Artificial intelligence, especially biased AI, poses discrimination risks in decision-making for professions like healthcare, law, and finance. Bias correction methods like algorithmic de-biasing and retrieval-augmented generation are being developed but their effectiveness is uncertain.

Google's "Model Explorer" tool unveils the secrets of complex AI systems, enhancing transparency. It offers hierarchical visualization for navigating large neural networks, promoting accountability and enabling deployment on mobile devices. This tool helps developers identify and solve issues quickly, supporting responsible AI deployment. Google's initiative to bring AI to devices emphasizes transparency and accountability, crucial for building user trust as AI advances rapidly. The "Model Explo

Key OpenAI researcher Jan Leike resigns, citing compromising safety culture for flashy products. He led dissolved Superalignment team focused on AI risks. Concerns about AGI impact prompt transition to proprietary tech. John Schulman takes over Leike's responsibilities. Tensions rise over developing AGI vs. safety protocols. Leike criticizes lack of resources for crucial work.

At Google I/O, "AI Overviews" were introduced, generating real-time summaries similar to Perplexity. Srinivas discussed competitors, OpenAI launching generative AI search, and Perplexity hiring industry leaders like Parakhin. OpenAI partners with Reddit, while Google earns from search. OpenAI aims to build a search engine, avoiding high user charges. Potential revenue growth is predicted. OpenAI CEO looks to crack the code of large language models and search. Venkataraman joins OpenAI as VP with

Google introduces AI theft detection, part of Android 15 update. New feature uses phone sensors to lock device if stolen. Enhanced security measures coming. Additional auth steps, private app space. Offline device locking to prevent remote locking prevention. Nerve-wracking phone theft experience addressed with remote locking option using phone number and prompts. AI added to search engine, Veo converts text to videos.

IMF Chief Economist warns AI impact on labor market, with potential 60% job disruption in developed economies. Urges preparation for productivity boost and social inequality risk. Microsoft survey shows 50% of employees fear AI, but 75% find it useful at work. Amazon doubles robots in fulfillment centers, creating new job opportunities. Need for government and tech companies to regulate AI evolution.

Ampere Computing to launch 256-core CPU next year; partnering with Qualcomm for cloud AI accelerators. Energy-efficient AI solutions key focus for data center sustainability. Collaboration with Qualcomm for AI processors. AmpereOne CPU offering 40% performance improvement.

OpenAI introduces GPT-4o, a model advancing human-machine interaction with real-time multitasking across text, audio, and visuals, enhancing applications like translation, customer service, and education.

At Tencent Cloud Summit, VP Jiang Jie disclosed Hybird Model's top-tier performance in AI, rivaling global GPT-4. The model excels in video generation, 3D modeling, as well as Google's Veo. AI video generation is set to revolutionize various industries.

Microsoft's climate goals face challenges as its greenhouse gas emissions rise by 30% in FY2023. The focus on AI is contributing to increased pollution, making it harder for the company to achieve carbon neutrality by the end of the century. This shift to AI has led to a $13 billion investment in OpenAI and a $50 billion commitment in the past fiscal year.

Xiaomi's self-developed large language model, MiLM, has passed filing and will be integrated into various products. MiLM-6B has 6.4 billion parameters, excelling in STEM but needing improvement in abstract subjects. Xiaomi plans to enhance the model for better performance.

OpenAI's new GPT-4o model offers faster, more cost-effective image generation with improved realism and accuracy. It analyzes multimedia data directly without text conversion, setting it apart from previous models like DALL-E 3. Despite not being publicly available yet, OpenAI is working to make its image generation capabilities accessible soon.

Meizu launched the new Flyme AIOS system, integrating cutting-edge AI tech for improved user experience. Features include AI search, call assistant, image processing, and an intelligent agent store. The update aims to enhance user personalization and convenience in smart devices.

UK-based AI startup Stability AI is reportedly facing financial challenges, with Q1 2024 revenue below $5 million and losses exceeding $30 million. The company owes nearly $100 million to partners, prompting discussions for a potential sale. Industry analysts are closely monitoring the situation for implications on Stability AI's future in the competitive AI market.

Vivo upgrades "vivo Vision" with Lanxin large model integration, becoming industry's first multimodal visual assistance product. Supports voice broadcast, text recognition, barcode scanning. Enhances accessibility for visually impaired. Launches "Accessibility Care Month" event.

OpenAI partners with Reddit, leveraging user-generated content for ChatGPT. The collaboration aims to improve AI training with diverse, up-to-date data. Reddit gains AI-driven features and financial benefits. The move reflects a trend of AI companies seeking quality training data.

OpenAI enhances data analysis in ChatGPT with GPT-4o, allowing direct file uploads from cloud services. Users can interact with tables/charts in an expandable view, customize/download charts, and ensure security/privacy. Updates will be available soon for all users.

Hugging Face invests $10 million in free shared GPUs for AI development, aiming to support small developers and counter tech giants' dominance. ZeroGPU project provides cost-effective, energy-efficient Nvidia A100 GPU devices through Hugging Face's hosting platform Spaces to promote community-wide accessibility of state-of-the-art AI technology. Delangue emphasizes the importance of open-source AI to level the competitive landscape.

Hugging Face introduces LeRobot, a machine learning model for real-world robot applications. It offers a versatile library for functions like data sharing and model training. LeRobot simplifies project initiation and aims to lower the barrier to robot learning. Operated as an open-source project on GitHub, it encourages global participation.datasets available cover various scenarios for robot learning tasks. Its potential in simplifying robot development makes it a promising resource for the AI

This week, OpenAI and Google presented their latest AI technologies focusing on multimodal capabilities for more natural interactions. GPT-4 Omni by OpenAI processes video and audio simultaneously, while Google's Project Astra is still in early stages. The race for advanced AI models has OpenAI leading with GPT-4o, which handles audio, video, and text within a single model. AI wearables like Humane AI Pin and Rabbit R1 are emerging with multimodal features.