AI Snake Oil
What Artificial Intelligence Can Do, What It Can’t, and How to Tell the Difference
RSS feed: https://www.aisnakeoil.com/feed
Recent posts
Last post published on Fri Jul 26 2024
The text discusses the limitations of using probability estimates to assess the risk of AI and suggests alternative approaches to AI policy-making.
The development and evaluation of AI agents, highlighting challenges and proposing recommendations for improvement.
The article discusses the limitations of improving language models through scaling and the need for a more nuanced approach to AI progress.
The text discusses the limitations and pitfalls of using AI in scientific research, highlighting the need for transparency and reproducibility.
The text discusses the limitations of traditional AI agent evaluation methods and proposes a new approach using Pareto curves to consider both accuracy and cost.
Preorder available for AI Snake Oil book, focusing on foundational AI knowledge, separating hype from reality, and addressing concerns about AI and society.
Reasons for cynicism about tech policy and cautious optimism, with examples of recent work on AI policy
AI safety measures, limitations of securing AI models, and the importance of external defenses against misuse
Open letter for legal, technical protections in AI models research
A paper's contributions to the debate on societal impact of open foundation models, proposing a risk assessment framework.
Explores the potential impact of AI on the legal profession, emphasizing socio-technical assessments.
Discusses the copyright infringement lawsuit by the New York Times against OpenAI and potential methods to avoid copyright violations in generative AI.
The text discusses the risks and benefits of open foundation models (FMs) and the implications for policymaking and public consultation.
The limitations of Reinforcement Learning with Human Feedback (RLHF) as a method for aligning language models with human values and the need for alternative approaches to prevent harm from AI.