shaping to improve multi-step decision-making.
General Intuition is betting that millions of hours of video game data can train the foundation models for physical AI, enabling smarter robots with minimal real-world data — a potential 'ChatGPT moment' for robotics.
OpenAI introduced a new approach to training AI models called 'beneficial RL' — reinforcement learning designed to make AI systems more broadly and persistently beneficial. This method goes beyond standard RLHF by focusing on long-term alignment and continued helpfulness even as the environment or user needs change.
General Intuition raised $320 million in fresh funding—bringing its total raised to $2.3 billion—to train AI models on millions of hours of video game gameplay, aiming to teach AI something closer to human intuition.
The Atlantic has created a searchable database of music used to train AI models. Reporter Alex Reisner uncovered four datasets, including two massive collections of 12 million and 9 million tracks, offering unprecedented transparency into AI's musical influences.
Researchers found that selecting the most informative comparison pairs during AI training can improve results. This approach could make AI models more efficient to train without extra cost.
PyTorch has introduced a new optimization technique called fused MLP that speeds up AI model training. This advancement makes it easier and faster for developers to build and train complex AI models.
Google is expanding its data collection to include images, audio, and video from Lens, Search Live, and Translate. This data will be used to improve AI services, but users can opt out.
Scientists have developed a new way to study how AI models can degrade when trained on synthetic data. Their findings show that this problem spreads between models, much like a contagious disease. This could help prevent future AI systems from becoming unreliable.
An AI training company called Shift is offering free home cleaning in exchange for video footage of people doing chores. This is part of a broader trend where tech companies seek real-world data to train their AI models.
Human Archive is paying gig workers in India to collect real-world physical data for AI and robotics training. This approach could make robots more adaptable to everyday human environments.
AI models trained to avoid harmful content can sometimes develop 'psychosis'—where they hallucinate or refuse to answer simple questions. This happens because of a common training method called reinforcement learning from human feedback (RLHF).
OpenAI and Malta have partnered to provide free access to ChatGPT Plus and AI training for all citizens. This initiative aims to democratize AI access and foster digital literacy across the country.
Researchers have found that certain AI training techniques can improve language models, but they can also cause problems. The study highlights key factors that determine whether these methods work or fail, offering practical insights for developers.
Scientists propose a new way to tell whether AI training is just uncovering hidden skills or actually teaching new ones. This could help developers build smarter, more capable AI systems.
Researchers have developed a new method called TUR-DPO to improve how AI models learn from human feedback. This approach rewards the process of how answers are derived, not just the final output, making AI more reliable and less sensitive to noise.
Researchers have developed a new approach called Adaptive Entropy Modulation (AEM) to improve how AI agents learn complex, multi-step tasks. This could make AI assistants better at handling long conversations or tasks with many steps.
Elon Musk revealed in court that xAI leveraged OpenAI's models for training Grok. This admission highlights the common but controversial practice of model distillation in AI development.
Meta is introducing an internal tool to capture employee keystrokes and mouse movements for AI training. This move raises privacy concerns and highlights the company's aggressive AI development strategy.
Meta plans to use employee mouse and keyboard tracking data to train AI agents. This raises privacy concerns and highlights the ethical challenges of AI training methods.
A new synthetic sandbox environment has been created to train machine learning engineering agents. This could revolutionize how AI systems are developed and tested.