The Science
Our meta-research across disciplines: behavioral economics, cognitive science, complexity science, computer science, decision science, design, neuroscience, philosophy, and psychology. Science is changing because boundaries between disciplines are dissolving. Our research dissects the latest books and papers. Highly curated, an antidote to information overload.
The Hidden Cost of ChatGPT is the Erosion of the Digital Commons
A new study suggests that the rise of ChatGPT may be eroding the digital commons. If users turn more and more to ChatGPT and other AI models for answers and assistance, rather than posting their questions and solutions publicly, the digital commons that these models rely on will begin to decline.
Agentic AI and Self-Driven Exploration
New research on developing AI that builds robust world models shows an AI's ability to seek out surprise, motivation, and novelty, enabling it to navigate and understand the complexities of the world through self-driven exploration rather than just following predetermined reward pathways.
How AI Can Help Us Envision More Diverse Intelligence
AI will force us to broaden our view of intelligence. The real success in AI development will be in discovering forms of intelligence that go beyond anything we've known, transforming how we understand and interact with the world around us.
LLMs Can Set Prices and Collude Without People Knowing
Recent research demonstrates through empirical evidence that GPT-4 can autonomously develop pricing strategies that edge towards collusion without explicit human direction or inter-firm communication.
Why RAG Beats Fine-tuning AI
Enterprises face a critical choice in their generative AI adoption strategy: fine-tuning or Retrieval-Augmented Generation (RAG)? While fine-tuning has been the go-to approach for early adopters, a new study suggests that RAG may be the more powerful and sustainable path forward.
AIs That Understand Other AIs
By enabling different AI models to 'speak' to each other and combine their strengths, CALM opens up new possibilities for solving complex problems across various domains and tackling tasks with expertise and precision, in a data and compute efficient way.
Tools to Understand Appropriate Reliance on AI
By emphasizing critical engagement, transparency, bias mitigation, deliberate decision-making, user autonomy, and continuous education, Microsoft's research offers valuable guidelines for designing AI systems that promote appropriate reliance and user empowerment.
Worried That AI Might Make You Obsolete?
In 2016, AI experts predicted radiologists would be obsolete within years as machines outperform humans. This did not transpire.
Craft Better Prompts: Using AI to Improve Your Predictions
A review of research by Phil Tetlock and other experts on crafting better prompts by investigating if human forecasting can be improved through the use of a large language model.
How Network Theory Might Explain Emergent Abilities in AI
This research opens up vast possibilities for AI's role in solving complex problems but also underscores the importance of understanding and this emergent behavior especially as we head towards a world of multimodal models and agentic AI.
Meta-Prompting and What it Tells us About a Bias for Code
This research shows how flexible these models are: meta-prompting aids in decomposing complex tasks, engages distinct expertise, adopting a computational bias when using code in real-time which further enhances performance, then seamlessly integrates the varied outputs.
Gemini 1.5 Pro: An Ultra-Efficient, Multimodal System
The introduction of Gemini 1.5 Pro's ability to handle unprecedented context lengths, its superior performance compared to its predecessors, and the sustained relevance of power laws in its design underscore the breadth and depth of Google's long term capabilities.