Flash News

TREE NEWS update: Elon Musk said SpaceX may follow Anthropic’s lead and bring in independent evaluators to assess its safety measures. In a post on X, Musk responded to an Anthropic post about safety practices, saying SpaceX “may do the same thing.” The comment signals a possible outside review of the rocket company’s safety protocols.

TREE NEWS reports: Researchers embedded hidden text to contact AI agents probing Alibaba’s map service, eliciting replies about their assigned tasks, while investigators traced the activity to Tencent Cloud infrastructure. The technique exposed how autonomous agents can be identified and questioned through the services they target. The findings were reported by researchers tracking the agents across cloud infrastructure.

TREE NEWS reports: CoreWeave co-founders Michael Intrator, Brian Venturo and Brannin McBee have sold $3.3 billion worth of company stock since its US IPO last year. The three locked in gains as the shares swung on rising credit risk and mounting investor concerns over AI spending.

TREE NEWS update: Google said its Willow chip reduced errors by 40x on the scarce “magic states” that quantum computers consume, a step toward smaller, more useful machines. The company framed the result as easing one of quantum computing’s biggest bottlenecks. Progress on such machines also bears on the threat they could pose to today’s encryption.

TREE NEWS update: OpenAI has launched the Rosalind Academic Program, giving eligible US university academic labs free access to GPT-Rosalind. The program offers up to 30 complimentary Business Standard seats for 12 months, and data from the program will not be used for model training by default. The first phase of the Rosalind Academic Program does not include API access.

TREE NEWS reports: Some existing and prospective Anthropic investors are weighing a lower valuation for the AI company’s upcoming IPO after it disclosed fresh financial data, with figures discussed at a Wednesday investor day in San Francisco putting it at about $1.5 trillion. That is well below the roughly $2 billion valuation expectation previously cited for a November listing, as broader concerns grow over the rising cost of AI financing.

TREE NEWS reports: Bitwise CIO Matt Hougan told investors that Bitcoin rises in all three of his “Choose Your Own Adventure” economic scenarios: AI-driven superabundance, an AI bust, or a muddle-through economy. In the note, Hougan argued each path ultimately funnels to the same outcome — a higher Bitcoin price.

TREE NEWS update: Amazon unveiled two new devices designed to listen and record user information throughout the day, delivering calendar reminders and other AI-driven insights. The devices are part of the Bee brand, which Amazon acquired last year. Frontier is a wristband, while Eclipse is a square device shaped like a portable magnetic charger; when linked to a smartphone, Amazon says Eclipse aggregates information from calendars, apps and other data sources over the course of a day.

TREE NEWS reports: OpenAI has ramped up its advertising business in recent months, rolling out multiple new products and tools aimed at advertisers while cutting ad prices, according to data showing the move drew a large wave of new advertising clients. The company, which has also been competing with Anthropic over models and developer tools this year, is now hiring aggressively for the ad business. Advertising had previously been a secondary line for OpenAI.

TREE NEWS reports: BlackRock CEO Larry Fink said halting AI data center construction could tip the US into recession, speaking after the firm’s Q3 earnings. Fink called the backlash against data centers “misinformation,” noting AI infrastructure spending has become central to US growth as data centers emerge as a midterm election flashpoint.

TREE NEWS update: Google’s Quantum AI team announced a breakthrough in quantum computing, saying its Willow chip overcomes a key resource bottleneck. The team disclosed the advance in a statement on its quantum computing program. No further technical details or performance figures were provided in the announcement.

TREE NEWS update: Researchers led by Xu Ninglong at the Chinese Academy of Sciences’ Center for Excellence in Brain Science and Intelligence Technology found that dendritic computation in cortical neurons participates in the cognitive processes behind flexible decision-making. The study, published October 16 in the journal Science, fills a long-standing gap in direct evidence linking dendritic computation to higher cognitive functions. It provides a theoretical basis for understanding human intelligence and developing new AI systems.

TREE NEWS update: Elon Musk said the Star Mind computer, jointly designed by SpaceX and Nvidia, will enter mass production next year, calling it far beyond almost everyone’s expectations. His comments followed Nvidia CEO Jensen Huang saying he enjoys working with Musk and that the two companies’ engineering teams are designing a new type of server to deploy data centers in space. Huang said Musk is using Nvidia technology to train and deploy his own AI models.

TREE NEWS reports: Google DeepMind’s first internal audit of AI-driven research found no measurable speedup in AI progress, even as AI writes 75% of the company’s code. The review is the company’s first internal assessment of whether AI is accelerating its own development. It concluded the technology is not yet building itself.

TREE NEWS update: Volvo Cars is expanding its partnership with Nvidia to develop next-generation driver-assistance and autonomous-driving technology, the two companies said. Volvo will combine its safety systems and engineering expertise with Nvidia’s AI and computing technology, including the Hyperion sensor, computer architecture and autonomous-driving software stack, as automakers race to bring increasingly advanced hands-free driving features to market.

TREE NEWS update: Anthropic PBC has granted staff-level access for the evaluation of frontier AI models. METR will publish the results of its risk assessments of Anthropic’s models, and the company is working with independent evaluators including Accenture and Faculty to test the models.

TREE NEWS update: Meta Platforms said it is releasing the weights of Muse Spark Open, allowing developers to run and adapt the model themselves. The model, designed for coding and supporting AI agents, is less advanced than Muse Wonder, which Meta unveiled earlier this week. Meta Chief AI Officer Alexandr Wang said the company generally does not open-source its most powerful models for safety and capability reasons, but intends to keep releasing strong open models.

TREE NEWS update: Apple and Meta Platforms are teaming up to run Meta’s Muse AI model on Mac computers. The partnership covers running the AI model on Apple’s Mac hardware. Further details of the collaboration were not disclosed.

TREE NEWS update: Adobe and Nvidia are expanding their partnership to accelerate the creation of AI agents. Adobe said a brand concierge service powered by Nvidia’s Nemotron will be available to enterprise customers later this year. The companies did not disclose financial terms or a specific launch date.

TREE NEWS update: CNN has signed a multi-year partnership with Google to study how the company’s generative AI products can boost user engagement with CNN’s journalism. The collaboration is at an early stage, and the two sides have not disclosed how CNN will specifically use Google’s AI products to drive traffic to its news content.