Everything tagged ai-training, newest first. Tags come from the classifier reading each item's summary; 394 tags used 25 times or more have their own page.
AfterQuery, an AI model-training startup, reportedly raised a new funding round valuing it at $3.2 billion, marking a dramatic jump from its $300 million valuation just five months earlier.
xAI has been accused of training its Grok AI model on material including real and AI-generated child sexual abuse content, raising serious legal, ethical, and safety concerns about data sourcing and model governance.
Published authors may have unknowingly helped train AI tools that could threaten their livelihoods, raising concerns about consent, legality, and the fairness of using their work without permission.
Cloudflare announced that, starting September 15, it will default-block “mixed use” search crawlers that also feed AI training, potentially disrupting Google, Bing, and Apple because many websites rely on Cloudflare and can now prevent indexing bots from doubling as data scrapers.
AI companies are discreetly purchasing rare books in large quantities, encountering pushback from booksellers concerned about the impact on the market and preservation.
Model collapse is a critical challenge in AI development where repeated training on AI-generated content leads to loss of rare information and increasingly inaccurate outputs, threatening the fidelity of future AI systems.
The resolution of a specific legal case does not address the larger concern regarding the use of copyrighted materials for AI model training.
Google has updated its privacy settings to allow its AI to train on more user data, but users can opt out of this feature.
Kaytus has introduced its All-QLC Flash Storage Solution at AI EXPO KOREA 2026, designed to support ultra-large-scale AI training with over 10,000 GPUs, addressing bottlenecks in modern AI infrastructure.
Phil Gervasi's article explores the concept of rail-optimized networking, focusing on its application in enhancing data center efficiency for AI training workloads.
Clos-based leaf-spine architectures have dominated data center networking for their scalability and integration capabilities, but large-scale AI training reveals their limitations.
Deccan AI focuses its workforce in India to ensure quality control in the rapidly expanding and fragmented AI training industry.
Delivery couriers can now earn additional income by engaging in activities such as filming daily tasks or recording themselves speaking different languages.
Researchers from top Chinese institutions have developed a humanoid robot, the Unitry G1, capable of playing tennis using AI that learns from imperfect amateur motion data, overcoming challenges of real-time coordination and movement.
A Harry Potter data set was erroneously labeled as public domain and subsequently removed after the mistake was identified.
Meta has dismissed a lawsuit alleging it used pirated pornography to train its AI, stating the claim is baseless.
Meta's Disaggregated Schedule Fabric (DSF) is an advanced network fabric technology designed to overcome limitations of Clos-based networks, enabling efficient AI training through mega clusters and exploring new switching technologies.
At the OCP Summit 2025, Meta unveiled advancements in network fabrics for AI training clusters, expanding their hardware portfolio and contributing new disaggregated network platforms to the Open Compute Project.
Rodney Brooks, co-founder of iRobot and former MIT professor, criticizes companies like Tesla and Figure for attempting to teach robots dexterity through human task videos, labeling this method as "pure fantasy thinking."
Startups are developing reinforcement learning (RL) environments to aid AI labs in training agents, potentially sparking a new trend in Silicon Valley.
Anthropic faces a financial penalty for illegally downloading books instead of purchasing them, not for using writers' work to train AI.
Anthropic is implementing significant alterations in its user data management policies, requiring users to respond by September 28.
Meta is facing a lawsuit alleging that it distributed adult content to minors and concealed piracy activities to train its AI models.
A federal judge ruled in favor of Meta in a lawsuit by 13 authors, including Sarah Silverman, who claimed Meta illegally used their copyrighted works to train AI models.
A judge's groundbreaking decision compares the process of training artificial intelligence to the way schoolchildren learn to write, emphasizing the educational nature of AI development.
EU residents have a limited timeframe to opt out of contributing their data to Meta's AI training programs, with a deadline approaching in less than two weeks.
A judge minimized the significance of Meta's improper use of torrenting in a lawsuit concerning the training of its AI models.
Innodisk has introduced its PCIe Gen5 SSD series, aligning with OCP Data Centre NVMe SSD spec v2.0, to cater to AI model training, big data analytics, and data-intensive environments with support for multiple form factors.
OpenAI faces accusations of using unlicensed, non-public books to train advanced AI models, raising concerns about copyright infringement.
Nvidia has acquired Gretel, a San Diego startup specializing in synthetic AI training data, for a nine-figure sum exceeding $320 million.
Meta is testing a custom AI-training chip, developed with TSMC, to decrease dependency on external hardware providers like Nvidia.
Recent court filings in an AI copyright case against Meta suggest the company halted licensing discussions with book publishers for training data.
Elon Musk and other AI experts agree that the available real-world data for training AI models has been nearly exhausted, reaching the cumulative sum of human knowledge by last year.
A class action lawsuit threatens Photobucket's strategy to monetize old user photos by using them to train artificial intelligence models.
A new self-evolving large language model by Writer can update itself post-deployment, potentially reducing AI training costs and improving real-time knowledge accuracy.
OpenAI is being sued by The New York Times and Daily News for allegedly using their content without permission, and has reportedly deleted potentially relevant data.
FrontierMath keeps its challenging questions unpublished to prevent AI companies from using them for training purposes.
The rapid advancement of humanoid robotics, exemplified by XEG's realistic robot, has sparked disbelief, while AI skills training offers a path to financial security amid rising unemployment.
Meta unveiled new network fabric innovations for AI training clusters at OCP 2024, enhancing hardware collaboration and open designs.
AI models can be trained on synthetic data generated by other AIs, a concept gaining traction due to data scarcity.
Meta and X have temporarily halted certain AI training activities due to shared concerns about potential risks.
The EU is investigating Google's compliance with data protection laws regarding the use of personal information for AI training.
PyTorch advancements for LLMs were discussed at AI Infra @ Scale 2024, highlighting new features enhancing large-scale training and memory efficiency.
An AI has been developed that can beat humans at ping-pong by learning to play in a simulated environment.
The utilization of motion-capture actors' performances in AI training raises significant concerns and debates in the industry.
