Top News
Tesla Full Self-Driving Comes Out Of Beta, But Must Be Supervised
After recent significant improvements to its self-driving capabilities, Tesla has officially removed the “Beta” label from its Full Self-Driving (FSD) software, now referring to it as “Full Self-Driving (Supervised).” This change comes after a year of anticipation following CEO Elon Musk’s announcement of the upcoming shift. The new name aims to clarify that while the vehicle can make certain decisions independently, it still requires supervision from the driver. Despite the name change, Tesla has faced criticism from regulatory bodies worldwide for its choice of branding, with some arguing that it is misleading and has led to a series of legal issues. The company has also expanded access to FSD by offering a one-month free trial to all owners with compatible vehicles, potentially increasing the adoption rate of the $12,000 add-on.
‘The machine did it coldly’: Israel used AI to identify 37,000 Hamas targets
The Israeli military used an AI-powered database, named Lavender, to identify 37,000 potential targets linked to Hamas during its bombing campaign in Gaza, according to intelligence sources. Developed by the Israel Defense Forces’ elite intelligence division, Unit 8200, Lavender processed large amounts of data to identify potential “junior” operatives. The system’s use has raised legal and moral questions about the relationship between military personnel and machines in warfare. The intelligence sources also claimed that the Israeli military had a preference of bombing targets when they were at home and allowed a large number of Palestinian civilian casualties, particularly in the early stages of the conflict, with pre-authorized allowances for estimated civilian deaths before a strike was authorized.
Amazon Ditches ‘Just Walk Out’ Checkouts at Its Grocery Stores
Amazon is discontinuing its “Just Walk Out” technology in its grocery stores, a system that used cameras and sensors to track customers’ purchases and allowed them to leave the store without physically checking out. The technology was not reliable enough and depended on over 1,000 people in India reviewing and labeling videos for accuracy, sometimes manually assigning shopping items to individual shoppers. “Just Walk Out†will be replaced by Dash Carts, a scanner and screen embedded in shopping carts that allows customers to check out as they shop. Despite the change, Amazon will continue to use the “Just Walk Out” technology in a few of its Amazon Fresh stores in the UK and some Amazon Go convenience stores.
A conversation with OpenAI’s first artist in residence
MIT Tech Review interviews Alexander Reben, the first artist in residence at OpenAI. Reben has been working with the AI firm for several years, even before his official appointment in January. Reben was invited to experiment with an early version of GPT-3, OpenAI’s language prediction model, five years ago and has since created several artworks using the system. With a background in math and robotics from MIT’s Media Lab, Reben is particularly interested in the interaction between humans and machines. He views AI as a tool rather than a defining aspect of his work, and is intrigued by how humanity and technology co-evolve over time. Currently, he serves as the director of technology and research at Stochastic Labs, a nonprofit incubator for artists and engineers.
Other News
Tools
Introducing Stable Audio 2.0 – Stable Audio 2.0 introduces a new AI-generated audio model that can produce high-quality, full tracks with coherent musical structure up to three minutes in length at 44.1kHz stereo, expanding creative possibilities for artists and musicians.
OpenAI’s Sora just made its first music video and it’s like a psychedelic trip – OpenAI’s Sora AI engine creates a psychedelic music video for the song “Worldweight” by August Kamp, showcasing its potential for unique and ethereal visual storytelling, while also raising questions about its widespread adoption and impact on content creation.
Microsoft’s new safety system can catch hallucinations in its customers’ AI apps – Microsoft has developed new safety features for Azure AI to detect and block hallucinations, malicious prompts, and vulnerabilities, aiming to prevent controversial and unintended responses from AI models.
Microsoft 365’s Copilot gets a GPT-4 Turbo upgrade and improved image generation – Microsoft 365’s Copilot AI assistant is getting a GPT-4 Turbo upgrade, providing priority access to the latest model for business subscribers with improved image generation and expanded capabilities for analyzing documents.
OpenAI makes it easier for developers to fine-tune and build AI models – OpenAI expands its Custom Models Program and fine-tuning API, offering assisted fine-tuning and custom-trained models to help organizations optimize AI models for specific use cases.
Microsoft is working on an Xbox AI chatbot – Microsoft is testing an AI-powered Xbox chatbot for support tasks, part of a broader effort to bring AI features to the Xbox platform and developer tools, including AI-generated art and assets for games, AI game testing, and AI-powered Copilots for safety and moderation tasks.
With Brave Leo on iOS today, the browser AI assistant is now available on all platforms – Brave’s privacy-preserving AI assistant, Leo, is now available on iOS, allowing users to ask questions, summarize pages, translate content, create written content, and more directly from the browser address bar, with voice-to-text capability and access to high-quality large-language models.
Opera allows users to download and use LLMs locally – Opera now allows users to download and use Large Language Models (LLMs) locally on their computer, selecting from over 150 models from more than 50 families, as part of its AI Feature Drops Program.
ChatGPT no longer requires an account – but there’s a catch – OpenAI’s ChatGPT is now accessible without an account, but users will have limited features and their chats will still be used for training data unless they opt out.
Introducing Command R+: A Scalable LLM Built for Business – Introducing Command R+, a state-of-the-art RAG-optimized model designed for enterprise-grade workloads, with multilingual support, advanced tool use, and a commitment to data privacy and security.
Business
In a first, FDA approves AI-driven test to predict sepsis in hospitals – FDA approves AI-driven test to predict sepsis in hospitals, developed by Prenosis, a company that trained its algorithm to recognize health measures associated with developing sepsis and now produces a snapshot that classifies a patient’s risk of sepsis in four categories.
Will A.I. Boost Productivity? Companies Sure Hope So. – Artificial intelligence’s potential to boost productivity is questioned by economists and officials, who are skeptical that it has spread enough to show up in productivity data already.
Apple Explores Home Robotics as Potential ‘Next Big Thing’ After Car Fizzles – Apple is exploring the potential of home robotics, with teams working on automated devices that can follow users around their homes and move a display around.
Waymo self-driving cars are delivering Uber Eats orders for first time – Waymo self-driving cars are now delivering Uber Eats orders in the Phoenix metropolitan area, marking the official launch of their delivery partnership.
Yahoo leaps toward AI-powered news with Artifact acquisition – Yahoo acquires AI news startup Artifact to integrate its recommendation engine into Yahoo News, aiming to deliver curated content based on individual news-reading habits.
Sam Altman gives up control of OpenAI Startup Fund, resolving unusual corporate venture structure – Sam Altman has transferred control of the OpenAI Startup Fund to Ian Hathaway, resolving an unusual corporate venture structure.
Research
Many-shot jailbreaking – A study investigates a “jailbreaking” technique called “many-shot jailbreaking” that exploits the longer context window of large language models, leading to potentially harmful responses, and discusses various mitigation strategies.
ReALM: Reference Resolution As Language Modeling – Using language modeling, this paper demonstrates the effectiveness of LLMs in resolving references of various types, including non-conversational entities, and shows significant improvements over existing systems.
Aurora-M: The First Open Source Multilingual Language Model Red-teamed according to the U.S. Executive Order – Aurora-M is a 15B parameter multilingual open-source model trained on multiple languages and code, continually pretrained from StarCoderPlus, and rigorously evaluated to align with the concerns articulated in the U.S. Executive Order on the Safe, Secure, and Trustworthy Development and Use of Artificial Intelligence.
Showing AI just 1000 extra images reduced AI-generated stereotypes – Feeding AI just 1000 extra images reduced AI-generated stereotypes and made the image generator more culturally sensitive and accurate.
Bigger is not Always Better: Scaling Properties of Latent Diffusion Models – Smaller latent diffusion models often outperform larger ones in generating high-quality results, leading to new strategies for enhancing generative capabilities within limited inference budgets.
Octopus v2: On-device language model for super agent – On-device language model Octopus v2 surpasses GPT-4 in accuracy and latency, reducing context length by 95% and enhancing latency by 35-fold compared to Llama-7B with a RAG-based function calling mechanism.
CityGaussian: Real-time High-quality Large-Scale Scene Rendering with Gaussians – Real-time high-quality large-scale scene rendering with Gaussians is achieved through CityGaussian, a technology that embraces openness, community, excellence, and user data privacy.
Concerns
Big Tech usually dismisses fears that AI kills jobs. Now it’s studying them. – Big Tech companies are forming a group to study how AI might affect technology jobs, as concerns about AI taking away jobs from humans grow more urgent.
Meta’s AI image generator really struggles with the concept of interracial couples – Meta’s AI image generator struggles to accurately create images of interracial couples, generating mostly same-race couples despite repeated attempts to prompt it otherwise.
Google Books Is Indexing AI-Generated Garbage – Google Books is indexing low quality, AI-generated books that could impact Google Ngram viewer and may change the understanding of cultural change reflected in books.
Policy
Billie Eilish, Pearl Jam, Nicki Minaj Among 200 Artists Calling for Responsible AI Music Practices – Over 200 artists, including Billie Eilish, Pearl Jam, and Nicki Minaj, have signed an open letter urging tech companies and AI developers to use artificial intelligence responsibly and cease practices that devalue human creativity in music.
Washington state judge blocks use of AI-enhanced video as evidence in possible first-of-its-kind ruling – Judge in Washington state blocks use of AI-enhanced video as evidence in a triple murder case, citing the technology’s opaque methods and potential to confuse eyewitness testimony.
Meta overhauls rules on deepfakes, other altered media – Meta announces major changes to its policies on digitally created and altered media, including the application of “Made with AI” labels and more prominent labels for high-risk altered media, ahead of U.S. elections.
Expert Opinions
An A.I. Researcher Takes On Election Deepfakes – AI researcher Oren Etzioni warns about the threat of AI-generated deepfakes in influencing elections and releases free tools to identify digital disinformation.
Copyright © 2024 Skynet Today, All rights reserved.
Source: Read MoreÂ
