AI News

NIST faces internal unrest over the potential appointment of Paul Christiano to the US AI Safety Institute, with staff threatening to resign due to concerns about his ties to the effective altruism movement affecting objectivity. Critics question the transparency of the hiring process, risking the loss of key scientists and the institute's integrity.

A paper from 23 AI experts emphasizes the need for 'safe harbor' protections to facilitate independent evaluations of AI products. It critiques major AI companies' restrictive terms of service that hinder research, urging for legal changes to support public interest investigations while ensuring transparency and safety in AI development.

Abacus AI has launched Liberated-Qwen1.5-72B, an uncensored, AI-driven open-source LLM tuned for real-world applications, particularly in customer support. It outperforms prior models on benchmarks, yet remains uncensored, necessitating user-implemented safeguards. The model is available under a near-MIT license.

Cognizant has launched an AI lab in San Francisco to accelerate AI commercialization, addressing market demand and executive concerns. Led by Babak Hodjat, the lab focuses on innovative AI applications in business, expanding its team and partnerships. Cognizant is investing $1 billion in generative AI, aiming to enhance productivity and shape the future of work.

Inflection AI launched Inflection-2.5, a foundation model aimed at improving performance in STEM subjects and competing with OpenAI's GPT-4. This new model enhances emotional intelligence and user experience via the Pi assistant. It achieves 94% of GPT-4's performance with efficient training.

Senate Majority Leader Chuck Schumer announced $10 million for the National Institute of Standards and Technology to establish the U.S. Artificial Intelligence Safety Institute, enhancing AI safety and accountability. Despite funding concerns and criticisms over transparency, the initiative aims to develop AI standards and collaborate with 200 organizations.

Nvidia's GTC 2024, taking place March 18-21 in San Jose, anticipates over 16,000 in-person attendees and 300,000 online participants. CEO Jensen Huang will keynote, focusing on generative AI and upcoming certifications. The event features 1,700 presenters, discussions on responsible AI, and highlights community engagement.

Hugging Face is launching a new open-source robotics project led by ex-Tesla scientist Remi Cadene, who seeks engineers in Paris. The initiative focuses on developing low-cost robotic systems integrating AI. This marks a shift for Hugging Face, previously software-centric, into robotics amidst growing industry interest.

A former Google engineer, Linwei Ding, was arrested for allegedly stealing over 500 AI-related trade secrets to share with Chinese firms. He faces up to 40 years in prison for theft. This case highlights ongoing U.S.-China tensions over AI and trade secret thefts, reflecting national security concerns.

Meta is revamping its video recommendation system on Facebook, using AI technology to unify algorithms across Reels and other video content by 2026. This strategy, announced by Tom Alison, aims to enhance engagement and compete with TikTok, showing a 8-10% increase in watch time during trials.

Zapier has launched Zapier Central, an AI workspace allowing users to create bot workflows using natural language, with no coding needed. This feature distinguishes it from competitors like OpenAI’s GPT Builder. The company also acqui-hired Vowel’s team to boost its AI initiatives.

InvGate has launched an AI Hub that transforms incident response conversations into knowledge base articles, enhancing efficiency for IT teams. Using advanced AI models, it allows users to create relevant, compliant content easily, improving support processes for companies like NASA and McDonald's.

Content creators lead busy lives, often dedicating over 20 hours a week to produce fresh content for their unique online spaces. This commitment leaves little room for meaningful audience engagement. However, Mark Zuckerberg, CEO of Meta, believes that artificial intelligence (AI) could provide a solution.

Google will retire Duplex on the Web at the end of 2022 to focus on enhancing Duplex voice technology, citing user feedback. Launched in 2019, it allowed Android users to make automated online purchases, but complexities in training AI for web navigation impacted its effectiveness.

Sogou is enhancing audiobooks with AI-generated avatars of authors Yue Guan and Bu Xin Tian Shang Diao Xian Bing, using advanced text-to-speech technology. This aims to create a more lifelike listening experience, although previous AI news anchors showed limitations in realism.