LLMs

Twitter

Facebook

Copy Link

Key Facts

Year2018

Key PlayersOpenAI, Google, Anthropic

Top News

Princeton University ends 133-year-old tradition as AI cheating forces return of exam proctors

Princeton University has abandoned its 133-year-old Honor Code tradition of proctor-free exams, marking one of the most significant policy reversals in higher education. The change comes after surveys revealed 30% of seniors admitted to cheating, with generative AI tools making academic dishonesty easier than ever. Starting July 1, human proctors will monitor all exams at the Ivy League institution.

Entertainment5 sources

Princeton University ends 133-year-old tradition as AI cheating forces return of exam proctors

Entertainment5 sources

Sipeed K3 RISC-V SBC runs 30B AI LLMs locally at 15 tokens per second for $299

Sipeed has launched its K3 series RISC-V-powered single board computers that can run 30B-parameter AI LLMs locally at up to 15 tokens per second. Built with SpacemiT's Fusion Architecture and featuring up to 32GB LPDDR5 memory plus a 60 TOPS NPU, the K3 offers an accessible open-source alternative to proprietary AI hardware starting at $299.

Technology2 sources

Sipeed K3 RISC-V SBC runs 30B AI LLMs locally at 15 tokens per second for $299

Technology2 sources

AI models corrupt 25% of documents in long-running tasks, Microsoft researchers reveal

Microsoft researchers tested 19 AI models across 52 professional domains and found even the most advanced LLMs introduce substantial errors in multi-step workflows. Frontier models like Gemini 3.1 Pro, Claude 4.6 Opus, and GPT 5.4 lost an average of 25% of document content over 20 interactions, with weaker models degrading by 50%. Only Python programming met the readiness threshold.

Technology2 sources

AI models corrupt 25% of documents in long-running tasks, Microsoft researchers reveal

Technology2 sources

Netflix tests AI-powered voice search that understands mood-based requests on smart TVs

Netflix is testing an AI-powered voice search feature with select US subscribers that lets them find content using natural language queries like "I need a good cry" or "help me stay awake." The beta feature works on Google TV devices but bypasses platform voice assistants, keeping users within Netflix's app while delivering surprisingly accurate viewing recommendations.

Technology4 sources

Netflix tests AI-powered voice search that understands mood-based requests on smart TVs

Technology4 sources

Anthropic Mythos finds 271 Firefox security vulnerabilities but cURL creator calls it marketing

Anthropic Mythos identified 271 security vulnerabilities in Firefox with almost no false positives, helping Mozilla ship 423 bug fixes in April 2026. But cURL creator Daniel Stenberg questions the hype after the AI model found just one confirmed vulnerability in his widely-tested codebase, calling it primarily a marketing stunt.

Technology7 sources

Anthropic Mythos finds 271 Firefox security vulnerabilities but cURL creator calls it marketing

Technology7 sources

Moonshot AI raises $2 billion at $20 billion valuation as Chinese AI labs attract major investors

Moonshot AI, creator of the popular Kimi chatbot, has closed a $2 billion funding round at a $20 billion valuation, led by Meituan's venture arm. The Beijing-based AI lab has raised $3.9 billion in just six months, with annual recurring revenue surpassing $200 million. The deal signals surging investor appetite for Chinese AI startups as they compete with OpenAI and Anthropic.

Startups5 sources

Moonshot AI raises $2 billion at $20 billion valuation as Chinese AI labs attract major investors

Startups5 sources

Siri gets complete AI-powered rebuild in iOS 27 with chatbot features and standalone app

Apple is finally delivering on its AI-supercharged Siri promise with iOS 27, announced at WWDC 2026. The virtual assistant will be completely rebuilt as an always-on agent with ChatGPT-like capabilities, featuring a standalone Siri app, Dynamic Island integration, and the ability to perform actions across apps using natural conversations.

Technology3 sources

Siri gets complete AI-powered rebuild in iOS 27 with chatbot features and standalone app

Technology3 sources

Subquadratic claims 1,000x AI efficiency with 12 million token context window, researchers skeptical

Miami-based Subquadratic emerged from stealth with $29 million in seed funding, claiming its SubQ model is the first large language model built on fully subquadratic architecture. The company says it reduces compute by almost 1,000 times at 12 million tokens compared to frontier models, but AI researchers are demanding independent proof of the extraordinary claims before validating what could be a major inflection point in how AI systems scale.

Technology2 sources

Subquadratic claims 1,000x AI efficiency with 12 million token context window, researchers skeptical

Technology2 sources

OpenAI launches GPT-5.5 Instant as ChatGPT's new default model, cutting hallucinations by 52.5%

OpenAI rolled out GPT-5.5 Instant as ChatGPT's new default model, claiming it reduces hallucinations by 52.5% on high-stakes prompts in medicine, law, and finance. The update delivers shorter, clearer responses while introducing a memory sources feature that shows users exactly what context informed their answers. Enhanced personalization pulls from past chats and connected services.

Technology13 sources

OpenAI launches GPT-5.5 Instant as ChatGPT's new default model, cutting hallucinations by 52.5%

Technology13 sources

Cybercriminals complain about AI slop flooding their forums as study reveals limited impact

A new study analyzing 97,895 conversations on cybercrime forums reveals an unexpected trend: cybercriminals are frustrated with AI-generated content flooding their communities. Researchers from the University of Edinburgh and University of Cambridge found that while AI adoption was hyped, it's mostly used for low-skill activities like SEO spam and romance scams rather than sophisticated hacking.

Technology4 sources

Cybercriminals complain about AI slop flooding their forums as study reveals limited impact

Technology4 sources

Major Publishers Sue Meta and Zuckerberg Over Alleged Copyright Infringement in AI Training

Five major publishers including Elsevier, Hachette, and Macmillan have filed a class-action lawsuit against Meta and CEO Mark Zuckerberg, alleging the company illegally used millions of copyrighted books and research papers to train its Llama AI models. The case marks the first AI copyright lawsuit brought by major scientific publishers and could reshape the legal landscape of AI development.

Policy22 sources

Major Publishers Sue Meta and Zuckerberg Over Alleged Copyright Infringement in AI Training

Policy22 sources

Reddit shares surge 16% as AI-driven ad growth delivers 69% revenue jump to $663 million

Reddit rallied in premarket trading after reporting a strong revenue forecast powered by artificial intelligence advertising tools. The social media company's AI-optimized ad platform helped deliver 69% revenue growth while daily active users climbed 17% to 126.8 million. CEO Steve Huffman positioned Reddit as essential fuel for AI development through its vast user-generated content library.

Business7 sources

Reddit shares surge 16% as AI-driven ad growth delivers 69% revenue jump to $663 million

Business7 sources

News Publishers Block AI Training Access to Web Archives Amid Copyright Infringement Concerns

Over 240 news organizations across nine countries are blocking web archives like Common Crawl and the Internet Archive's Wayback Machine to prevent AI companies from using their content without permission or compensation. Major outlets including The New York Times, CNN, and USA Today are restricting access, citing copyright violations as AI firms use archived news content to train large language models. The move raises concerns about preserving public records and historical accountability.

Policy3 sources

News Publishers Block AI Training Access to Web Archives Amid Copyright Infringement Concerns

Policy3 sources

AI Diagnosis Outperforms Emergency Room Doctors in Harvard Study, Raises Questions on Patient Care

A Harvard Medical School study published in Science reveals that OpenAI's o1 model achieved 67% diagnostic accuracy at emergency room triage, outperforming two attending physicians who scored 55% and 50%. The research tested AI against doctors using real patient data from Beth Israel Deaconess Medical Center, with blinded reviewers unable to distinguish AI output from human diagnoses. While the findings demonstrate AI's potential as a diagnostic tool, researchers emphasize the urgent need for clinical trials and accountability frameworks before deployment in actual patient care settings.

Science16 sources

AI Diagnosis Outperforms Emergency Room Doctors in Harvard Study, Raises Questions on Patient Care

Science16 sources

Meta's business AI tools hit 10 million weekly conversations as ad revenue surges

Meta revealed its business AI tools now facilitate 10 million conversations per week, a tenfold increase from early 2026. While currently free, CEO Mark Zuckerberg hinted at future monetization as the company expands AI-powered features across its platforms. The surge comes alongside strong adoption of GenAI ad tools by over 8 million advertisers, driving higher conversion rates and engagement.

Technology3 sources

Meta's business AI tools hit 10 million weekly conversations as ad revenue surges

Technology3 sources

Prev 1…567…25 Next