AI Apocalypto Newsletter | Week 17
AI News This Week | AI Insights
Welcome to Week 17 of AI Apocalypto, where we explore the latest in AI technology, from ethical applications and groundbreaking advancements to the risks and unethical uses. This week, we cover AI's role in enhancing healthcare, security, and multimedia, while also examining potential pitfalls and innovations.
Ethical Usage of AI
Discover how AI is revolutionizing healthcare and financial security, with Visa's VAAI score and Lexi Bogan's speech restoration.
South Korea to Host Second Global AI Safety Summit
• Follows Britain's inaugural event.
• Aims to foster international cooperation on AI safety.
• Future summits every six months, with France hosting in 2024.
Visa Introduces VAAI Score
• Utilizes generative AI to detect real-time enumeration attacks.
• Trained on 15 billion transactions.
• Assign risk scores based on historical patterns.
• Reduces false positives by 85%.
• Initially available to U.S. financial institutions.
MIT and University of Basel Develop AI Framework
• Uses generative models for phase diagram mapping.
• Approximates probability distributions for phase transition detection.
• Offers a more knowledgeable approach than discriminative methods.
AI Aids Lexi Bogan's Speech Restoration
• Brain tumor removal impaired speech.
• AI restores voice using a teenage recording.
• The app uses OpenAI's Voice Engine technology.
• Promising approach potentially aids millions worldwide.
AI in Primary Care
• Aids in early detection of chronic kidney disease.
• Accelerates drug development through data analysis.
• Importance of ethical AI use and collaboration.
Unethical Usage of AI
Google's real-time voice call scanning sparks privacy concerns, highlighting the risks of centralized censorship and surveillance.
OpenAI Dissolves Superalignment Team
• Key members Ilya Sutskever and Jan Leike leave.
• Work integrated into other research efforts.
• Disagreements over company priorities and resources.
Google's AI Demo: Real-Time Voice Call Scanning for Financial Scams
• Sparks privacy concerns.
• Represents Apple's abandoned CSAM scanning plan.
• Experts question centralized censorship.
• Raises questions about privacy erosion and societal control through AI surveillance.
Time Magazine covers how to Pause AI
• AGI poses significant risks.
• ChatGPT's 2022 release spurred AI advancements.
• Importance of AI safety and alignment.
• Proactive measures needed to mitigate threats.
Department of Justice Concerns Over AI's Election Interference
• Attorney General Garland and Deputy Attorney General Lisa Monaco express concerns.
• Pledge collaboration with the FBI for defense.
• Proposal for enhanced sentences for AI-manipulated election criminals.
Sony Music Group Warns AI Companies
• Warns against AI use of content.
• Sends letters to over 700 companies.
• Protects intellectual property from AI training.
AI Advancements
OpenAI's GPT-4o model leads the latest AI innovations, with South Korea hosting a global AI safety summit and new releases from Google and Samsung.
OpenAI Unveils GPT-4o AI Model
• Enables realistic voice conversation and integration across text and image.
• Upgraded ChatGPT offers real-time voice interactions.
• Aim to expand user base.
• Comes amid competition pressures.
• Ahead of Alphabet's AI-related announcements.
OpenAI Co-Founder Ilya Sutskever Leaves
• Sutskever leaves after a leadership crisis with Sam Altman.
• Jakub Pachocki replaces Sutskever as chief scientist.
• Sutskever's departure coincides with the AI model launch and ChatGPT feature expansion.
OpenAI's GPT-4 Passes Turing Test
• Ethereum co-founder Vitalik Buterin confirms.
• UC San Diego researchers find 56% misinterpretation of GPT-4.
• AI achieves significant human-like conversational ability milestone.
Technology Innovation Institute Releases Falcon 2
• Small, powerful AI model optimized for cost-effectiveness and scalability.
• Surpasses Meta's Llama 3 on benchmarks.
• Offers multimodal capabilities and multilingual support.
• Plans to expand with the Mixture of Experts version.
Google Unveils LearnLM AI Model
• AI model developed in collaboration with DeepMind.
• Aim to tutor and enhance learning experiences.
• Integrated into Google products for lesson planning, problem-solving, and personalized study guidance.
• Limitations typical of generative AI models.
OpenAI's Preferred Publishers Program
• Offers partnership opportunities for news outlets.
• Financial incentives: guaranteed and variable payments based on user engagement.
• Aim to enhance user experience by prioritizing publishers' content.
Samsung Malaysia Unveils 2024 AI-Powered TVs
• Neo QLED 8K, 4K, OLED models.
• Advanced AI processors enhance picture quality and sound.
• AI customization and energy-saving modes.
• Rewards and prizes are available until June 30, 2024.
Google Introduces "Web" Filter
• Allows access to text-based links in search results.
• Acknowledges diverse web content preferences.
• Reflects shift in search dynamics towards varied content types and AI responses.
• May reshape SEO strategies and user search behaviors.