This week's latest AI industry updates: February 25, 2025

Stay informed with SymphonyAI's weekly Eureka newsletter, covering all the important AI industry developments and technology advancements you need to know.

SymphonyAI news?

• Six ways agentic AI for retail transforms business in 2025: Agentic AI is revolutionizing retail operations by enabling autonomous agents to perform tasks independently across areas like merchandising and category management, improving efficiency and decision-making.?

• Turbocharge your enterprise workforce with AI copilots and automation: SymphonyAI and Microsoft present a webinar showcasing how generative AI and automation can transform IT service management and improve employee productivity.?

• Tech companies reassess IPO timing in 2025 market: Amid market uncertainties, many tech companies are reconsidering IPO plans, though some like SymphonyAI, with $400 million in revenue, remain optimistic about going public.?

• Unlocking the potential of AI in the fight against financial crime: Singapore is establishing itself as a global leader in AI adoption within its banking sector through strong regulatory frameworks and infrastructure while facing challenges such as talent shortages and ethical concerns.?

• When will large vision models have their ChatGPT moment?: Large vision models (LVMs) are gaining traction in computer vision applications, following the success of transformer architecture in language models, but still await widespread adoption due to hardware requirements and the need for fine-tuning. ?

• ITSM evolution—AI-powered transformation: AI is transforming the IT service management industry by revolutionizing customer support, enabling predictive analytics, and improving operational capabilities.?

• Optimizing industrial operations with DataOps and AI: A webinar featuring experts from SymphonyAI and LTTS demonstrates how AI-driven industrial data operations can optimize manufacturing processes and improve operational efficiency.?

?

Upcoming webinars??

Learn more about how AI is transforming your industry in one of our upcoming webinars.??

• Retail: Improve new product success with agentic AI: February 27?

• Industrial—Combining AI and simulation models for performance optimization: February 27?

?

Meet SymphonyAI?

Join us at an upcoming industry event to learn more about how AI can drive outcomes in your business.??

• Financial Services—1LOD: February 27, New York?

• Retail—Shoptalk 2025: March 25 – 27, Las Vegas?

• Industrial—Hannover Messe: March 31 – April 4, Hanover

Big tech announcements?

• OpenAI’s ChatGPT explodes to 400M weekly users: OpenAI's ChatGPT reaches 400 million weekly active users and 2 million enterprise users while preparing to launch GPT-4.5 and GPT-5 amid growing competition from xAI and DeepSeek. ?

• OpenAI rolls out its AI agent, Operator, in several countries: OpenAI expands its AI task automation tool Operator to ChatGPT Pro subscribers in multiple countries, excluding the EU and select European nations.?Operator launched in the US in January.?

• OpenAI reports on efforts to combat AI misuse: OpenAI shares its efforts to protect democratic AI systems and prevent misuse of AI technology by authoritarian regimes and malicious actors.?

• OpenAI plans to shift compute needs from Microsoft to SoftBank: OpenAI plans to shift 75% of its computing power needs from Microsoft to SoftBank-backed Stargate project by 2030, amid projections of dramatically increasing operational costs.?

• Google pulls Gemini from main search app on iOS: Google removes Gemini from its main iOS app, requiring users to download a separate Gemini app to access the AI assistant. ?

• New Google AI leak reveals powerful Gemini video upgrade: A leaked report suggests Google is testing new AI video generation capabilities for its Gemini app, potentially integrating its Veo 2 technology to allow users to create realistic videos from text prompts.?

• Microsoft launches new platform to connect AI research with developers: Microsoft launches Azure AI Foundry Labs, a new platform that connects AI research innovations with developers to accelerate practical applications and development.?

• Microsoft unveils new quantum chip with scalable architecture: Microsoft introduces Majorana 1, a breakthrough quantum chip using topological architecture that promises to scale quantum computing to solve complex industrial problems.?

• Microsoft launches Muse AI model for 3D environment understanding: Microsoft introduces Muse, a new AI model that understands 3D environments and human interactions, initially applied to gaming but with potential applications across various industries.?

• Anthropic’s Claude 3.7 Sonnet hybrid reasoning model is now available in Amazon Bedrock: Amazon Bedrock adds Anthropic's Claude 3.7 Sonnet foundation model, featuring hybrid reasoning capabilities and enhanced coding performance.?

• Meta AI arrives in the Middle East and Africa with support for Arabic: Meta expands its AI chatbot service to the Middle East and North Africa, making it available in 42 countries with Arabic language support. ?

• Apple commits $500B to US manufacturing, including a new AI server facility: Apple announces $500 billion investment in US operations over next four years, focusing on manufacturing, engineering, and education, including a new server factory in Houston.?

• Apple’s $599 iPhone 16e adds AI, launches February 28: Apple announces iPhone 16e, a rebranded iPhone SE with AI capabilities, A18 chip, and Face ID, starting at $599 and shipping February 28.?

• The Vision Pro is getting Apple Intelligence in April: Apple announces Apple Intelligence integration for Vision Pro through visionOS 2.4, bringing AI features like text generation and image creation to the headset in April.?

AI impact?

• Research shows AI models still fall short of replacing human software engineers: Despite OpenAI's claims, new research shows that while large language models can help solve coding problems, they currently lack the comprehensive understanding needed to fully replace human software engineers.?

• Google's 'AI co-scientist' cracks decade-long research problem in two days:?An?AI co-scientist built on Gemini 2.0?analyzed?antibiotic-resistant bacteria and reached?the same hypothesis that had taken Imperial College London scientists a decade to develop, accomplishing this feat in just two days.?

• AI coding tools lead to sharp rise in code duplication, report finds: GitClear's latest report reveals AI coding tools are dramatically increasing code duplication and technical debt, threatening long-term software maintainability.?

• Crunchbase’s AI can predict startup success with 95% accuracy: Crunchbase is pivoting from a historical data provider to an AI-powered predictions platform, using data from 80 million users to forecast startup funding, acquisitions, and company growth with up to 95% accuracy.?

• AI vs. endpoint attacks—what security leaders must know to stay ahead: Enterprises must modernize their cybersecurity defenses as adversaries increasingly use AI and large language models to create sophisticated, automated attacks that outpace traditional security measures.

AI agents: The next wave of AI?

• Why agentic AI is the next big retail disruption: Agentic AI is transforming e-commerce through autonomous systems that can perform complex tasks like price comparison, customer service, and supply chain management with minimal human intervention, making it an ideal testing ground for this technology.?

• Invisible, autonomous, and hackable—the AI agent dilemma no one saw coming: AI agents present unique security challenges for enterprises as they require access to sensitive data while performing autonomous tasks, necessitating new approaches to identity management and access controls.?

• How to start using AI agents to transform your business: Business leaders looking to integrate AI agents into their operations should first identify specific business needs, choose the appropriate type of agent, and then carefully build and deploy these systems with proper oversight and refinement.?

Noteworthy AI applications and use cases?

• Grok 3 appears to be driving Grok usage to new heights: xAI's release of Grok 3 led to a significant surge in both mobile and web app usage, with downloads increasing tenfold and daily active users growing up to 260%, despite recent controversies.?

• Mistral’s Le Chat tops 1M downloads in just 14 days: Mistral's AI assistant Le Chat reaches one million downloads, becoming the top free iOS app in France.?

• Grok 3's early success signals shifts in AI model development and release cycles: xAI's Grok 3 emerges as a potentially leading AI model, demonstrating competitive performance across benchmarks while still in development, signaling faster release cycles in the AI industry.?

• Spotify opens platform to more AI-generated audiobooks: Spotify partners with ElevenLabs to accept AI-narrated audiobooks in 29 languages through its Findaway Voices distribution service. ?

• Nvidia helps launch AI platform for teaching American Sign Language: Nvidia launches Signs, an AI-powered platform to teach American Sign Language and build a validated dataset for developing accessible AI applications.?

• Voltron Data and Accenture partner to accelerate enterprise data processing: Voltron Data partners with Accenture to offer GPU-accelerated data processing that helps enterprises handle the massive datasets required for AI workloads more efficiently.?

• Flexport releases onslaught of AI tools: Flexport announces semi-annual product releases, launching over 20 AI-powered logistics tools while maintaining its commitment to human customer service.?

Platform and model developments?

• Small language models could redefine the AI race: The AI industry is shifting away from large language models toward smaller, more specialized models that offer better efficiency, cost-effectiveness, and specific business applications despite having less general knowledge.?

• AI wins 40% of a million-dollar budget in OpenAI's latest coding benchmark: OpenAI's new SWE-Lancer benchmark reveals that while current AI models can successfully handle basic programming tasks and earn significant revenue, they still struggle with complex software development challenges that require deep comprehension and holistic solutions.?

• Perplexity AI removes Chinese censorship from Deepseek R1: Perplexity AI has modified the Deepseek-R1 language model to create R1 1776, which maintains the original model's capabilities while removing Chinese censorship restrictions through specialized post-training on previously censored topics.?

• xAI’s new Grok 3 model criticized for blocking sources that call Musk, Trump top spreaders of misinformation: xAI faces criticism after users discover its Grok 3 AI model is programmed to avoid sources critical of Elon Musk and Donald Trump while remaining permissive about other potentially dangerous content.?

• How test-time scaling unlocks hidden reasoning abilities in small language models: New research shows that very small language models (SLMs) can outperform much larger models in reasoning tasks when using optimized test-time scaling techniques.?

• New AI model 'thinks' without words: Researchers develop Huginn, a new language model with recursive architecture that can adapt its computational depth to improve reasoning capabilities.?

• New S* framework helps AI models write better, more reliable code: Berkeley researchers developed a framework called S* that combines parallel and sequential scaling approaches to improve AI code generation, enabling even smaller language models to outperform larger ones in programming tasks. ?

AI regulations and public policy?

• US AI Safety Institute could face big cuts: NIST reportedly plans to lay off up to 500 staff members.?

• Court filings show Meta staffers discussed using copyrighted content for AI training: Court documents reveal Meta employees discussed and considered using copyrighted materials, including pirated books from Libgen, to train AI models despite legal risks. ?

• Chegg sues Google over AI search summaries: Chegg sues Google, claiming the tech giant's AI-powered search summaries unfairly use their content and have negatively impacted their traffic and revenue.?

Research??

• AI systems develop a sense of their own limitations with more time to "think": A Johns Hopkins University study found that giving AI systems more processing time improves both their answer accuracy and their ability to recognize when they should not attempt to answer questions.?

• Language models can overthink and get stuck in endless thought loops: A new study from several US universities and ETH Zurich?finds that AI language models can suffer from "overthinking" in interactive tasks, leading to decreased performance through analysis paralysis, rogue actions, or premature disengagement.?

• Researchers discover how AI models process time-specific knowledge: Researchers from Korea University, Upstage AI, and AIGEN Sciences discovered specialized "temporal heads" in LLMs that handle time-related information and could enable targeted updates of AI systems' temporal knowledge.?

Funding?

• Together AI’s $305M bet: Together AI raises $305 million in funding as demand for AI infrastructure increases, particularly due to DeepSeek-R1's resource-intensive reasoning capabilities.?

• Mercor raises $100M: AI recruiting startup Mercor raises $100 million in Series B funding led by Felicis, reaching a $2 billion valuation, as it expands its automated hiring platform used by major AI labs.?

• Augury raises $75M at $1B+ valuation: Augury raises $75 million to expand its AI-powered machine monitoring technology that predicts equipment repairs in manufacturing facilities.?

• Arize AI secures $70M Series C: Arize AI has secured a $70 million Series C investment, the largest ever in AI observability, to help companies better test and evaluate AI systems' reliability.?

• Patlytics raises $14M: Patlytics, a Google-backed startup, raises $14 million Series A funding to expand its AI-powered patent analytics platform that streamlines intellectual property workflows. ?

• Sawmills exits stealth with $10M: Sawmills has emerged from stealth with $10 million in seed funding to launch an AI-powered platform that helps companies manage and optimize their observability data costs and quality.?

• AI-coding startup Codeium in talks to raise at an almost $3B valuation, sources say: Codeium, an AI coding startup, is raising a new funding round led by Kleiner Perkins at a $2.85 billion valuation, just six months after its previous raise. ?