LIVE PULSE
3.9 Anthropic CEO Amodei calls for slower AI development and shared safety rules — 11 src2.1 Agility Robotics unveils Digit 5 humanoid for warehouses and factories — 2 src2.0 Apple ships rebuilt Siri with Google Gemini, but not in the EU — 2 src1.7 Siri AI in macOS 27 Golden Gate: FAQ, Germany availability, privacy questions — 2 src1.4 Sam Altman says OpenAI will not go public in 2026, citing AI safety concerns — 5 src1.1 OpenAI contractors review real ChatGPT conversations to rate responses, report says — 2 src1.1 Anthropic data retention policy prompts firms to limit Claude use for sensitive work — 1 src1.1 Study examines issue bias in LLMs used as writing assistants before Swedish 2026 election — 1 src1.1 Study Audits Misalignment in Multi-Modal World Models — 1 src1.1 Retrieval-Grounded Reasoning Approach Proposed for Universal Multimodal Embeddings — 1 src3.9 Anthropic CEO Amodei calls for slower AI development and shared safety rules — 11 src2.1 Agility Robotics unveils Digit 5 humanoid for warehouses and factories — 2 src2.0 Apple ships rebuilt Siri with Google Gemini, but not in the EU — 2 src1.7 Siri AI in macOS 27 Golden Gate: FAQ, Germany availability, privacy questions — 2 src1.4 Sam Altman says OpenAI will not go public in 2026, citing AI safety concerns — 5 src1.1 OpenAI contractors review real ChatGPT conversations to rate responses, report says — 2 src1.1 Anthropic data retention policy prompts firms to limit Claude use for sensitive work — 1 src1.1 Study examines issue bias in LLMs used as writing assistants before Swedish 2026 election — 1 src1.1 Study Audits Misalignment in Multi-Modal World Models — 1 src1.1 Retrieval-Grounded Reasoning Approach Proposed for Universal Multimodal Embeddings — 1 src
HEATPULSEAI MAGAZINES
FLIP · FOLLOW · SAVE
Log inSign up

MUSE

model6 events
papersTODAY 04:00 UTC

K-Bench benchmark evaluates LLM unlearning in agentic settings

A new arXiv paper introduces K-Bench, a benchmark designed to test whether unlearning holds up when language models act as agents rather than just answering questions directly. The authors argue that existing benchmarks like TOFU and MUSE certify forgetting only from a model's final response, so a model that simply declines to answer is treated as having forgotten the target knowledge. They show this model-level certification does not carry over to agentic deployments, where the model's behavior unfolds over multiple steps.

benchmarks/unlearning/agents/safety
papersTODAY 04:00 UTC

MUSE: Theory-Guided Story Engine for LLM Narrative Generation

A new arXiv paper in computational linguistics introduces MUSE, a story-generation engine that draws on narrative theory to steer how plot, character, and language choices fit together across planning, drafting, and revision. The authors frame story guidance as facing two bottlenecks, pointing to the quality of the guidance and how it is supplied. The abstract as posted is truncated, so full details of the method and results were not available.

story-generation/llm/narrative/arxiv
productsSEP 10 12:27 UTC

Meta unveils Muse, an AI agent in WhatsApp for shopping and bookings

Meta introduced Muse, an AI agent inside WhatsApp that can shop, book travel, draft emails and negotiate prices on a user's behalf. Payments are handled through Stripe's Link, positioning Meta ahead of OpenAI, which pulled direct checkout from ChatGPT. Meta also showed a separate security-focused agent named Sentinel.

agents/meta/whatsapp/payments
productsSEP 10 08:48 UTC

Meta announces Muse AI agent inside WhatsApp for shopping, bookings and email

Meta introduced Muse, an AI agent that operates through WhatsApp and can handle tasks such as booking travel, making purchases, writing emails and negotiating prices. The agent includes payment support through Stripe's Link service, going beyond OpenAI, which has discontinued its direct checkout feature in ChatGPT. A separate safety agent is also mentioned.

agents/meta/whatsapp/payments
productsSEP 9 23:57 UTC

Meta's AI agent Muse takes over social media handles of band Muse

Meta's newly announced AI agent shares its name with the British rock band Muse, and the band's social media accounts were reportedly reassigned to the tech company's product. Username transfers of this kind typically happen when platforms reclaim or hand over handles, leaving the group without accounts it had used for years. Neither party has publicly detailed how the switch occurred.

meta/agents/social-media/naming
productsSEP 9 16:43 UTC

Meta AI username tool reassigns Muse's social media handles

The rock band Muse reportedly lost control of social media handles tied to their name after Meta's new AI tool claimed or reassigned them. The tool appears to generate Instagram and Threads usernames automatically, including ones already associated with established accounts or brands. Meta has not detailed how it handles conflicts with existing handle owners.

meta/social-media/instagram/usernames