Alibaba's Qwen team recently shared a preview of the latest OpenAI o1 rival. QwQ-32B-Preview leverages the test-time compute technique to deliver enhanced math and coding skills. However, the team notes that QwQ struggles with common sense reasoning and nuanced language understanding.
A recent study by the Tow Center for Digital Journalism revealed that ChatGPT's new search capabilities are unreliable, correctly identifying sources less than 25% of the time, frequently fabricating information, and potentially misattributing or citing plagiarized content from various news outlets.
ElevenLabs has added a new feature to its Reader mobile app that allows users to generate AI-powered co-hosted podcasts based on any text source they have imported into the app. The GenFM is currently available for iOS Reader app users and will roll out for Android users shortly.
Inflection has acquired three enterprise AI startups to enhance its enterprise platform's capabilities in automation, data analysis, and communication while shifting its strategy to focus exclusively on enterprise products and moving away from competing in the foundation model market.
A group of artists protested OpenAI's Sora video generation technology by creating a front-end tool on Hugging Face, criticizing the company for allegedly exploiting artists through unpaid labor and restrictive access to the AI video generation platform.
IMAX has partnered with Camb.ai to use AI-powered speech translation technology to localize its original content. IMAX plans to release translated content in stages, beginning with high-resource languages.
The Allen Institute for AI (AI2) has released the OLMo-2 model family, featuring fully open-source 7B and 13B-parameter models that demonstrate competitive performance against other open-weight models through innovative training techniques and transparency in model development.
Some educational institutions, like the University of South Australia, are reconsidering the usefulness of oral assessments after facing the challenges of evaluating students using traditional written methods at a time when generative AI can quickly output plausible-sounding prose.
Anthropic has open-sourced the Model Context Protocol (MCP), an innovative standard designed to help AI systems break through data silos by providing a universal, secure framework for connecting AI assistants to diverse data sources and enterprise tools.
The US sees China as a contender in the AI race; DeepSeek launches R1-Lite-Preview, an open-source "reasoning" model; Amazon has invested an additional $4B in Anthropic; AI announcements in Microsoft's Ignite 2024; OpenAI launches Advanced Voice Mode for the web; and more.
Anthropic and Amazon have recently expanded their partnership, with Amazon making another $4B investment in the startup. Correspondingly, Anthropic has named AWS its main cloud and training partner, and committed to contribute to optimizing AWS's hardware and software stack.
OpenAI has expanded the availability of Advanced Voice Mode to the web experience for paid subscribers, who can talk to ChatGPT right from their browser. OpenAI CPO Kevin Weil confirmed the company is looking to enable free users to use AVM from their browser soon.