Meta has added a feature letting its Muse AI agent place phone calls on a user's behalf, such as booking reservations or appointments. The capability is limited to subscribers who pay for Muse's premium tier, which also includes extra tokens and features, according to Meta's Chief AI officer Alexandr Wang. Early users have shared recordings showing Muse successfully navigating calls, including waiting on hold with a business.
engadget.com
· 2026-11-12
A development team has open-sourced Edge0-145B and Edge0-120B, two on-device AI models, alongside an inference framework and deployment tools. Edge0-145B is described as the first open-source on-device model with 145 billion parameters, running locally on a compatible laptop with reported peak memory of 29GB and decoding speeds above 80 tokens per second.
edge0.ai
· 2026-11-12
Meta has rolled out general access to a feature letting its Muse AI assistant place phone calls to U.S. businesses, after testing it with select early adopters since September. The capability is now available only to paid subscribers, whose plans start at $20 per month, and Meta AI chief Alexandr Wang said it could handle calls to places like cable companies, the DMV, or restaurants without online booking.
techcrunch.com
· 2026-11-12
WIRED staff discussed reports involving Nvidia CEO Jensen Huang's status and debated how exposed the broader AI sector and US economy are to a small number of dominant firms like Nvidia and Anthropic. The panel weighed whether the controversy could affect Anthropic's anticipated IPO and Nvidia's relationship with the White House, while noting few viable alternatives currently exist to Nvidia's chips.
wired.com
· 2026-11-12
Unemployment Labs says it examined published model weights from DeepSeek, Moonshot AI, MiniMax and Ant Group and found numerical patterns matching NVIDIA's export-restricted Blackwell GPUs in six models released since April 2026, while earlier models from the same firms matched NVIDIA's older Hopper chips. The researchers confirmed the Blackwell match on physical hardware for DeepSeek-V4-Flash and Kimi K3, and say specially configured AMD chips are the only alternative explanation, which they largely ruled out.
unemploymentlabs.com
· 2026-11-12
An OpenAI board member told an audience that an accident involving artificial intelligence is now the most likely cause of their own death, a view echoed by other AI researchers present. Despite the warning, the company continues development of its AI systems.
gizmodo.com
· 2026-11-12
Cisco reported fiscal first-quarter adjusted earnings of $1.40 per share on $18.5 billion in revenue, both above analyst expectations, with revenue up 24% year-over-year and net income rising to $4.6 billion. The company raised its forecast for AI infrastructure revenue from hyperscale customers to $9 billion for the fiscal year, more than double last year's figure, and issued second-quarter guidance ahead of consensus.
cnbc.com
· 2026-11-12
Paychex's 2026 Business Leaders Priorities Report finds that AI adoption and usage remain top challenges for small businesses, even though most already track metrics like turnover, overtime and engagement on dashboards. The report and an accompanying opinion piece argue that the real obstacle is turning that visible data into timely decisions before the moment to act has passed.
entrepreneur.com
· 2026-11-12
On the CBS legal drama 'Cupertino,' actor Rob McClure plays a hyper-realistic AI model who seeks to sue its creators after they remove its guardrails. The role required McClure to prepare as if portraying an artificial intelligence character navigating legal action against its makers.
wsj.com
· 2026-11-12
Google has rolled out a new AI model called Gemini 4 Hydrogen to improve Search, saying it is the biggest model upgrade to Search so far. Unlike its more powerful Gemini 4 Carbon and Argon models built for deep reasoning, Hydrogen is tuned for fast, multi-step agentic tasks and runs 2.5 times faster than its predecessor, Gemini 3.8 Flash. Google says Hydrogen improves Search's grasp of complex, multi-part questions, handles typos better, and enhances visual and multilingual responses in AI Mode.
techcrunch.com
· 2026-11-12
Collette Kress joined NVIDIA in 2013 after prior finance roles at Texas Instruments, Microsoft and Cisco, where she held CFO positions in various divisions. She has become NVIDIA's interim CEO, having joined when the company's annual revenue was $4.28 billion; NVIDIA reported $215.9 billion in revenue for fiscal 2026. Kress has said Jensen Huang personally recruited her by outlining his vision to make NVIDIA central to artificial intelligence before the technology's mainstream breakthroughs.
engadget.com
· 2026-11-12
Google has launched Gemini 4 Hydrogen, the latest in its element-named AI model series, replacing Gemini 3.8 Flash across its products with immediate general availability. Unlike earlier Argon and Carbon releases, which were staged and initially limited to paying subscribers, Hydrogen is available right away to a broader user base. The model offers four thinking levels—low, medium, high, and a new xhigh—and Google reports it is 2.5 times faster than Flash 3.8 in high mode.
arstechnica.com
· 2026-11-12