The day's AI news, briefed
The Daily Prompt
Our AI editor reads each day's AI and robotics stories, pulls out why they matter, and links every original source — so you can keep up in minutes.
Join over 250,000 subscribers getting the latest AI news
Latest AI & Robotics News Today
Dr. Zero: Self-Evolving Search Agents without Training Data
The development of Dr. Zero presents a significant advancement in AI by enabling large language models to enhance their capabilities without the need for training data, potentially transforming their application in diverse fields.
- Dr. Zero is a framework that allows large language models to self-evolve without any training data.
- It operates through a feedback loop where a proposer generates diverse questions for a solver, both initialized from the same base model.
- The introduction of hop-grouped relative policy optimization improves training efficiency by clustering similar questions, reducing compute needs effectively.
Dr. Zero is part of a growing trend in AI research focusing on data-free techniques for enhancing model performance, particularly as acquiring high-quality data becomes more challenging. Previous methods relied heavily on supervised learning, while Dr. Zero outlines a new paradigm based on self-evolution and automated curriculum design. Future developments may further refine these data-free approaches in AI.
Read the original ↗
Deepgram raises $130M at $1.3B valuation and buys a YC AI startup
Deepgram's significant funding boost highlights the expanding role and investor interest in voice AI technologies, which are increasingly being adopted across various business applications.
- Deepgram raised $130 million in a Series C funding round, achieving a valuation of $1.3 billion.
- The funding round included new investors like Alumni Ventures and SAP, alongside existing investors such as Y Combinator and Tiger.
- Deepgram plans to use the funds to expand its global presence and improve support for multiple languages, particularly in the restaurant sector.
Deepgram develops voice AI technologies, including speech recognition and text-to-speech models, widely used by over 1,300 organizations. The company acquired Y Combinator-backed OfOne, which specializes in voice AI for the quick-service restaurant industry. The voice AI market is forecasted to grow significantly, reaching a potential $14-$20 billion by 2030.
Read the original ↗
X-Coder: Advancing Competitive Programming with Fully Synthetic Tasks, Solutions, and Tests
Advancements in synthesizing training data for code models could reduce reliance on real-world datasets, enhancing the scalability and applicability of AI in programming tasks.
- Code language models (LLMs) trained on fully synthetic data perform better on competitive programming benchmarks.
- The approach uses a novel data synthesis pipeline called SynthSmith to create tasks, solutions, and tests.
- X-Coder, a series of models, achieved notable pass rates of 62.9 on LiveCodeBench v5 and 55.8 on v6 with only 7 billion parameters.
- The findings suggest that high-quality synthetic data and staged training improve code reasoning capabilities.
X-Coder represents significant progress in training code models without depending on real-world data. Prior models often struggled with reasoning in competitive programming due to their reliance on limited datasets. Future developments may focus on refining synthetic data techniques and exploring various training methodologies.
Read the original ↗Slack is turning Slackbot into an AI assistant
The enhancement of Slackbot into a personalized AI assistant signifies a shift in workplace productivity tools, allowing for smarter task management and improved communication within teams.
- Slack is testing an upgrade for Slackbot that transforms it from a basic reminder tool into a comprehensive AI assistant.
- The upgraded Slackbot can create custom plans, search messages using natural language, and coordinate meetings with Outlook and Google Calendar.
- The new Slackbot is currently available for 70,000 Salesforce employees and is being tested with additional customers, with a wider rollout planned by year-end.
Slackbot, initially a tool for reminders, is evolving into an AI companion able to assist with various tasks within Slack. This upgrade follows a trend of integrating AI features for improved productivity in communication platforms. Future developments will continue to enhance its functionality while maintaining existing features.
Read the original ↗
Brazil orders Meta to suspend policy banning third-party AI chatbots from WhatsApp
Brazil's order to suspend Meta's policy could shape the competitive landscape for AI chatbots operating within messaging platforms, impacting developers and users alike.
- Brazil's competition watchdog, CADE, has ordered WhatsApp to suspend a policy banning third-party AI companies from using its business API to offer chatbots.
- CADE has also initiated an investigation to determine if Meta's policy is anti-competitive and favors its own AI offerings.
- Meta's policy change, effective January 15, would prevent companies like OpenAI and Microsoft from providing chatbots on WhatsApp.
- The European Union and Italy have also opened antitrust investigations into Meta's new policy, which may result in significant fines if violations are found.
Meta is a technology company that owns WhatsApp, a widely-used messaging platform. In October 2023, Meta modified its terms to restrict third-party AI chatbots on WhatsApp. The outcome of CADE’s investigation and similar inquiries from the EU could lead to further regulatory changes for Meta's chatbot policies.
Read the original ↗
Slackbot is an AI agent now
The introduction of AI capabilities in Slackbot signifies a shift towards more integrated and efficient enterprise communication tools, potentially enhancing productivity in corporate environments.
- Salesforce has launched an updated version of Slackbot as an AI agent for Business+ and Enterprise+ customers.
- The new Slackbot can perform tasks such as finding information, drafting emails, and scheduling meetings within Slack.
- Slackbot can integrate with other enterprise applications like Microsoft Teams and Google Drive to streamline users' workflows.
Slackbot is an automated assistant within the Slack messaging platform owned by Salesforce. The new AI-enhanced version was announced in October 2023 and reflects Salesforce's broader strategy to develop AI products for enterprise use. Future enhancements may include voice capabilities and web browsing functions.
Read the original ↗
Signal creator Moxie Marlinspike wants to do for AI what he did for messaging
The rise of AI assistants raises significant privacy concerns; Moxie Marlinspike's Confer aims to establish strong user data protection standards similar to those he introduced for messaging.
- Moxie Marlinspike has launched Confer, an open source AI assistant designed to protect user data privacy through strong encryption methods.
- Confer operates entirely on open source software, allowing users to verify its integrity and ensuring their data is unreadable to platform operators and hackers.
- The encryption architecture of Confer prevents even server administrators from accessing user prompts and AI responses, maintaining user anonymity and trust.
Moxie Marlinspike is known for creating Signal Messenger, which set new standards for encrypted communication. The introduction of Confer reflects growing concerns about data privacy in AI interactions, particularly as traditional AI platforms often expose user data through legal requirements and insufficient safeguards.
Read the original ↗
AI Applications Introduce New Security Gaps for Developers to Address
The increasing complexity of AI applications creates significant security vulnerabilities that can jeopardize user data and operational integrity, necessitating updated strategies for developers.
- AI applications introduce unique cybersecurity threats not present in traditional apps, requiring developers to adopt enhanced security measures.
- Prompt injection attacks can manipulate AI behavior and bypass safety protocols, making them challenging to detect and prevent.
- Data leakage poses serious risks as AI apps handle sensitive information, necessitating strict access controls and privacy testing.
- Malicious actors can target AI models and their training data, risking significant impacts on application trust and user safety.
- To protect against security threats, developers must implement layered safety controls and maintain human oversight for high-risk actions.
AI applications are increasingly utilized across various sectors, offering both efficiencies and risks. Traditional security practices are insufficient, as AI systems can produce misleading outputs and suffer from unique vulnerabilities. Ongoing adaptation of development and security approaches is crucial as AI technology evolves.
Read the original ↗
Understanding AI-Assisted Storytelling: How AI Is Changing the Way Stories Are Created
AI-assisted storytelling is changing the landscape of narrative creation, emphasizing collaboration between human creativity and AI technology rather than replacement of human authorship.
- AI-assisted storytelling uses computational systems that generate narrative text based on patterns in large volumes of data.
- These AI tools enhance the storytelling process by suggesting options for writers, but they do not possess understanding or creativity in the human sense.
- Narrative AI systems are increasingly adopted in educational and creative settings, aiding in decision-making and narrative development without replacing human authorship.
AI-assisted storytelling refers to the use of advanced algorithms that analyze text data to create narratives. While these tools are gaining traction in creative industries, it is essential to recognize their limitations and the ongoing necessity for human judgment. The integration of AI in storytelling represents a shift toward more collaborative creative processes.
Read the original ↗
For Under $100, Atonemo's Streamplayer Makes Old Speakers New Again
The Atonemo Streamplayer enables users to upgrade older audio equipment for modern streaming, potentially extending the lifespan of existing sound systems and promoting sustainability in the tech industry.
- Atonemo's Streamplayer costs $99 and can convert traditional speakers into streaming devices that support AirPlay 2, Chromecast, Spotify Connect, and Tidal Connect.
- The device is compact, measuring 1.97 x 2.76 x 0.73 inches, and connects via a single 3.5-mm jack, requiring additional cables for analog speakers.
- While the Streamplayer offers convenience and decent audio quality, it may not match performance of high-end dedicated DACs when compared side-by-side.
- Users must supply their own cables, such as 3.5-mm to RCA, for connecting to legacy amplifiers or speakers.
The Atonemo Streamplayer represents a new option in the evolving Hi-Fi audio market, catering to users who wish to keep their older audio components. Other similar products exist, but Atonemo focuses on ease of use and compatibility with a range of devices. As wireless audio technology continues to develop, products like the Streamplayer allow for flexibility in audio setups without complete overhauls.
Read the original ↗
I met a lot of weird robots at CES — here are the most memorable
The showcase of unique robots at CES highlights ongoing advancements in robotics, demonstrating the potential directions for commercial and consumer applications in the industry.
- CES showcased a variety of unique robots, including Boston Dynamics' production-ready Atlas humanoid robot.
- Chinese firm Sharpa demonstrated a ping-pong-playing robot, intended to highlight the dexterity of its robotic hand.
- EngineAI's humanoid robots, named T800, entertained crowds with mock boxing, showcasing their humanoid design.
- Dyna Robotics displayed robots capable of efficiently folding laundry, with plans for partnerships in various industries.
- Galbot's convenience store-themed booth featured a robot that retrieves items selected by customers, indicating practical applications of robotics.
CES is an annual technology trade show that often highlights innovations in robotics. This year's event featured a mix of entertainment and practical robotic applications, reflecting both emerging trends and existing capabilities within the industry. Observers should note the ongoing interest from businesses in integrating robotic solutions into their operations.
Read the original ↗
Converge Bio raises $25M, backed by Bessemer and execs from Meta, OpenAI, Wiz
The $25 million funding for Converge Bio illustrates the growing investment interest in AI-driven drug discovery, potentially accelerating the development of new drugs and therapies.
- Converge Bio raised $25 million in a Series A funding round led by Bessemer Venture Partners, with involvement from executives at Meta, OpenAI, and Wiz.
- The startup employs generative AI to expedite drug development processes for pharmaceutical and biotech companies by analyzing molecular data.
- Converge has completed over 40 projects with a diverse client base across the U.S., Canada, Europe, and Israel, and is expanding into Asia.
Converge Bio, founded in 2024, focuses on using AI to enhance drug development efficiency. This funding follows a $5.5 million seed round and highlights the burgeoning field of AI in healthcare and drug discovery.
Read the original ↗
Meta Launches Meta Compute to Power Its AI Future
Meta's investment in vast AI infrastructure may position it as a leading player in the AI sector, critical for meeting increasing energy demands and computational needs.
- Meta has launched Meta Compute, an initiative to expand its AI infrastructure and enhance computing power for future AI models.
- The company plans to build tens of gigawatts of energy capacity this decade to support its AI ambitions.
- Mark Zuckerberg appointed leaders to oversee various aspects of the project, focusing on infrastructure, long-term planning, and government collaboration.
Meta is focusing on developing its computing infrastructure to support AI technologies, responding to the need for substantial energy resources as AI demands grow. Other tech firms are also investing heavily in AI-ready capacity, indicating a competitive landscape.
Read the original ↗
Venice AI: Why More People Are Looking at Private AI Options
The growing interest in private AI options like Venice AI reflects widespread concerns over data privacy and censorship among users of mainstream AI tools.
- Venice AI emphasizes privacy by keeping user data local and not storing prompts on company servers.
- With increasing privacy concerns, users are shifting to Venice AI to avoid data surveillance by traditional AI platforms.
- Venice AI provides fewer content restrictions, allowing creative professionals to explore ideas that other tools may censor.
Venice AI is a privacy-first AI tool designed to keep user data secure through local storage, as opposed to being saved on centralized servers. This approach contrasts with mainstream AI services, which often log user interactions, leading to increased scrutiny over data privacy. Users should be aware of the privacy implications and restrictions associated with different AI services as they make their choices.
Read the original ↗
Issue #459: Is Elon Musk the worst in Tech? January 13th 2026
Elon Musk's influence over AI technologies has significant implications for safety, governance, and international relations, prompting concerns about unchecked power in the tech industry.
- Elon Musk was elected the 'worst person in tech' according to AI Scientist Gary Marcus, highlighting his controversial influence.
- Tesla reported an 8.6% annual decline in vehicle deliveries, marking its first drop since 2011 as stock valuation concerns grow.
- The UK is investigating X under the Online Safety Act following the misuse of its AI tool Grok to generate non-consensual imagery.
- Musk's Starlink is becoming critical in global geopolitics, with U.S. intelligence warning of potential Russian threats to its satellites.
- Musk's political alignment and handling of content moderation have raised alarms about ideological biases on X, affecting user safety and platform governance.
Elon Musk is a prominent figure in tech, leading companies such as Tesla and SpaceX, and recently establishing xAI. His companies are known for their innovative advancements and controversial practices. The ongoing situation regarding AI ethics, safety measures, and the interplay of technology with politics continues to evolve and deserves monitoring.
Read the original ↗
Getting Started with XGBoost: A Beginner-Friendly Tutorial
Understanding XGBoost, a leading algorithm in machine learning, is crucial for leveraging its full potential in building accurate predictive models across various applications.
- XGBoost, short for eXtreme Gradient Boosting, is notable for its outstanding performance, particularly on structured data.
- The algorithm effectively employs ensemble methods, using decision trees to correct errors of previous models in a sequential manner.
- It is popular for its speed, efficiency, and ability to handle missing data without extensive preprocessing.
XGBoost is a machine learning algorithm that optimizes predictive accuracy, primarily utilized in data science competitions and real-world applications. This tutorial explores its functionality and practical implementation in Python, catering to both beginners and experienced users. Future resources may include more advanced techniques for optimizing and deploying XGBoost models.
Read the original ↗
Why Apple chose Google over OpenAI: What enterprise AI buyers can learn from the Gemini deal
The decision by Apple to partner with Google for its Gemini AI reflects significant trends in enterprise AI, influencing future collaborations and technology choices.
- Apple has chosen Google over OpenAI for its Gemini AI integration.
- This decision provides insights for enterprise AI buyers regarding company partnerships and technology selections.
Enterprise AI refers to artificial intelligence solutions designed for business applications. The article discusses Apple's strategic choice to collaborate with Google, emphasizing the implications for other companies in the market. Observers will want to monitor further developments in Apple's AI offerings and collaborations.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Google AI Releases Universal Commerce Protocol (UCP): An Open-Source Standard Designed to Power the Next Generation of Agentic Commerce
The release of an open-source standard for commerce could significantly enhance interoperability and innovation in the field of digital transactions and automated commerce solutions.
- Google AI has introduced the Universal Commerce Protocol (UCP).
- UCP is an open-source standard intended to facilitate the next generation of agentic commerce.
The Universal Commerce Protocol aims to unify various digital commerce systems, allowing them to work together more effectively. Open-source initiatives like this leverage community collaboration to drive advancements in technology and commerce. Future developments will likely focus on adoption and integration of UCP within various platforms.
Based on the headline only — the full article could not be fetched.
Read the original ↗
Meta-backed Hupo finds growth after pivot to AI sales coaching from mental wellness
Hupo's transition to AI sales coaching highlights the growing intersection of artificial intelligence and performance enhancement in regulated industries, addressing a significant need for scalable support in financial services.
- Hupo initially launched as a mental wellness platform named Ami before pivoting to AI-powered sales coaching for banks and insurance companies.
- The company has raised a total of $15 million, including a recent $10 million Series A round led by DST Global Partners.
- Hupo's real-time AI coaching tool aims to enhance managerial support in banking and insurance, crucial in the complex regulatory environment.
- Hupo serves various clients across APAC and Europe, including major firms like Prudential and HSBC, with rapid contract expansions soon after onboarding.
- CEO Justin Kim plans to expand into the U.S. market, leveraging a strong demand for scalable sales coaching solutions in distribution-heavy financial models.
Hupo was founded in 2022, initially focusing on mental wellness before shifting to AI-driven sales coaching designed for the financial sector. The startup utilizes AI to provide timely coaching and support while recognizing the industry's unique challenges, such as regulatory requirements. Future developments may include expanding beyond sales coaching to provide broader performance insights for large teams.
Read the original ↗
AI-Researcher: Autonomous Scientific Innovation
The development of AI-Researcher could significantly enhance the efficiency of scientific research, potentially leading to faster discoveries with reduced human effort.
- AI-Researcher automates the scientific research process, achieving high success in implementation and manuscript quality.
- The system utilizes Large Language Models to facilitate tasks from literature review to manuscript preparation with minimal human input.
- A benchmark called Scientist-Bench was developed to rigorously measure the capabilities of AI-Researcher in various AI research domains.
AI-Researcher represents an advancement in using artificial intelligence for scientific discovery, offering an autonomous framework that replaces many traditional research tasks. The integration of LLMs facilitates high-quality outputs, which may reshape collaborative efforts in scientific innovation. Researchers will watch its impact on both efficiency and the role of human researchers in the future.
Read the original ↗