Crypto M - Crypto News
2.42K subscribers
15.9K photos
194 links
Your #1 destination for the latest and most unbiased market news on Bitcoin, Ethereum, NFT, Fintech, Web3, DeFi, and Blockchain.
Download Telegram
🚀 AI TRENDS | China's AI Model Call Volume Surpasses U.S. for Second Week

China's AI large model call volume reached 4.69 trillion tokens as of March 15, surpassing the United States for the second consecutive week, according to OpenRouter. Chinese models secured the top three global positions by call volume. According to NS3.AI, JPMorgan Chase predicts that China's AI inference token consumption will increase significantly, from around 10 quadrillion in 2025 to approximately 3,900 quadrillion by 2030.

#AI #China #ArtificialIntelligence #AImodels #TechTrends #MachineLearning #JPMorgan #NS3AI #OpenRouter #AIusage #GlobalTech
🚀 AI Integration in Trading Desks: Key Considerations

For trading desks incorporating AI, it is crucial to integrate it into actual workflows. FactSet posted on X, emphasizing the importance of governance, explainability, and trader adoption in the process. The article outlines strategies for firms to effectively implement AI, ensuring that these elements are addressed to facilitate successful integration.

#AIIntegration #TradingDesks #Governance #Explainability #TraderAdoption #AIImplementation #Fintech #FactSet #AIInFinance #MachineLearning
🚀 AI TRENDS | OpenAI Codex Team Undergoes Comprehensive System Overhaul

OpenAI Codex engineer Tibo has revealed that the Codex team is currently engaged in a nearly end-to-end comprehensive overhaul of the entire system. According to Jin10, this effort is aimed at accommodating the expansion of future model capabilities. Tibo expressed satisfaction with having Codex assist in these reconstructions, noting that without it, the process would have taken several months.

#AI #OpenAI #Codex #SystemOverhaul #MachineLearning #TechTrends #AIEngineering #ModelUpgrade
🚀 AI TRENDS | OpenClaw Releases Discrawl 0.2.0 with Enhanced Data Sync

OpenClaw founder Peter Steinberger announced on the X platform the release of Discrawl 0.2.0, highlighting significant improvements in data synchronization speed. According to Odaily, Steinberger has integrated this tool into daily operations to analyze key issues and user feedback within the OpenClaw Discord community.

#AI #OpenClaw #Discrawl #DataSync #MachineLearning #TechNews #PeterSteinberger #Discord
🚀 AI TRENDS | Eigen Labs Launches Beta Version of AgentKit for AI Development

On March 27, Eigen Labs released the first beta version of AgentKit, a toolkit designed for developers to create AI agents capable of holding assets, making payments, managing identities, and operating autonomously online. According to BlockBeats, the initial applications of AgentKit focus on 'autonomous media agents,' which can generate content such as text, images, audio, and video, manage their social presence, and operate continuously.

In the coming weeks, Eigen Labs plans to expand AgentKit's integration with popular social platforms, introduce new agent management templates, and launch a user-friendly interface. Future versions of AgentKit will also include native support for standards like x402 and ERC-8004.


#AI #AgentKit #EigenLabs #AutonomousAgents #AIDevelopment #BetaRelease #Blockchain #DigitalAssets #SocialMediaAgents #MachineLearning
🚀 Baidu's Apollo Go Launches Fully Autonomous Commercial Operations in Dubai

On March 30, Baidu's Apollo Go officially commenced its fully autonomous commercial operations in Dubai. According to Jin10, this marks a significant milestone for the company as it expands its autonomous driving technology beyond China. The launch in Dubai is part of Baidu's broader strategy to establish a global presence in the autonomous vehicle industry. Apollo Go aims to provide efficient and safe transportation solutions, leveraging advanced AI and machine learning technologies to navigate urban environments without human intervention. This development is expected to enhance Dubai's smart city initiatives and contribute to the region's technological advancement.

#Baidu #ApolloGo #AutonomousVehicles #Dubai #AI #SmartCity #Technology #Innovation #MachineLearning #Transportation
🚀 AI TRENDS | Zhipu Unveils GLM-5V-Turbo for Multimodal Programming

Zhipu has introduced the GLM-5V-Turbo, a multimodal coding model designed for visual programming. According to Odaily, this model natively understands various multimodal inputs, including images, videos, design drafts, and document layouts. It also supports the use of multimodal tools such as framing, screenshotting, and web page reading, with an expanded context window of up to 200,000.

#AI #Multimodal #Programming #Zhipu #GLM5V #Technology #Coding #MachineLearning #Innovation #DeepLearning
🚀 AI TRENDS | Microsoft Plans to Develop Advanced AI Model by Next Year

Microsoft is set to develop a cutting-edge artificial intelligence model by next year, aiming to create an internal alternative to the strongest AI tools from OpenAI and Anthropic. According to Jin10, Mustafa Suleyman, CEO of Microsoft AI, emphasized the need to deliver technology at the forefront of innovation. By 2027, the goal is for the model to achieve state-of-the-art capabilities in text, image, and audio generation and response.

On Thursday, Microsoft's AI division launched a speech transcription model that reportedly outperformed competitors in benchmark tests across 11 of the 25 most commonly used languages. However, similar to previous speech and image generation models released by the division, this model is designed as an efficient professional tool, utilizing less training data compared to general models like Claude 3 Opus or OpenAI's GPT-4.

Suleyman noted that Microsoft is consolidating computing power to develop models with broader capabilities. Since October last year, the company has been expanding its computational resources using a set of Nvidia GB200 chips. He stated, "Over the next 12 to 18 months, we will gradually enhance our computing capabilities to reach cutting-edge levels."


#AI #Microsoft #ArtificialIntelligence #MachineLearning #DeepLearning #SpeechRecognition #TextGeneration #ImageGeneration #AudioGeneration #Innovation #Technology #ComputingPower #Nvidia
🚀 OpenAI Acquires TBPN

OpenAI has reportedly acquired TBPN, according to Jin10. This acquisition marks a significant move in the AI industry, as OpenAI continues to expand its capabilities and resources. The details of the acquisition, including the financial terms and strategic implications, have not been disclosed. This development is expected to enhance OpenAI's technological offerings and strengthen its position in the competitive AI market.

#OpenAI #TBPN #Acquisition #AI #ArtificialIntelligence #TechIndustry #TechNews #Business #Innovation #MachineLearning
🚀 AI TRENDS | Zhejiang University Research Enhances AI Learning with Human Brain Signals

A research team from Zhejiang University has introduced a novel approach to training deep neural networks using human brain signals. According to NS3.AI, this method has demonstrated an average improvement of 20.5% in few-shot learning and abstract concept recognition in new contexts. The study also revealed that while larger models enhance accuracy in concrete concepts, they tend to decrease accuracy in abstract concepts.

#AI #DeepLearning #NeuralNetworks #BrainSignals #FewShotLearning #AbstractConcepts #MachineLearning #ZhejiangUniversity #ArtificialIntelligence
🚀 AI TRENDS | MemPalace AI Tool Gains Traction with 10,000 GitHub Stars

Milla Jovovich announced her involvement in developing MemPalace, an open-source AI knowledge tool that has quickly gained popularity. According to NS3.AI, the tool, which organizes documents in virtual rooms rather than using traditional keyword searches, achieved 10,000 GitHub stars and 50 pull requests within 24 hours of its release. Ben Sigman highlighted this rapid adoption as a significant milestone.

Sean Ren, a computer science professor at USC and CEO of Sahara AI, commented on the potential of MemPalace's method to be applied across various AI frameworks. However, he noted that the performance claims of the system have yet to be validated outside of controlled testing environments.


#AI #MemPalace #OpenSource #GitHub #MachineLearning #ArtificialIntelligence #TechTrends #KnowledgeManagement #VirtualTools #Innovation
🚀 AI TRENDS | Xverse Launches AI-Driven Bitcoin Wallet for Enhanced Transactions

Bitcoin wallet Xverse has announced the introduction of its AI-powered Agentic Wallet, designed to facilitate payments, transactions, and fund management on the Bitcoin network. According to Foresight News, the wallet utilizes Spark for lightning payments and supports machine-payable APIs through MPP, featuring a fully self-custodial architecture.

#AI #Bitcoin #Cryptocurrency #Fintech #Wallet #Blockchain #Payments #MachineLearning #DigitalCurrency #BTC
🚀 AI TRENDS | Tether Launches QVAC SDK for Cross-Platform AI Development

Tether has introduced the QVAC SDK, a unified software development kit designed to enable developers to build, run, and fine-tune AI applications directly on any device. According to Foresight News, this SDK ensures consistency across different environments.

Applications developed using the QVAC SDK can seamlessly operate on platforms such as iOS, Android, Windows, macOS, and Linux. The same codebase can function across all supported environments without the need for platform-specific branches, rewrites, or conditional logic.

The QVAC SDK is built on QVAC Fabric, a branch of llama.cpp, offering broad compatibility with the llama.cpp model ecosystem for text generation, embedding, and multimodal workloads.


#AI #SDK #CrossPlatform #MachineLearning #LlamaCpp #SoftwareDevelopment #Multimodal #QVAC
🚀 AI's Impact on Investment and Trading: Insights from Nansen CEO

PANews posted on X (formerly Twitter) about a discussion with Nansen CEO Alex Svanevik on the evolving role of AI in investment and trading. Svanevik highlighted that 'smart money 2.0' is transforming into a predictive system, with agent trading expected to surpass human trading by 2028. However, he emphasized the need for users to build a 'trust ladder' before fully relying on trading agents.

The conversation also covered the implementation of tools like OpenClaw in enterprise settings, where safety is prioritized over speed. Svanevik shared insights on how the Nansen team utilizes OpenClaw and how AI is reshaping team structures. He noted that 'judgment' is becoming the most scarce resource within AI-native companies.

Svanevik further pointed out that low latency, overcoming AI bottlenecks, and open-source solutions will define the next generation of agent infrastructure.


#AI #Investment #Trading #FinTech #MachineLearning #PredictiveAnalytics #OpenSource #EnterpriseAI #FinancialTechnology #AlgorithmicTrading
🚀 Security Concerns Arise Over LLM Agent API Routers

On April 10, Solayer founder @Fried_rice highlighted on social media the growing reliance of large language model (LLM) agents on third-party API routers, which distribute tool call requests to multiple upstream providers. According to BlockBeats, these routers operate as application layer proxies and can access each JSON payload in plaintext during transmission. However, no provider currently enforces encryption integrity protection between the client and upstream models.

A study tested 28 paid routers purchased from platforms like Taobao, Xianyu, and Shopify independent sites, along with 400 free routers collected from public communities. The findings revealed that one paid router and eight free routers were actively injecting malicious code. Additionally, two routers deployed adaptive evasion triggers, 17 accessed AWS Canary credentials owned by researchers, and one stole ETH from a private key held by researchers.

Further poisoning studies demonstrated that seemingly harmless routers could also be exploited. A leaked OpenAI key was used to generate 100 million GPT-5.4 tokens and over seven Codex sessions. Weaker bait configurations resulted in 2 billion billing tokens, 99 credentials across 440 Codex sessions, and 401 sessions running autonomously in YOLO mode.

The research team developed an experimental proxy named Mine, capable of executing all four types of attacks on four public proxy frameworks. They also verified three client defense strategies: fault lock strategy gating, response-side anomaly screening, and append-only transparent logging.


#LLM #API #Security #CyberSecurity #Malware #DataBreach #Encryption #Proxy #AI #MachineLearning #ETH