OpenAI and MIT conducted studies on ChatGPT's emotional impact, finding varied responses among users. Women were less social after use, while gender-swapping led to higher loneliness. Further research is needed to understand the emotional interactions with AI.
Leaked chats reveal detailed plans by General Services Administration staff, prompting anger among employees.
A leaked memo reveals plans by USAID to adopt blockchain technology for aid distribution. Experts question the necessity and potential negative impact on NGO partners. Staffers at USAID concerned about future under Trump administration's plan to fold agency into State Department.
Developers are using Minecraft as a creative benchmark for AI models due to its familiarity and ease of evaluation. MC-Bench allows users to compare models by voting on their Minecraft creations, providing a unique way to assess generative AI capabilities beyond conventional benchmarks.
Chinese AI models like DeepSeek censor politically sensitive topics, but compliance varies by language. American models also show different responses based on language, highlighting the challenges of cultural context in AI development.
OpenAI introduces new transcription and voice-generating AI models with improved realism and customization options. The models aim to enhance user control and accuracy in speech synthesis and transcription.
Nvidia partners with EPRI and tech companies to use AI models to address power grid issues caused by rising demand from AI. The Open Power AI Consortium aims to develop solutions to the growing electricity demands, with focus on renewable energy and efficient grid management.
Audience Choice voting is open for TechCrunch Sessions: AI, with six finalists for breakout session speakers. Topics include monetizing AI, diversity in tech, secure AI deployment, AI policy, safeguarding AI, and identifying exceptional AI founders. Vote for your favorite session now!
Microsoft is adding 475 megawatts of solar power to its data centers in the Midwest. Tech companies are turning to quick and inexpensive renewable energy sources to meet the growing demand for computing power, using solar with battery storage for continuous electricity supply. The shift towards renewables is outpacing new nuclear power plants due to cost and speed advantages.
Meta is launching its AI assistant, Meta AI, in the European Union with limited features due to regulatory concerns. The tool will be available in several EU countries and on platforms like Facebook, Instagram, and WhatsApp. Despite regulatory challenges, Meta aims to expand AI capabilities in Europe over time.
Satellite internet providers are expanding rapidly to meet the demand for AI-powered agents that need constant connectivity. Companies like Starlink, Amazon, and Google are launching satellites to provide global coverage for IoT devices and AI innovations. Interoperability is key to ensuring seamless connectivity, but concerns about sustainability and speed remain. Ultimately, satellite internet may become a standard feature in all phones.
European startup Pruna AI is making its AI model optimization framework open source, offering efficiency methods like caching and quantization. The framework evaluates compressed model quality and performance gains. It aims to offer a comprehensive tool for various compression methods, helping developers save money on inference costs.
Dubai-based startup ClearGrid, backed by $10 million, uses AI to modernize debt collection in emerging markets. It automates processes, improves recovery rates, and aims to build trust with borrowers. It has already gained traction in the UAE and plans to expand to Saudi Arabia.
OpenAI has launched o1-pro, a more powerful AI model in its API for select developers at a high cost. Despite some initial struggles, it aims to offer better responses to complex problems.
Noam Brown of OpenAI discusses the potential for earlier development of "reasoning" AI models had researchers known the right approach. He emphasizes the importance of academia in exploring less computing-intensive areas and collaborating with frontier labs. Brown also highlights the need for better AI benchmarks.
Raj Aggarwal is leaving AWS after leading generative AI efforts. He plans to start a new company after successful stints at Localytics and Demand Sage.
The California-based policy group led by Fei-Fei Li suggests laws requiring transparency in AI model development to address potential unseen risks. The report emphasizes the need for proactive AI regulation to prevent extreme consequences. Various industry stakeholders support the report, which is seen as a positive step for AI safety governance in California.
Nvidia CEO Jensen Huang emphasizes the importance of getting the fastest chips for serving AI in cloud providers, announcing new systems and dismissing competition from custom chips.
Elon Musk's AI company xAI adds image generation capabilities to its API with a model generating up to 10 images per request at $0.07 each. xAI aims for revenue growth and potential $10 billion funding round.
Researchers propose a new AI scaling law called "inference-time search," claiming it boosts model performance. Experts are skeptical due to limitations in real-world applications. Further research is needed in the quest for more efficient scaling techniques in AI.
