FlagThis AI

Home Assistant has released a new voice assistant device called ‘Voice Preview Edition’. This privacy-focused device enables local voice control for smart homes, processing commands offline. It supports custom wake words like ‘Hey Jarvis’ and can function without an internet connection, emphasizing user privacy and local processing capabilities. The device is built around a ESP32-S3 chip and an XMOS XU316 audio processor and available for $59.

UnitedHealthcare’s Optum had an AI chatbot used by employees exposed to the internet. This chatbot, designed for employees to inquire about claims, was accessible publicly. The exposure raises concerns about the security of sensitive data and the potential for unauthorized access. This incident highlights the risks associated with deploying AI tools without adequate security measures. The AI chatbot exposure occurred amid broader scrutiny of UnitedHealthcare for its use of AI in claims denials.

Microsoft’s new AI feature ‘Recall’ for Copilot+ PCs stores screenshots of sensitive data, including credit cards and social security numbers, even when a ‘sensitive information’ filter is enabled. This has raised serious privacy and security concerns among users. This feature takes continuous screenshots of everything a user does. The data is stored locally but sent off to Microsoft’s LLM for analysis. This has prompted an investigation by the UK Information Commissioner’s Office. This incident highlights the potential risks of AI-powered surveillance features and the importance of user privacy.

OpenAI introduced ‘Santa Mode’ in ChatGPT and Advanced Voice Mode with live video and screen sharing features for ChatGPT Plus, Team and Pro users. Santa Mode is a temporary voice feature for the month of December which lets users talk to ChatGPT with the voice of Santa Claus. Additionally, the Advanced Voice Mode was expanded to include live video and screen sharing capabilities, enhancing the user interface for collaborative tasks. These features mark an advancement in how users interact with AI and its potential in practical applications. These features were launched after an earlier global outage of ChatGPT services, which was also resolved by OpenAI.

DeepMind’s Genie 2 is a generative AI model that creates rich, interactive 3D environments from text or images. While limited to brief simulations, it excels as a creative prototyping tool and for AI agent evaluation. The model raises questions about intellectual property and ethical use but represents a major advancement in AI-driven world modeling. It’s a significant leap in AI’s ability to generate and interact within 3D virtual worlds, with implications for game development, virtual reality, and AI research. Further development could lead to more realistic and complex simulations, pushing the boundaries of AI capabilities and opening new avenues for research and application.

NVIDIA and AWS are collaborating to accelerate AI and robotics development in the cloud. This partnership integrates NVIDIA’s CUDA-Q platform with Amazon Braket, allowing for development and testing of hybrid quantum-classical workflows using GPU-accelerated simulators.

Meta has joined Elon Musk in opposing OpenAI’s transition to a for-profit company, arguing it could have major consequences for Silicon Valley. Meta sent a letter to California Attorney General Rob Bonta expressing their concerns that this shift would have seismic implications for the tech industry. This alliance between Meta and Musk highlights the ongoing debate and scrutiny over the ethical and competitive landscape of AI development.

OpenAI’s new o1 model and its Pro version offer enhanced math, coding and image processing capabilities. The o1 model is available to ChatGPT Plus and Team users, whereas the Pro version offers more advanced features. This upgrade marks a significant step in AI model evolution, showcasing improved reasoning and multimodal functionality. The Pro version of the model appears to use multiple attempts to get better answers, and offers significantly increased usage, higher resolution, and longer duration options.

Apple Intelligence’s notification summaries are generating inaccurate and false news. This issue is creating misleading summaries, leading to concerns about the reliability of AI-generated content. The BBC has filed a formal complaint due to their news being misrepresented, highlighting the need for strict quality control in AI-driven news aggregation and summarization. This incident raises serious questions about the responsible deployment of AI in news dissemination and the potential consequences of misinformation from trusted sources.

Former OpenAI researcher Suchir Balaji, who raised concerns about the company’s copyright practices related to training AI models, was found dead in his San Francisco apartment. His death has sparked discussions about the ethics of AI training data and the impact on online content creators. Balaji’s work involved gathering data for models like GPT-4, and he expressed concerns about the potential harm to online communities, particularly due to the free copying of data used in training AI models, and the implications of fair use.

Waymo is expanding its autonomous vehicle testing to Tokyo, marking its first international deployment. This initiative is in partnership with local taxi operators and is part of Waymo’s “road trips” program. The tests will focus on mapping key areas of Tokyo and assessing the performance of their AI systems in a new and complex urban environment. The move signifies a major step for Waymo in their global expansion plans and the challenges of adapting self-driving technology for diverse international markets.

A survey from Tray.ai indicates that many enterprises are not fully prepared to support autonomous AI agents. The survey shows that most enterprises require a technology stack upgrade to properly deploy AI agents. These results highlight the need for better planning and infrastructure to effectively leverage AI agents.

Google has introduced a new tool to detect AI-faked celebrities on YouTube by using a database called the CAA Vault that contains digital copies of celebrities’ faces, bodies and voices. Google has also launched an AI video generator, Veo 2, which they claim has better audience scores than OpenAI’s Sora. Additionally, they have debuted a new version of their image generation model Imagen 3 which produces richer, more detailed photos. They have also launched a new feature in NotebookLM which allows users to talk with the AI hosts.

YouTube is developing a tool in partnership with Creative Artists Agency (CAA) to enable creators and celebrities to detect and manage AI-generated content that uses their likeness, including faces and voices. The tool will allow them to submit removal requests for unauthorized AI-generated content. CAA provides a database of digital copies for the celebrities which helps find the AI deepfakes. This initiative aims to address the growing issue of deepfakes and ensure content creators have more control over the use of their digital identities online.

NVIDIA has released the Jetson Orin Nano Super Developer Kit for $249, targeting generative AI applications at the edge. It offers 1.7 times more generative AI performance, reaching 67 INT8 TOPS, making it suitable for robotics and other edge AI tasks. This kit is aimed at developers and hobbyists looking for a cost-effective yet powerful platform for AI development. The product is designed to disrupt the market with high performance at a lower price, targeting the generative AI at the edge market.

Backflip, a startup specializing in AI-powered 3D modeling, has secured $30 million in Series A funding. The company is developing AI models that can generate 3D designs using text, sketches, or photos, aiming to simplify the design process for engineers. The funding will support further development and expansion of its AI-driven platform for 3D model creation.

Google has unveiled Gemini 2.0 Flash, a new AI model that is twice as fast as Gemini 1.5 Pro and optimized for speed and multimodal functionality. It supports real-time interaction with video and is designed to be the foundation for future AI agents. Google also released Android XR, a new operating system for virtual and augmented reality devices.

Waymo’s autonomous vehicles are shown to be safer than human-driven vehicles based on a study with Swiss Re, a major insurance company. The study found a 92% reduction in bodily injury claims and a significant decrease in property damage claims when comparing Waymo’s autonomous driving data with traditional vehicles. The research analyzed 25.3 million fully autonomous miles, highlighting the safety benefits of self-driving technology.

GitHub has launched a free tier for its Copilot AI code completion tool, offering developers up to 2,000 code completions and 50 chat messages per month. This new free tier includes access to both GPT-4o and Claude 3.5 Sonnet models, making advanced AI-assisted coding accessible to a broader audience. This initiative aims to expand the reach of AI capabilities to more developers and further encourage adoption of AI in software development workflows.

OpenAI’s new o3 model has achieved a breakthrough performance on the ARC-AGI benchmark, demonstrating advanced reasoning capabilities through a ‘private chain of thought’ mechanism. The model searches over natural language programs to solve tasks, with a significant increase in compute leading to a substantial improvement in its score. This approach highlights the use of deep learning to guide program search, pushing the boundaries beyond simple next-token prediction. The o3 model’s ability to recombine knowledge at test time through program execution suggests a significant step towards more general AI capabilities.