Isaac Kohane’s Post

Introducing Genex: Generative World Explorer ✨ 🌎 🧠 Humans mentally explore unseen parts of the world, revising their beliefs with imagined observations. To safely explore the world, we can first imagine what could be out there before we physically venture out. ✨ Genex replicates this human-like ability, pushing the boundaries of embodied AI. our paper has been featured as the #1 entry in the Hugging Face Daily Papers: https://lnkd.in/e55Ad-XF our website is live (for high-resolution video): https://lnkd.in/e9unKybR ; arxiv: https://lnkd.in/exR5QpR9 ; see the X threads for more informed introduction: https://lnkd.in/eMHGeUqr 🎉 many thanks to the amazing collaborators Taiming Lu Tianmin Shu Daniel Khashabi & Alan Yuille

"QwQ: Reflect Deeply on the Boundaries of the Unknown" (https://shorturl.at/U3dJn) JK--I'd like to ask the authors of this study to turn down the "pretentious" knob on their otherwise interesting research project into advanced AI reasoning.

chatomics! The best way to learn is to work through a toy dataset! Generative AI: Text generation using Long short-term memory (LSTM) model https://lnkd.in/eFq28xNb

What Is Retrieval-Augmented Generation, aka RAG? Retrieval-augmented generation (RAG) is a technique for enhancing the accuracy and reliability of generative AI models with facts fetched from external sources. Well articulated Rick Merritt https://lnkd.in/dZ_PhPcF

Excited to share our latest research study on the metacognitive capabilities of Generative AI compared to human metacognition. Our study compared the performance of five advanced LLMs with human participants on an ICF-mimicking exam. The results reveal that LLMs outperformed humans across all metacognitive metrics, particularly in reducing overconfidence. However, both LLMs and humans exhibited less adaptability in ambiguous scenarios. This study sheds light on the potential for Generative AI to engage in human-like metacognitive processing, without conscious awareness. Read the full study here: https://bit.ly/3WBt1WK.

This is an important TED talk on AI.

📰 Tech News Snippet #1524 A groundbreaking study from King's Business School and Wazoku is shaking up traditional problem-solving paradigms through AI. This report highlights that 46% of Wazoku's vast network, consisting of 700,000 diverse problem solvers, have incorporated generative AI

In this month’s update on the Information Domain, we reflect on the key trends observed in 2024 and what to expect in 2025. Read more about this in our January 2025 report on how the maturation of Generative AI technologies has made it more difficult for both humans and automated systems to effectively identify AI-driven misinformation and disinformation.

I am excited to share our latest blog post focusing on LLaVA-O1, a groundbreaking approach that enhances the reasoning capabilities of vision-language models through a step-by-step reasoning process. This innovative framework not only advances the efficiency of model interactions but also significantly improves the understanding and interpretation of visual information in conjunction with language. Dive into the details and explore the potential implications for various applications in AI. You can read the full article here: https://ift.tt/OUJciAM.

Today! DIJ researcher Carolin Fleischer-Heininger https://lnkd.in/g4g7Bwpg will present "Desirable AI for all: Approaching future technologies through a disability studies lens" in the Leverhulme Centre for the Future of Intelligence Cross-Cultural Approaches to Desirable AI seminar series https://lnkd.in/gcXcb-J3