AI's Tightrope Walk: Safeguarding Democracy with New Treaties and Mitigating Educational Disparities
Niel de Kock
Editor of 'The AI Way' a weekly email newsletter focussed on Education and AI. | Pioneering AI in Education & Self-Learning | Explore AI's Frontier with My Weekly Newsletter |1010+ Subscribers & Growing
Education AI News
Lionsgate Partners With AI Video Startup Runway for New Creative Endeavors
Lionsgate, the studio behind John Wick and The Hunger Games, partnered with AI video startup Runway to create a custom AI model to assist the studio's creative teams in content production. The partnership aims to "augment" the work of producers and directors by offering innovative, cost-efficient ways to create content. However, details on how the AI model will function remain unclear, especially considering the AI regulations agreed upon after last year's SAG-AFTRA and WGA strikes. Lionsgate's partnership with Runway comes amid a challenging year for the studio following several box-office disappointments. Runway, like other AI companies, is also facing legal challenges from visual artists over copyright infringement.
UN Advisory Body Issues Seven Key Recommendations for Global AI Governance
The UN advisory body on AI governance released its final report with seven key recommendations to address AI-related risks and governance gaps. These proposals aim to ensure responsible development and use of artificial intelligence globally, considering the rapid growth and potential dangers of AI technology. The recommendations include:
These recommendations will be discussed further at the upcoming UN summit in September.
Stanford Researchers Warn AI Could Exacerbate Racial Disparities in Education
Researchers from Stanford University, working with the United Nations, have raised concerns about how artificial intelligence (AI) could worsen existing disparities in education. While AI offers personalized learning and predictive analytics, it risks reinforcing racial biases due to biased historical data. Their report highlights that AI tools designed to predict student success may disadvantage racial minorities.
The researchers recommend involving teachers, students, and marginalized communities in AI development, promoting responsible AI use through public education programs, and ensuring open access to AI tools. Their findings aim to guide future efforts to ensure AI mitigates, rather than exacerbates, racial inequalities in education.
Educators More Comfortable with AI than Students, Study Reveals
A recent survey from the 2024 AI in Academia Study by Copyleaks highlights a growing divide in AI adoption between educators and students. The study surveyed 1,000 students and 250 educators in the U.S. and found that 34% of educators frequently use AI for tasks like drafting or reviewing assignments, compared to only 24% of students. Despite enthusiasm for AI integration—70% of educators versus 58% of students—there remains a significant gap in trust and ethical use of AI. While both groups acknowledge the potential of AI for personalized learning, educators tend to be more comfortable with AI tools. However, the study points to an increasing need for clear guidelines and educational initiatives to foster responsible and effective AI use in the classroom.
AI News
YouTube Introduces New Generative AI Features for Video Creation, Music, and Content Inspiration
YouTube has announced a set of new AI-driven tools to enhance content creation. CEO Neal Mohan revealed features like six-second AI-generated video clips for YouTube Shorts, AI-enhanced responses to comments, and music creation tools like Dream Track. These new capabilities, powered by Google’s DeepMind, will help creators generate video ideas, improve video accessibility with auto-dubbing, and boost engagement with features like the "Hype" button. To ensure transparency, YouTube will watermark all AI-generated content.
(https://www.nbcnews.com/tech/tech-news/youtube-ai-generative-features-date-announcement-rcna171661)
领英推荐
AI Models in Home Surveillance Show Inconsistent and Biased Outcomes, Study Finds
A study by MIT and Penn State researchers reveals that large language models (LLMs) used in home surveillance could inconsistently recommend calling the police, even when no crime is occurring. The study found that these AI models often flagged videos differently, with biases emerging based on neighborhood demographics, particularly in predominantly white neighborhoods, without considering the skin tone of individuals in the videos. This inconsistency, termed "norm inconsistency," raises concerns about deploying AI in high-stakes environments, with researchers calling for more transparency and careful consideration before using generative AI models for tasks like home surveillance.
Enhancing LLM Collaboration for Smarter, More Efficient Solutions
MIT CSAIL researchers have developed "Co-LLM," an algorithm that improves collaboration between a general-purpose large language model (LLM) and an expert LLM, resulting in more accurate and efficient responses. Co-LLM enables the base model to call on the expert model only when necessary, using a "switch variable" to decide which parts of the response require specialized knowledge. The system boosts performance in areas like medical and mathematical queries by combining general models' flexibility with expert ones' precision. Co-LLM's collaboration strategy mimics human teamwork, allowing LLMs to deliver more factual and reliable answers. This innovative approach can outperform independent LLMs, enhancing accuracy and efficiency in high-stakes applications.
AI Tools
Homeworkify: A Comprehensive Online Tutoring Platform
Homeworkify is an online tutoring platform developed by a Berlin-based web developer to assist students with homework across various subjects. It offers personalized tutoring, study guides, textbook solutions, and 24/7 accessibility to help students understand and complete their assignments. AI integration ensures accurate and tailored answers. Users can submit questions, track progress, and choose from three pricing plans: a free Basic Plan, Plus Plan for $9.99/month, or Premium Plan for $19.99/month. Homeworkify aims to make learning more accessible and enjoyable while maintaining privacy and security.
CoderKit: AI-Powered Coding Assistant for Xcode
CoderKit is a free Xcode extension that enhances developer productivity through AI-powered code autocompletion. It integrates with Codeium or GitHub Copilot to provide real-time suggestions as developers type, which can be accepted or rejected seamlessly within Xcode. The extension syncs with Xcode’s color theme for a smooth visual experience. While CoderKit is free, users need a Codeium or GitHub Copilot subscription to access its AI autocompletion features. Although not open-source, the developers are inspired by the open-source community and aim to contribute in the future.
How might the introduction of YouTube's new generative AI tools, such as AI-generated video clips and auto-dubbing, inadvertently contribute to misinformation, and what safeguards could be implemented to minimize these risks?
Enjoy your week till next week.