AI News on June 27th
1. Salesforce Accelerates AI Workload Automation, 30%-50% of Tasks Automated with 93% Accuracy
#Nvidia #ChipShortage #DeepSeek #AI #ExportControls
Salesforce CEO Marc Benioff announced that the company is accelerating AI-driven workload automation, which now handles 30%-50% of tasks with approximately 93% accuracy. He described AI’s rise in the workforce as a "digital labor revolution," enabling employees to focus on higher-value work. Earlier this year, Salesforce restructured around AI, cutting over 1,000 jobs. Other tech firms like Klarna (reducing staff by 40% due to AI investments) and Amazon (planning AI-driven job cuts) have taken similar steps.
2. U.S. Export Controls Cause Nvidia Chip Shortages, DeepSeek R2 Model Launch Potentially Delayed
#Nvidia #ChipShortage #DeepSeek #AI #ExportControls
Due to U.S. export restrictions, Nvidia GPU shortages may delay the release of DeepSeek’s highly anticipated next-gen large language model, R2. A Chinese cloud service provider distributing DeepSeek’s models stated that domestic Nvidia server chip shortages could slow R2’s widespread adoption compared to its predecessor, R1. DeepSeek has not yet confirmed a release date.
3. FLUX.1 Kontext [dev] Goes Open-Source! 12B-Parameter Model Rivals Top Closed-Source Tools
#OpenSource #AI #ImageEditing #BlackForestLabs #FLUX1
12-billion-parameter model that delivers image-editing performance comparable to proprietary tools while running on consumer hardware. Released under a non-commercial license, it supports research and non-commercial use, integrates with ComfyUI, Hugging Face Diffusers, and TensorRT, and is now available on Hugging Face with ready-to-use API endpoints.
4. Meta Hires OpenAI’s Key Researcher Bansal to Boost AI Reasoning Model Development
#Meta #OpenAI #AIRecruiting #AIReasoning
Meta has recruited former OpenAI researcher Trapit Bansal, a key contributor to OpenAI’s first reasoning model (o1), to join its new AI superintelligence team. Bansal’s expertise will help Meta develop cutting-edge reasoning models to compete with industry leaders. OpenAI’s upcoming open AI reasoning model may intensify competition.
5. CoreWeave CEO Becomes Billionaire in 3 Months as AI Cloud Firm’s Stock Soars 300%
#CoreWeave #IPO #AI #CloudComputing #Nvidia
CoreWeave CEO Michael Intrator saw his net worth surge to ~$10B within three months of the company’s IPO, the largest tech IPO of 2025 ($1.5B raised). The Nvidia GPU-powered AI cloud provider, serving Microsoft and OpenAI, has seen its stock jump nearly 300% since March despite $8.8B in debt and a $315M Q1 net loss.
6. Ant Group Launches "AQ" Health AI: Medical LLM Powers Consultations, 200+ Hospitals Offer Virtual Visits
#AntGroup #HealthcareAI #MedicalLLM #Telemedicine
Ant Group unveiled "AQ", an AI health app powered by its medical LLM, which mimics doctor-patient interactions, connects users to 5,000+ hospitals and nearly 1M doctors, and offers virtual visits from 200+ hospitals. It also features AI avatars of top physicians, voice-call support for seniors, and personalized health tracking via wearables.
7. Alibaba’s FY2025 Revenue Nears ¥1 Trillion, Net Profit Up 77% to ¥125.98B
#Alibaba #Earnings #AI #CloudComputing
Alibaba reported FY2025 revenue of ¥996.35B ($137B), with net profit soaring 77% YoY to ¥125.98B ($17.3B). Strong AI demand drove double-digit cloud growth, while Taobao’s 88VIP membership surpassed 50M. International commerce revenue rose 29%, and open-source AI models like Tongyi gained traction. Alibaba plans heavy AI investments, positioning "AI + Cloud" as a growth engine.
8. Hengbot’s Sirius Robotic Dog: Plays Soccer, Dances, Integrates OpenAI LLM for Voice Chat
#Robotics #AI #OpenAI #Hengbot
Hengbot launched Sirius, an agile robotic dog with 14 motion axes, capable of soccer, dancing, and natural movement via "Neurocore" joints. It integrates OpenAI’s LLM for personality simulation, voice interaction, and data privacy. Priced at $1,299 (pre-order), it offers 40-60 mins of active use and ships this fall.
9. Google Releases Open AI Model Gemma 3n: Multimodal, Runs on 2GB RAM Devices
#Google #Gemma #OpenAI #MultimodalAI
Google unveiled Gemma 3n, an open, lightweight AI model supporting text, image, audio, and video processing on devices with as little as 2GB RAM. Optimized for coding and reasoning, it uses MatFormer architecture for efficiency and is available on Hugging Face. Unlike Gemini, Gemma is designed for developer customization.
10. Gartner Predicts Over 40% of Agent AI Projects Will Be Scrapped by 2027 Due to Cost, ROI Issues
#Gartner #AITrends #AgentAI #ROI
Gartner forecasts that >40% of agent AI projects will be canceled by 2027 due to high costs, unclear value, or poor risk control. Many current "AI agent" initiatives are rebranded hype, with immature models offering limited ROI. By 2028, 15% of daily work decisions may be agent-driven, and 33% of enterprise apps will embed AI agents.