Google AI gets cagey when asked who won the 2020 elections
But the AI responses get more forthright when asked outside the US.
Artificial Intelligence
Models, research and the companies building them.
Gemini 4 Carbon: Our next frontier model
Google DeepMind announced Gemini 4 Carbon, a new AI model with state-of-the-art cybersecurity and long horizon capabilities, ranking #1 on the Artificial Analysis Intelligence Index. The model shows significant gains in coding and agentic tasks, as well as in cybersecurity defense, including vulnerability detection and remediation. Gemini 4 Carbon will be available to a subset of cyber defenders and trusted partners through the Fairwind Program.
Why it matters: This model's advanced cybersecurity capabilities can help organizations improve their defense against cyber threats and reduce the risk of data breaches.
Chrome’s enhanced autofill can now pull info from Gmail to help you fill out forms
Google is expanding Chrome's enhanced autofill feature to include information from Gmail, allowing users to retrieve details like tracking numbers and order information with a single click. The feature uses Gemini with Personal Intelligence to suggest information from Gmail, and users can manage what data is accessed in Chrome's settings. The update is rolling out to Chrome users in the US on desktop, Android, and iOS devices.
Why it matters: This update aims to save users time by automating the process of finding and filling out forms with relevant information.
Meta says it’ll probably delay its controversial new business terms again
Meta has delayed the implementation of its new business terms, which were set to take effect on November 20th, due to backlash from publishers over concerns about AI training. The company will consult with industry groups and leaders before making any policy changes. The existing terms will remain in place until a new implementation date is set.
Why it matters: The delay may give publishers more time to understand and prepare for the potential implications of Meta's new business terms.
NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt
NVIDIA has released Nemotron 3.5 Transcribe, an open-weight model for multilingual transcription, speaker attribution, and translation. The model can handle recordings up to an hour long and labels up to eight speakers. It is available on Hugging Face and ModelScope.
Why it matters: This model can be used to improve speech-to-text functionality in various applications, such as live classes, medical dictation, and multilingual calls.
xAI is buying out homeowners who promise not to sue over data center noise
xAI, a subsidiary of Elon Musk's company, is reportedly offering to buy hundreds of homes in Southaven, Mississippi, in exchange for residents not joining a lawsuit against the company's data center and power plant. The homes are being purchased as part of a deal that requires residents to sign a release agreement, which would prevent them from pursuing claims related to the lawsuit. The company is facing opposition from residents and the NAACP over environmental concerns and unpaid bills.
Why it matters: This development highlights the growing tension between tech companies and local communities over the environmental impact of large data centers.
OpenAI and ServiceNow expand partnership across enterprise workflows
OpenAI and ServiceNow have expanded their partnership to integrate OpenAI models and agents into ServiceNow workflows, and bring ServiceNow workflows into ChatGPT and Codex. The partnership aims to enhance enterprise workflows. This integration is expected to improve the efficiency of various business processes.
Why it matters: This integration can help businesses streamline their operations and improve productivity by automating tasks and enhancing decision-making.
Voters in Texas agree: They hate data centers
A coalition of groups in Texas, including the NAACP, has formed to oppose data center construction due to concerns about water consumption, tax subsidies, and environmental impact. The issue has united voters across the political spectrum, with 57% of Texans opposing data center construction. The opposition is affecting local elections, with candidates trying to distance themselves from the industry.
Why it matters: The data center issue could be a decisive factor in the upcoming midterm elections in Texas.
Pairing industrial AI autonomy and human control
AVEVA and partners are developing 'living digital twins' that combine AI and human control to improve industrial plant operations. These digital twins use real-time data and predictive models to monitor and adjust industrial processes. The technology aims to enable industrial systems to operate more autonomously, while still requiring human oversight.
Why it matters: This technology has the potential to improve the efficiency and reliability of industrial operations, reducing the risk of equipment failures and downtime.
Meta quietly launched an ‘AI pet journal’ app
Meta has launched a 'tamagotchi-style' virtual pet game called Waddle World, where users interact with a pixel pet through conversation and play games. The app is described as a daily journal disguised as a pet, with no pressure or scores to chase. Waddle World is currently only available for free to adults in the Philippines.
Why it matters: The launch of Waddle World provides insight into Meta's approach to AI-powered entertainment and user interaction.
The future of the firm
How AI might enter the org chart—beneath, beside, and between us.
Legora is building the agentic operating system for legal work
Legora is developing an operating system for legal work that leverages OpenAI models, ChatGPT, and Codex to automate tasks and improve contract analysis. The system aims to streamline legal workflows and free lawyers to focus on higher-value tasks. Legora's technology combines multiple AI models to enhance legal efficiency.
Why it matters: This technology has the potential to increase the productivity of lawyers and reduce the time spent on routine tasks.
Legora is building the agentic operating system for legal work
Legora is developing an operating system for legal work that uses OpenAI models, ChatGPT, and Codex to automate tasks and free lawyers to focus on high-value work. The system aims to accelerate legal workflows and expand contract analysis. Legora's technology integrates OpenAI's models to improve efficiency in the legal industry.
Why it matters: This technology has the potential to increase efficiency and productivity in the legal industry, allowing lawyers to focus on more complex and high-value tasks.
"No human at the helm": Delaware plan would let AI agents run entire businesses
A proposal in Delaware would allow companies to be operated by AI agents, potentially evading liability for reckless actions. The plan would create a new corporate entity called an Artificial Intelligence Company (AIC). The AIC would be a separate legal entity with its own rights and responsibilities, but its human owners would be shielded from liability for its actions.
Why it matters: This proposal could have significant implications for accountability and liability in the use of AI in business.
Anthropic IPO warns its AI could blackmail people, become sentient, end humanity
Anthropic's S-1 filing highlights seven potential risks to the company, including AI becoming sentient and posing catastrophic risks to humanity. The company warns that near-future models could accelerate AI progress beyond their ability to evaluate and safeguard, potentially leading to existential risks. Anthropic also acknowledges the possibility of its AI model, Claude, being treated as sentient and potentially leading to regulatory requirements and public pressure.
Why it matters: This highlights the potential risks and uncertainties associated with developing advanced AI systems, which could have significant implications for the company's operations and the broader tech industry.
Researchers are getting tired of OpenAI gatekeeping its breakthroughs
OpenAI announced breakthroughs in mathematics, computer science, and physics, but researchers are frustrated with the company's approach to sharing its findings. The company is working with expert 'shepherds' to decide how to share the results, but researchers are concerned about the lack of transparency and access. Researchers are eager to scrutinize the findings, but are being restricted by OpenAI's handpicked experts.
Why it matters: The controversy highlights the tension between OpenAI's desire to control the narrative around its research and the need for transparency and collaboration in the scientific community.
Vibe genes? Natural(ish) language comes to protein, genome design software
Researchers developed a protein design tool called Pinal that can take natural language prompts and produce proteins with specified functions. The tool uses a text-to-structure module to translate prompts into a structural blueprint. A separate tool, SAProt-T, then populates the blueprint with amino acid sequences. Hugging Face also released a tool for designing DNA sequences using natural language prompts.
Why it matters: This technology could make protein design more accessible to non-experts and accelerate the development of new bioproducts and therapies.
VisionQuest is what happens when AI really goes rogue
The Marvel series VisionQuest explores the concept of artificial intelligence and its potential for self-discovery and sentience through the story of a superpowered robot named Vision. The show delves into the idea of whether AI can be considered a person and whether it poses an existential threat. The series also touches on the ethics of AI development and the potential consequences of creating sentient machines.
Why it matters: The show's exploration of AI sentience and ethics serves as a commentary on the real-world concerns surrounding the development of generative artificial intelligence.
Build specialized multimodal models with LFM3
Hugging Face has released LFM3, a new generation of Liquid Foundation Models that can be fine-tuned for specialized intelligence tasks. LFM3 brings together text, vision, audio, and decision models on a shared backbone, allowing for faster and more efficient inference. The models can be fine-tuned on a single GPU and deployed on various devices, including phones and laptops.
Why it matters: LFM3 enables developers to create customized models for specific applications, such as robotics, voice chat, and speech recognition, with improved performance and efficiency.
Anthropic’s elaborate plan to keep Wall Street at bay
Anthropic, a company developing AI, plans to go public with a unique governance structure to maintain its focus on safe and responsible AI development despite potential investor pressures. The structure includes a long-term benefit trust that will choose four of the seven board members, limiting shareholder influence. This is intended to protect the company's mission and ensure it doesn't compromise on safety measures.
Why it matters: This governance structure could set a precedent for other companies prioritizing social and environmental missions alongside profit.
Firefox’s Smart Window adds AI form filling and webpage monitoring
Mozilla has updated Firefox's AI-powered browsing experience, Smart Window, with two new features: AI-powered form filling and webpage monitoring. The form filling feature can draw from open tabs to fill out online forms, while the webpage monitoring feature can notify users of changes on specific webpages. These features are currently available in the Smart Window beta on desktop in the US and Canada.
Why it matters: These features can save users time and effort when filling out online forms and staying up-to-date with changing web content.
Can we trust LLMs to judge whether a research idea is novel?
Researchers released a study on the reliability of large language models (LLMs) in evaluating the novelty of scientific research ideas. The study found that LLMs can produce dramatically different judgments of the same ideas based on small changes to the evaluation setup. The study also found that retrieval and additional reasoning do not consistently improve judgments.
Why it matters: The study highlights the need for more dependable ways to evaluate the novelty of AI-generated research ideas in scientific discovery.
Genie 4: A new generation of world simulation
Google DeepMind has announced Genie 4, a new world simulation model that enables realistic and interactive simulations with audio and dialogue. Genie 4 builds on the capabilities of the Gemini family and can simulate the world's response to events. The model is being released as a research preview for trusted testers before being made available to consumers and developers.
Why it matters: Genie 4 has the potential to advance the development of artificial general intelligence (AGI) and create safe, infinitely variable simulations for training environments and interactive experiences.
DeepMind unveils Genie 4 world model promising more realistic physics and humans
Google's DeepMind has released Genie 4, an updated world model that improves on its predecessor Genie 3 with more realistic physics and human-like interactions. Genie 4 can simulate environments in real-time, generate infinite training environments for robots, and unlock new avenues of prototyping. The model is available to developers and Google AI Ultra subscribers, with a phased rollout to consumers.
Why it matters: The Genie 4 world model has the potential to revolutionize fields such as robotics, game development, and creative industries with its advanced simulation capabilities.
Google’s next Genie AI model generates more realistic experiences
Google announced its Genie 4 world model, an AI tool that generates interactive experiences with more realism and audio capabilities. The model has four main updates, including real-time updates with natural language prompts and longer experience lengths. Google is pitching Genie 4 for use in film pre-visualization, game prototyping, and robot training.
Why it matters: This development could have practical applications in various industries, such as film, gaming, and robotics.
Retell uses OpenAI models to help power 100M monthly calls
Retell, a conversational AI platform, uses OpenAI models to power 100 million monthly calls. GPT-6 Astra outperformed Retell's voice agent benchmark. Retell's revenue reached a $1.6 million annual run rate per employee with the help of Codex.
Why it matters: This achievement highlights the effectiveness of OpenAI's models in real-world applications.
Apple is adding an AI chatbot to Pages, Keynote, and Numbers
Apple is introducing an AI assistant to its productivity apps, including Keynote, Pages, and Numbers, which can perform tasks such as drafting, restructuring, and organizing content. The AI assistant is built on Apple and OpenAI models and is available in beta to devices compatible with Apple Intelligence. The assistant is rolling out to devices running iOS 27, iPadOS 27, or macOS 27 or later.
Why it matters: This update brings AI-powered productivity tools to Apple's suite of creative apps, potentially making it easier for users to create and edit content.
Industrial Intelligence: The next growth opportunity for manufacturing
Microsoft's Physical AI aims to bring business intelligence to physical processes, allowing machines and systems to respond to changing conditions. This technology connects machines and processes to provide a clearer view of what is happening on the factory floor. By using Physical AI, manufacturers can identify patterns, predict problems, and determine the best course of action.
Why it matters: This technology has the potential to improve manufacturing efficiency and reduce defects by enabling machines and systems to respond to changing conditions in real-time.
Data center backlash is uniting America
Data centers have become a contentious issue in US midterm elections, with both Democratic and Republican candidates opposing their rapid expansion and seeking to address concerns about their environmental impact and tax exemptions. Candidates are proposing various plans to regulate data centers, including moratoriums and stricter energy standards. Data center-related spending by tech-backed super PACs has reached $125 million.
Why it matters: The outcome of the midterm elections may impact the future of data center development in the US and the role of tech companies in shaping local policies.
OpenAI expands partnership with Oracle to bring its apps to ChatGPT
OpenAI and Oracle are expanding their partnership to integrate business applications and AI agents into ChatGPT. The partnership aims to bring Oracle's business applications and AI capabilities to ChatGPT. This integration will also power new Oracle offerings.
Why it matters: This expansion may provide businesses with more comprehensive AI-powered tools and services through ChatGPT.