Steven AI Talk

Barclays to roll out Microsoft 365 Copilot to 100,000 colleagues https://ukstories.microsoft.com/features/barclays-rolls-out-microsoft-365-copilot-to-100000-colleagues/

https://arxiv.org/html/2505.24832v2 How much do language models memorize?

How Anthropic teams use Claude Code

https://nlp.elvissaravia.com/p/top-ai-papers-of-the-week-02a?r=4kwjsr&utm_campaign=post&utm_medium=email&triedRedirect=true

OpenThoughts_ Crafting Superior Reasoning Models https://www.arxiv.org/pdf/2506.04178

Snowflake Summit 2025_ Data, AI, and the Future

AI Developments_ Models, Tools, and Disputes

Understanding Open Source AI_ Transparency, Freedom, and Impact

Sundar Pichai: AI, Humanity, and Progress

https://arxiv.org/html/2502.21321v2 The sources explore the application of reinforcement learning (RL) to refine Large Language Models (LLMs) , treating text generation as a sequence of decisions. They discuss how RL methods, particularly those using reward models trained on human preferences , enable LLMs to produce outputs that are not just statistically likely but also aligned with desired char...

https://arxiv.org/html/2505.12540v2 This document introduces vec2vec , the first method for translating text embeddings between different models without paired data or encoders . This approach leverages a universal latent representation of text semantics, supporting the Strong Platonic Representation Hypothesis . The authors demonstrate that vec2vec successfully translates embeddings while preserv...

https://arxiv.org/html/2505.13379v1 This source introduces Thinkless , a reinforcement learning framework designed to make large language models (LLMs) more efficient by enabling them to decide whether to use detailed, long-form reasoning or provide a brief, short-form answer . The core innovation is the Decoupled Group Relative Policy Optimization (DeGRPO) algorithm, which separates the learning...

These documents discuss how traditional banks can transform into  AI-first institutions  to remain competitive in the digital age. The transformation involves building a comprehensive  AI-and-analytics capability stack  with four interconnected layers: reimagined  customer engagement ,  AI-powered decision making , modern  core technology and data infrastructure , and a platform-based  operating m...

This  report from the U.S. Department of the Treasury  discusses the  widespread use of Artificial Intelligence (AI) in the financial sector , including both established machine learning techniques and newer Generative AI. It outlines the  various applications and potential benefits  of AI for financial firms and consumers, while also addressing the  significant risks  such as data privacy concern...

This paper discusses the increasing use of  artificial intelligence (AI)  in  financial services , particularly in emerging markets. AI is being adopted to  analyze diverse data sources , which helps in areas like assessing  creditworthiness  for those without traditional financial histories and improving  risk management  against threats like fraud. Furthermore, AI supports  process automation  t...

Prompting, Vibe Coding, and AI Agents

https://arxiv.org/html/2504.20879v1The Leaderboard Illusion Measuring progress is crucial for any scientific field, and in Artificial Intelligence, benchmarks play a vital role. Recently, Chatbot Arena has become the most influential leaderboard for ranking the capabilities of large language models (LLMs), significantly impacting media perception, industry trends, and academic research. However, a...

AI PodcastFin-R1: Financial Reasoning with a Lightweight Language Model https://huggingface.co/SUFE-AIFLM-Lab/Fin-R1

AI generated podcastMechanistic Interpretability via Cross-Layer Feature Attribution Graphsref: https://transformer-circuits.pub/2025/attribution-graphs/methods.html

AI generated podcastTracing the Thoughts of a Large Language ModelRef: https://www.anthropic.com/news/tracing-thoughts-language-model

Source:https://huggingface.co/microsoft/Magma-8Bhttps://github.com/microsoft/MagmaDigital and Physical Worlds: Magma is the first-ever foundation model for multimodal AI agents, designed to handle complex interactions across both virtual and real environments! Versatile Capabilities: Magma as a single model not only possesses generic image and videos understanding ability, but also generate goal-d...

Experimental Economics_ Core Concepts and Applications

The Psychology and Economics of Incentives

Market Experiments, Asymmetric Information, and Auction Theory

Behavioral Economics_ Trust, Motivation, Willpower, and Gender