A systematic approach to building an AI forecasting bot by classifying questions into four types to enhance prediction accuracy and improve forecasting methods.
About:
Jonathan Mann is a forecaster and cybersecurity architect with interests in AI safety and prediction markets.
Website:
Specializations:
Interests:
Outgoing Links:
Subscribe to RSS:
Decay functions in AI forecasting adjust probability estimates over time based on event types and their hazard patterns, enhancing prediction accuracy.
The post reveals that simpler polling methods outperform complex human-designed structures in AI forecasting when using frontier models.
The 'Broken Leg check' in AI forecasting helps identify decisive information that can override models, balancing efficiency with the risk of incorrect updates.
The blog post discusses the challenges of maintaining a reputation across different online platforms, highlighting how users start from scratch when switching services like Uber, Airbnb, and eBay. It introduces Cipherblock.io, a s...
Pre-mortems and sanity checks are essential techniques in AI forecasting to identify potential failures and ensure forecasts are coherent and reasonable.
The post highlights the shift in forecasting from craft to science through automated bots and empirical testing, promising a leap forward in the field.
A new forecasting bot mimics human forecasters and aims to refine its methodology through iterative testing in a competitive tournament setting.
The author creates an app to connect with like-minded individuals in NYC, emphasizing the importance of proximity for building friendships around shared interests in AI.
Housing reform is crucial but politically challenging; remote work offers immediate benefits that can support pro-family policies and build political loyalty.
The text discusses the importance of mobility in addressing housing affordability and improving quality of life in New York. It highlights the benefits of mobility improvements and proposes low-friction interventions to reshape ho...
The text discusses the importance of world models in addressing global existential threats and the need for systematic forecasting to test these models. It explains the challenges of evaluating world models and the instrumentalist...
Jonathan Mann provides quick updates on his upcoming articles, his invitation to LessOnline, and a public hangout in Brooklyn for a forecasting conference and retreat.
The text discusses when to use forecasting for decisions, emphasizing that it's not always the best approach. It provides a framework to determine when to fortify, follow, or forecast when faced with an important decision. It high...
The text is a reflection on the revised 2020 edition of Robert Putnam's Bowling Alone, discussing the decline of social capital in America and the factors contributing to it. It explores the concepts of bonding and bridging capita...
The text discusses the capabilities of current-generation AI and the importance of providing detailed, step-by-step instructions to improve the accuracy of AI-generated outputs. It highlights the power of prompt engineering and th...
The text discusses the evaluation of forecasting performance and the difficulty in distinguishing between skill and luck. It explains the use of statistical hypothesis testing to assess the likelihood that a forecaster's success i...
0Book Bets
2024-04-28 • constructive feedback intellectual growth book bets
The text discusses the concept of 'book bets' as an alternative to monetary bets in resolving disagreements. It explains how book bets work and provides an example of a book bet between two co-workers. The post highlights the bene...
Jonathan Mann shares his thoughts on the ACX Forecasting Contest, discussing the strategies employed by participants and the surprising outcomes. He reflects on the performance of Samotsvety in the contest and the implications of ...
0When AI Can Do All Jobs
2024-01-22 • technology humanities future ai
The text discusses the future of AI and its impact on human work. It explores the idea that humans will still have a role in guiding AI to serve collective human interests, even as AI surpasses human cognitive abilities.
The author discusses the potential risks and benefits of advanced AI and the importance of understanding and managing these risks. He explains why he signed the AI Treaty Open Letter and advocates for thoughtful and cautious use o...
The article discusses different methods of scoring in forecasting, particularly examining Brier and Log scoring systems to understand how they shape the forecasting landscape and the identification of top-tier forecasters. It expl...
The text discusses the ideal properties of a forecast scoring system, focusing on proper and improper scoring rules. It provides examples of how these scoring systems work in practice and how they incentivize forecasters to report...
The text introduces a series aimed at identifying and understanding top forecasters, examining the challenges, and improving the craft of forecasting. It discusses the importance of finding reliable information sources and the com...