The Interpretable AI playbook: What Anthropic’s research means for your enterprise LLM strategy https://venturebeat.com/ai/the-interpretable-ai-playbook-what-anthropics-research-means-for-your-enterprise-llm-strategy/ #AI #interpretability
Administered by:
#interpretability
0 posts0 participants0 posts today
Circuit tracing for AI interpretability:
#ai #llm #interpretability #research #innovation
https://www.anthropic.com/research/open-source-circuit-tracing
www.anthropic.comOpen-sourcing circuit-tracing toolsAnthropic is an AI safety and research company that's working to build reliable, interpretable, and steerable AI systems.
Are LM more than their behavior? 
Join our Conference on Language Modeling (COLM) workshop and explore the interplay between what LMs answer and what happens internally 
See you in Montréal 
CfP: shorturl.at/sBomu
Page: shorturl.at/FT3fX
Reviewer Nomination: shorturl.at/Jg1BP
Unlock the Secrets of AI Learning! ????Ever wondered how generative AI, the powerhouse behind stunning images and sophisticated text, truly learns? Park et al.'s groundbreaking study, ‘Emergence of Hidden Capabilities: Exploring Learning Dynamics in Concept Space,’ offers a revolutionary new perspective. Forget black boxes – this research unveils a "concept space" where AI learning becomes a visible journey!By casting ideas into geometric space, the authors bring to life how AI models learn step by step, stripping bare the order and timing of their knowledge. See the crucial role played by the "concept signal" in predicting what a model is first going to learn and note the fascinating "trajectory turns" revealing the sudden "aha!" moments of emergent abilities.This is not a theoretical abstraction – the framework has deep implications in the real world:Supercharge AI Training: Optimise training data to speed learning and improve efficiency.Demystify New Behaviours: Understand and even manage unforeseen strengths of state-of-the-art AI.Debug at Scale: Gain unprecedented insights into the knowledge state of a model to identify and fix faults.Future-Proof AI: This mode-agnostic feature primes the understanding of learning in other AI systems.This study is a must-read for all who care about the future of AI, from scientists and engineers to tech geeks and business executives. It's not only what AI can accomplish, but how it comes to do so.Interested in immersing yourself in the captivating universe of AI learning?Click here to read the complete article and discover the secrets of the concept space! #AI #MachineLearning #GenerativeAI #DeepLearning #Research #Innovation #ConceptSpace #EmergentCapabilities #AIDevelopment #Tech #ArtificialIntelligence #DataScience #FutureofAI #Interpretability
Replied in thread
Continued thread
"Feature importance helps in understanding which features contribute most to the prediction"
A few lines with #sklearn: https://mljourney.com/sklearn-linear-regression-feature-importance/
ML Journey · Sklearn Linear Regression Feature Importance - ML JourneyDiscover how to determine feature importance in linear regression models using Scikit-learn. This comprehensive guide covers methods like
Replied in thread
#Lasso #LinearRegression "is useful in some contexts due to its tendency to prefer solutions with fewer non-zero coefficients, effectively reducing the number of features upon which the given solution is dependent"
https://scikit-learn.org/stable/modules/linear_model.html#lasso 
scikit-learn1.1. Linear ModelsThe following are a set of methods intended for regression in which the target value is expected to be a linear combination of the features. In mathematical notation, if\hat{y} is the predicted val...
Replied in thread
@data "practitioners can leverage #LASSO regression to construct more interpretable and predictive models that excel in scenarios involving high-dimensional data and intricate feature relationships."
https://datasciencedecoded.com/posts/12_LASSO_Regression_Feature_Selection_Predictive_Models
Continued thread
"The following sections discuss several state-of-the-art interpretable and explainable #ML methods. The selection of works does not comprise an exhaustive survey of the literature. Instead, it is meant to illustrate the commonest properties and inductive biases behind interpretable models and [black-box] explanation methods using concrete instances."
https://wires.onlinelibrary.wiley.com/doi/full/10.1002/widm.1493#widm1493-sec-0010-title
Model "#interpretability and [black-box] #explainability, although not necessary in many straightforward applications, become instrumental when the problem definition is incomplete and in the presence of additional desiderata, such as trust, causality, or fairness."
https://wires.onlinelibrary.wiley.com/doi/full/10.1002/widm.1493
When using #machinelearning for tasks in #geosciences, you should aim for #interpretability! Why this is the case and how to go about it is the topic of a brand-new article in the open access journal "Earth's Future" by Shijie Jiang and an interdisciplinary group of colleagues. Check it out!
Research in mechanistic interpretability and neuroscience often relies on interpreting internal representations to understand systems, or manipulating representations to improve models. I gave a talk at the UniReps workshop at NeurIPS on a few challenges for this area, summary thread: 1/12
#ai #ml #neuroscience #computationalneuroscience #interpretability #NeuralRepresentations #neurips2023
The Illusion of Understanding: MIT Unmasks the Myth of AI’s Formal Specifications
A study by MIT Lincoln Laboratory suggests that formal specifications, despite their mathematical precision, are not necessarily interpretable to humans. Participants struggled to validate AI behaviors using these specifications, indicating a discrepancy between theoretical claims and practical understanding. The findings highlight the need for more realistic assessments of AI interpretability.
https://scitechdaily.com/the-illusion-of-understanding-mit-unmasks-the-myth-of-ais-formal-specifications/ #AI #FormalSpecifications #interpretability #behavior
SciTechDaily · The Illusion of Understanding: MIT Unmasks the Myth of AI’s Formal SpecificationsSome researchers see formal specifications as a way for autonomous systems to "explain themselves" to humans. But a new study finds that we aren't understanding. As autonomous systems and artificial intelligence become increasingly common in daily life, new methods are emerging to help humans che
Anyone work at a company they like that's hiring #PhD #research interns for summer 2024? Interested in roles related to #ComputerScience #DataVisualization #ExplainableAI #Genomics #Bioinformatics #SoftwareEngineering #ML #interpretability #EDA #DataAnalytics
I'm a third year PhD student in Computer Science at an R1 university (completed MS coursework), and a dual citizen USA/EU :) Thanks for boosting! #GetFediHired
#XAI - This book is a great resource on #explainability / #interpretability methods for #AI and #ML:
"Machine learning has great potential for improving products, processes and research. But computers usually do not explain their predictions which is a barrier to the adoption of machine learning. This book is about making machine learning models and their decisions interpretable.
After exploring the concepts of interpretability, you will learn about simple, interpretable models such as decision trees, decision rules and linear regression. The focus of the book is on model-agnostic methods for interpreting black box models such as feature importance and accumulated local effects, and explaining individual predictions with Shapley values and LIME. In addition, the book presents methods specific to deep neural networks.
All interpretation methods are explained in depth and discussed critically. How do they work under the hood? What are their strengths and weaknesses? How can their outputs be interpreted? This book will enable you to select and correctly apply the interpretation method that is most suitable for your machine learning project. Reading the book is recommended for machine learning practitioners, data scientists, statisticians, and anyone else interested in making machine learning models interpretable.
christophm.github.ioInterpretable Machine LearningMachine learning algorithms usually operate as black boxes and it is unclear how they derived a certain decision. This book is a guide for practitioners to make machine learning decisions interpretable.
New #preprint: In this review manuscript we discuss established and emerging approaches to report and further analyze #subthreshold effects in #fMRI studies with the aim to improve #interpretability and #comparability in #neuroimaging https://psyarxiv.com/fyhst/
#Computerphile - #Glitch #Tokens In #LargeLanguageModels
#RobMiles talks about '#GlitchTokens', those mysterious words, which result in gibberish when entered into some large #LanguageModels.
https://www.youtube.com/watch?v=WO2X3oZEJOA&ab_channel=Computerphile
Continued thread
The TAYSIR competition about extracting small, interpretable models from neural language models will also be hosted at ICGI!
https://remieyraud.github.io/TAYSIR/
The first CFP is available: https://remieyraud.github.io/TAYSIR/TAYSIR-Call_for_participation_1.pdf
remieyraud.github.ioTAYSIR: Transformers+RNN: Algorithms to Yield Simple and Interpretable Representations
Continued thread
There are a four invited speakers, but I am only personally familiar with three (Cyril Allauzen @ Google, Will Merrill @ NYU/Google, and Dana Angluin @ Yale).
These three talks should be fantastic, especially if you are interested in #automata, #FormalLanguages, and #Interpretability in neural language models!
(Plugging https://flann.super.site/ if those sound cool to you)
FLaNN SeminarsFLaNN SeminarsWe organize a series of weekly online seminars on Formal Language Theory, Natural Language Processing, Machine Learning and Computational Linguistics in an informal setting.