Perplexity AI boosted app installs by 50% through a Super Bowl contest offering a $1 million prize for users asking five questions via its app. This approach, unlike Google and OpenAI's ads, emphasized practical AI use cases, driving engagement and productivity ranking growth to sixth in the US App Store, though it faced mixed results compared to tech giants' broader ad strategies.

Google has expanded NotebookLM Plus, its AI-powered note-taking and research tool, to individual users via the $20/month Google One AI Premium plan. Originally launched for enterprise users, this upgraded version offers enhanced features like increased usage limits, support for up to 500 notebooks, and advanced audio generation capabilities. Students in the U.S. aged 18+ receive a discounted rate of $9.99/month. Future plans include mobile apps, multi-language support, and continued improvements

Baidu's Wenxin Yiyan App for iOS has been updated to version 4.9.0, introducing the DeepSeek-R1 model. This enhances the snap-to-solve feature by providing detailed reasoning steps for math and physics problems, improving learning efficiency. The update reflects Baidu's advancements in AI education and commitment to user-centric improvements, aiming to strengthen its role in the educational market.

OpenAI is set to produce self-developed AI chips next year, reducing reliance on NVIDIA. The design will be finalized soon and manufactured by TSMC using 3-nanometer technology. Initial deployment will be limited, focusing on running AI models. Led by Richard Ho, OpenAI's chip team has expanded to 40 members, highlighting growing R&D efforts. This follows a collaboration with Broadcom for custom chips. Tech giants continue investing heavily in AI infrastructure despite some质疑from startups like D

DeepSeek R1 is a cost-effective AI inference model that matches or surpasses top competitors, reducing costs by 96%. Leveraging chain-of-thought reasoning, reinforcement learning, and MoE architecture, it excels in math and programming tasks while minimizing computational expenses. Iterative development ensures high performance in resource-constrained environments, making advanced AI both powerful and accessible.

Google's DeepMind team developed AlphaGeometry2, an AI capable of solving high school-level geometry problems from the International Mathematical Olympiad (IMO) with gold medal-level performance. The system integrates Google's Gemini language model and Alpha Proof for mathematical reasoning, enabling it to solve 42 out of 50 IMO questions tested, surpassing average human gold medalist performance.

Harness Inc. and Traceable Inc., led by Jyoti Bansal, are merging to integrate API security into software development processes. The combined entity will focus on autonomous AI for faster, secure application delivery, aiming to unify DevOps and security. With 1,100 employees, the company is projected to generate $250 million in annualized revenue for 2025, reflecting a 50% increase over last year.

Elon Musk and a group of investors submitted a $97.4 billion acquisition offer for OpenAI, aiming to return it to its original non-profit mission. Led by Musk's lawyer, the proposal was presented to OpenAI's board, opposing CEO Sam Altman's plan to convert it into a for-profit entity. Backed by Musk's AI firm xAI and major investors, this move adds uncertainty to OpenAI's future as industry players watch the unfolding negotiations. Altman humorously declined, referencing Musk's Twitter purchase.

Chinese researchers from Westlake University and others developed biodegradable robot parts using cotton cellulose and pigskin gelatin. Published in Science Advances, the study details a 240mm robotic arm with sensory capabilities, controlled via ion-conductive gelatin wires. The biodegradable components can serve as fertilizer post-use or degrade naturally in medical applications.

Ant Group's Alipay Baobaoxiang service now offers DeepSeek R1 and V3 models to all users, highlighting R1's visualized reasoning process. The DeepSeek series includes 32B and 7B versions, available for free. Users can integrate these models into conversational apps or workflow nodes via Baobaoxiang's basic or professional version, enhancing intelligence and efficiency in various scenarios.

The VideoWorld model, co-developed by the Dabao team, Beijing Jiaotong University, and USTC, is now open-source. It breaks industry norms by understanding the world without language models. Built on a latent dynamic model, it efficiently compresses video frame changes for improved knowledge learning. Achieving professional-level performance in tasks like 9x9 Go without reinforcement learning, VideoWorld opens new research avenues in video generation and cognition.

French AI startup Mistral plans to invest billions in its first French data center, aiming to control core data storage and processing capabilities. CEO Arthur Mensch emphasized controlling the AI value chain, aligning with France's initiative to become a data hub. Recently, Mistral launched AI assistant "Le Chat," praised by President Macron, and received government support via new contracts. Cerebras aids in computing power, while Mistral prepares for an IPO and expansion into Asia-Pacific and

Apple explores expressive movements in consumer-grade robots, emphasizing human-robot interaction through intention, attention, and emotion. Using a desk lamp as an example, similar to Pixar's "Luxo Jr.", the study shows how simple motions can trigger anthropomorphic perception. Accompanying research videos demonstrate potential applications, such as integrating Siri for voice responses or simulating head-turning actions. This aligns with rumors of Apple developing advanced smart home robots, po

11 tech leaders support Current AI's mission to provide AI tools and infrastructure for public good. Funded by $400 million from governments and organizations, Current AI aims to raise $2.5 billion for open data access, user-friendly tools, and measuring AI's social/environmental impact, aligning with sustainability goals ahead of the Global AI Summit in Paris.

Apple may launch a new iPhone SE with USB-C, Face ID, and an Apple-designed modem on February 11, alongside PowerBeats Pro 2 headphones. The event is low-key with no formal invitations. The iPhone SE targets markets like China and India, while PowerBeats Pro 2 could feature heart rate monitoring. Future releases may include M4 updates and a smart home hub with a display.

Sophisticated scammers impersonated Italy's Defense Minister using AI voice tech to target billionaires, tricking one into losing €1M. Milan prosecutors are investigating amid rising concerns over AI-enabled fraud. Meanwhile, global leaders discuss AI regulation at a Paris summit as defense partnerships strengthen between Italy and India.

European leaders face pressure to loosen AI regulations to compete with the US and China. A Paris summit hosted by Macron will explore global AI cooperation, emphasizing Europe's potential in talent and computing power. Tech executives and world leaders will attend as concerns grow about stringent EU regulations hindering innovation. Calls for sector-focused incentives and reduced burdens aim to position Europe as a leader in trustworthy AI, leveraging healthcare and energy sectors while develop

AI costs are plummeting rapidly, dropping to one-tenth annually and outpacing Moore's Law, according to Altman. He highlights three key trends: decreasing costs, intelligence scaling with resources, and super-exponential AI growth. These could drastically lower prices for many goods while increasing scarcity-driven luxury costs. Despite a January sell-off in AI stocks, tech giants are boosting AI investments, with combined capital expenditures exceeding $320 billion this year. Altman also discus

Safe Superintelligence Inc. (SSI), co-founded by Ilya Sutskever, is raising new funds at a $20 billion valuation, up from $5 billion last September. Backed by investors like Sequoia Capital and Andreessen Horowitz, SSI focuses on developing safeguarded superintelligent AI models with a novel research approach. Sutskever highlights the limitations of current LLM scaling methods due to data constraints and emphasizes reasoning challenges in AI systems. Meanwhile, Liquid AI Inc. is advancing liquid

Sam Altman outlines OpenAI's vision for AGI to empower individuals globally, emphasizing creativity and accessibility. He discusses the need for significant investment, potential societal impacts, and balancing safety with openness. Altman acknowledges challenges like mass unemployment and authoritarian misuse, advocating for human oversight and transparent practices while pursuing AGI development.