Liquid AI Releases LFM2-VL: Open-Weight Vision-Language Models for On-Device AI

Liquid AI has released LFM2-VL, a new family of open-weight vision-language foundation models created for low-latency, on-device deployment. The new models are designed to run efficiently on smartphones, laptops, wearables, and other embedded systems without needing to rely on cloud infrastructure. LFM2-VL is available in two versions: LFM2-VL-450M for highly resource-constrained devices and the more powerful LFM2-VL-1.6B. These models can process both text and images, offering up to twice the inference speed on a GPU compared to similar existing models. The models are available on Hugging Face under a license based on Apache 2.0, allowing for free academic and research use, as well as commercial use for smaller companies.

Sources:

Google Deploys Gemini-Powered AI Search in South Africa, Kenya, and Nigeria

Google has launched its ‘AI Mode’ in Search for users in South Africa, Kenya, and Nigeria, with the rollout beginning on August 21, 2025. This new feature is powered by a custom version of Google’s Gemini 2.5 model and is designed to handle complex, multi-part questions with more intuitive, AI-generated responses. Users can interact with AI Mode using text, voice, or images. The feature breaks down complex queries into subtopics to deliver richer insights and provides prominent links to web sources for deeper exploration. The AI Mode will appear as a dedicated tab on the Search results page and within the Google app for both Android and iOS.

Sources:

South Africa Unveils ‘Iris’: An AI Robot Tutor for All 11 Official Languages

South Africa has unveiled its first AI-powered robot tutor, named Iris, on August 19, 2025. Developed by innovator Thando Gumede and her company BSG Technologies, Iris is capable of teaching subjects from Grade R to the tertiary level in all 11 of the country’s official languages. The robot utilizes advanced natural language processing, emotional recognition, and adaptive learning algorithms to provide interactive and personalized lessons. The aim of Iris is to assist teachers by helping with tasks like marking and curriculum analysis, rather than replacing them. This innovation is seen as a significant step in addressing educational challenges of access and inclusivity in the country.

Sources:

Google Pixel 10 Series Debuts with On-Device Tensor G5 and Proactive AI

Google has unveiled its new Pixel 10 smartphone lineup, which includes the Pixel 10, 10 Pro, and 10 Pro Fold, with a significant focus on new artificial intelligence capabilities. A key new feature is ‘Magic Cue,’ an on-device AI assistant powered by the new Tensor G5 chip that proactively surfaces relevant information from apps like Gmail and Calendar based on the user’s context. For example, it can automatically display booking information when a user calls an airline. The new phones also introduce a ‘Camera Coach’ that provides real-time suggestions to improve photo composition and framing. Other AI-driven features include a new journaling app with AI suggestions and real-time call transcriptions.

Sources:

Nuro Raises $203M to Scale AI-First Autonomous Delivery Vehicles

Autonomous driving technology company Nuro has closed a $203 million Series E funding round, bringing its valuation to $6 billion. The investment round included participation from Uber, Baillie Gifford, Icehouse Ventures, Kindred Ventures, NVIDIA, and Pledge Ventures. This new capital is intended to scale Nuro’s AI-first autonomous technology and expand its commercial partnerships. The company has a technical collaboration with NVIDIA, utilizing its platforms for its compute module and for data processing and model training. Nuro has raised over $2.3 billion to date and is one of the few companies to have deployed autonomous vehicles at a city scale without a safety driver.

Sources:

Tesla is encountering growing legal and regulatory challenges concerning the marketing and performance of its Autopilot and Full Self-Driving (FSD) technologies. A federal judge has allowed a class-action lawsuit to proceed, alleging that the company has exaggerated its vehicles’ self-driving capabilities since 2016. This follows a lawsuit filed by the California Department of Motor Vehicles (DMV) in July, which claims that terms like ‘Autopilot’ and ‘Full Self-Driving’ mislead consumers into believing the vehicles are fully autonomous. Additionally, a Miami jury found Tesla partially liable in a fatal crash involving the Autopilot system, resulting in a significant financial award to the victim’s family. These legal actions could establish important precedents for advertising standards in the driver-assist technology sector.

Sources:

Microsoft and NFL Deepen Partnership with Generative AI for Football Operations

Microsoft and the National Football League (NFL) have announced an expanded multiyear partnership to integrate generative AI into team workflows. The deal involves deploying Microsoft’s Copilot AI and Azure cloud tools to assist in areas ranging from player scouting to in-game strategic decisions. As part of the initiative, over 2,500 new Surface “Copilot+” tablets equipped with AI-driven analytics will be distributed to all 32 NFL teams for players and coaching staff. A key feature includes a GitHub Copilot-powered tool that enables coaches to filter and analyze plays in real time, showcasing a significant application of AI in the sports industry.

Sources:

Huawei Commits to Tech Training in Malaysia, Focusing on AI and Cloud Skills

Huawei has announced a significant commitment to train thousands of individuals in Malaysia, a move aimed at strengthening the country’s growing technology ecosystem. This initiative will focus on developing skills in areas such as cloud computing, artificial intelligence, and 5G technology. By investing in local talent, Huawei aims to support Malaysia’s digital transformation goals and foster a new generation of tech professionals. The training programs are expected to be rolled out in collaboration with local universities and government agencies, providing participants with valuable industry-recognized certifications and hands-on experience.

Sources:

Microsoft Launches Quantum Safe Program to Transition to Post-Quantum Cryptography

Microsoft has launched its Quantum Safe Program (QSP), a strategic initiative to protect its infrastructure and customers from future cybersecurity threats posed by quantum computers. The company has set a goal to transition its products and services to post-quantum cryptography (PQC) by 2033, two years ahead of the 2035 deadline set by most major governments. The multi-year plan will be implemented in three phases, starting with the integration of PQC algorithms into foundational security components like SymCrypt. The program will then update core infrastructure services before a full rollout across all Microsoft services and endpoints, including Windows, Azure, and Microsoft 365. This initiative is in alignment with U.S. government guidelines and aims to address the “Harvest Now, Decrypt Later” threat, where encrypted data is stolen today to be decrypted by future quantum computers. Microsoft is also actively contributing to global PQC standards and open-source projects.

Sources:

Researchers Develop Magnetic Quantum Material for More Stable Qubits

Researchers from Chalmers University of Technology, Aalto University, and the University of Helsinki have developed a new exotic quantum material that uses magnetism to protect qubits from environmental disturbances. This breakthrough could lead to the development of more resilient and stable quantum computers. The new method utilizes magnetic interactions to create robust topological excitations, which are quantum states that are inherently more resistant to noise. This approach is significant because magnetism is a more common property in materials compared to the previously relied-upon spin-orbit coupling. The research team has also created a new computational tool to identify other materials with these desirable topological properties. This development is seen as a crucial step toward building practical, fault-tolerant quantum computers.

Sources:

IonQ’s Quantum Computing IP Portfolio Exceeds 1,000 Assets

IonQ, a company specializing in trapped-ion quantum computing, has announced that its intellectual property portfolio now exceeds 1,000 patents and patent applications. This milestone includes recently issued U.S. patents that strengthen its position in the field. The expanded IP estate covers various aspects of quantum computing and networking, including quantum circuit optimization, error mitigation techniques, and secure long-distance quantum networking. One of the new patents pertains to a portable quantum memory package designed for quantum network nodes, a key component for building a secure quantum internet. The company’s growing patent portfolio is a result of its long-term strategy of developing and owning quantum technologies across multiple applications.

Sources: