Episoder

  • Lots of people in the AI safety space worry about models being able to make deliberate, multi-step plans. But can we already see this in existing neural nets? In this episode, I talk with Erik Jenner about his work looking at internal look-ahead within chess-playing neural networks.

    Patreon: https://www.patreon.com/axrpodcast

    Ko-fi: https://ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/12/12/episode-38_3-erik-jenner-learned-look-ahead.html

    FAR.AI: https://far.ai/

    FAR.AI on X (aka Twitter): https://x.com/farairesearch

    FAR.AI on YouTube: https://www.youtube.com/@FARAIResearch

    The Alignment Workshop: https://www.alignment-workshop.com/

    Topics we discuss, and timestamps:

    00:57 - How chess neural nets look into the future

    04:29 - The dataset and basic methodology

    05:23 - Testing for branching futures?

    07:57 - Which experiments demonstrate what

    10:43 - How the ablation experiments work

    12:38 - Effect sizes

    15:23 - X-risk relevance

    18:08 - Follow-up work

    21:29 - How much planning does the network do?

    Research we mention:

    Evidence of Learned Look-Ahead in a Chess-Playing Neural Network: https://arxiv.org/abs/2406.00877

    Understanding the learned look-ahead behavior of chess neural networks (a development of the follow-up research Erik mentioned): https://openreview.net/forum?id=Tl8EzmgsEp

    Linear Latent World Models in Simple Transformers: A Case Study on Othello-GPT: https://arxiv.org/abs/2310.07582

    Episode art by Hamish Doodles: hamishdoodles.com

  • The 'model organisms of misalignment' line of research creates AI models that exhibit various types of misalignment, and studies them to try to understand how the misalignment occurs and whether it can be somehow removed. In this episode, Evan Hubinger talks about two papers he's worked on at Anthropic under this agenda: "Sleeper Agents" and "Sycophancy to Subterfuge".

    Patreon: https://www.patreon.com/axrpodcast

    Ko-fi: https://ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/12/01/episode-39-evan-hubinger-model-organisms-misalignment.html

    Topics we discuss, and timestamps:

    0:00:36 - Model organisms and stress-testing

    0:07:38 - Sleeper Agents

    0:22:32 - Do 'sleeper agents' properly model deceptive alignment?

    0:38:32 - Surprising results in "Sleeper Agents"

    0:57:25 - Sycophancy to Subterfuge

    1:09:21 - How models generalize from sycophancy to subterfuge

    1:16:37 - Is the reward editing task valid?

    1:21:46 - Training away sycophancy and subterfuge

    1:29:22 - Model organisms, AI control, and evaluations

    1:33:45 - Other model organisms research

    1:35:27 - Alignment stress-testing at Anthropic

    1:43:32 - Following Evan's work

    Main papers:

    Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training: https://arxiv.org/abs/2401.05566

    Sycophancy to Subterfuge: Investigating Reward-Tampering in Large Language Models: https://arxiv.org/abs/2406.10162

    Anthropic links:

    Anthropic's newsroom: https://www.anthropic.com/news

    Careers at Anthropic: https://www.anthropic.com/careers

    Other links:

    Model Organisms of Misalignment: The Case for a New Pillar of Alignment Research: https://www.alignmentforum.org/posts/ChDH335ckdvpxXaXX/model-organisms-of-misalignment-the-case-for-a-new-pillar-of-1

    Simple probes can catch sleeper agents: https://www.anthropic.com/research/probes-catch-sleeper-agents

    Studying Large Language Model Generalization with Influence Functions: https://arxiv.org/abs/2308.03296

    Stress-Testing Capability Elicitation With Password-Locked Models [aka model organisms of sandbagging]: https://arxiv.org/abs/2405.19550

    Episode art by Hamish Doodles: hamishdoodles.com

  • Mangler du episoder?

    Klikk her for å oppdatere manuelt.

  • You may have heard of singular learning theory, and its "local learning coefficient", or LLC - but have you heard of the refined LLC? In this episode, I chat with Jesse Hoogland about his work on SLT, and using the refined LLC to find a new circuit in language models.

    Patreon: https://www.patreon.com/axrpodcast

    Ko-fi: https://ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/11/27/38_2-jesse-hoogland-singular-learning-theory.html

    FAR.AI: https://far.ai/

    FAR.AI on X (aka Twitter): https://x.com/farairesearch

    FAR.AI on YouTube: https://www.youtube.com/@FARAIResearch

    The Alignment Workshop: https://www.alignment-workshop.com/

    Topics we discuss, and timestamps:

    00:34 - About Jesse

    01:49 - The Alignment Workshop

    02:31 - About Timaeus

    05:25 - SLT that isn't developmental interpretability

    10:41 - The refined local learning coefficient

    14:06 - Finding the multigram circuit

    Links:

    Differentiation and Specialization of Attention Heads via the Refined Local Learning Coefficient: https://arxiv.org/abs/2410.02984

    Investigating the learning coefficient of modular addition: hackathon project: https://www.lesswrong.com/posts/4v3hMuKfsGatLXPgt/investigating-the-learning-coefficient-of-modular-addition

    Episode art by Hamish Doodles: hamishdoodles.com

  • Road lines, street lights, and licence plates are examples of infrastructure used to ensure that roads operate smoothly. In this episode, Alan Chan talks about using similar interventions to help avoid bad outcomes from the deployment of AI agents.

    Patreon: https://www.patreon.com/axrpodcast

    Ko-fi: https://ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/11/16/episode-38_1-alan-chan-agent-infrastructure.html

    FAR.AI: https://far.ai/

    FAR.AI on X (aka Twitter): https://x.com/farairesearch

    FAR.AI on YouTube: https://www.youtube.com/@FARAIResearch

    The Alignment Workshop: https://www.alignment-workshop.com/

    Topics we discuss, and timestamps:

    01:02 - How the Alignment Workshop is

    01:32 - Agent infrastructure

    04:57 - Why agent infrastructure

    07:54 - A trichotomy of agent infrastructure

    13:59 - Agent IDs

    18:17 - Agent channels

    20:29 - Relation to AI control

    Links:

    Alan on Google Scholar: https://scholar.google.com/citations?user=lmQmYPgAAAAJ&hl=en&oi=ao

    IDs for AI Systems: https://arxiv.org/abs/2406.12137

    Visibility into AI Agents: https://arxiv.org/abs/2401.13138

    Episode art by Hamish Doodles: hamishdoodles.com

  • Do language models understand the causal structure of the world, or do they merely note correlations? And what happens when you build a big AI society out of them? In this brief episode, recorded at the Bay Area Alignment Workshop, I chat with Zhijing Jin about her research on these questions.

    Patreon: https://www.patreon.com/axrpodcast

    Ko-fi: https://ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/11/14/episode-38_0-zhijing-jin-llms-causality-multi-agent-systems.html

    FAR.AI: https://far.ai/

    FAR.AI on X (aka Twitter): https://x.com/farairesearch

    FAR.AI on YouTube: https://www.youtube.com/@FARAIResearch

    The Alignment Workshop: https://www.alignment-workshop.com/

    Topics we discuss, and timestamps:

    00:35 - How the Alignment Workshop is

    00:47 - How Zhijing got interested in causality and natural language processing

    03:14 - Causality and alignment

    06:21 - Causality without randomness

    10:07 - Causal abstraction

    11:42 - Why LLM causal reasoning?

    13:20 - Understanding LLM causal reasoning

    16:33 - Multi-agent systems

    Links:

    Zhijing's website: https://zhijing-jin.com/fantasy/

    Zhijing on X (aka Twitter): https://x.com/zhijingjin

    Can Large Language Models Infer Causation from Correlation?: https://arxiv.org/abs/2306.05836

    Cooperate or Collapse: Emergence of Sustainable Cooperation in a Society of LLM Agents: https://arxiv.org/abs/2404.16698

    Episode art by Hamish Doodles: hamishdoodles.com

  • Epoch AI is the premier organization that tracks the trajectory of AI - how much compute is used, the role of algorithmic improvements, the growth in data used, and when the above trends might hit an end. In this episode, I speak with the director of Epoch AI, Jaime Sevilla, about how compute, data, and algorithmic improvements are impacting AI, and whether continuing to scale can get us AGI.

    Patreon: https://www.patreon.com/axrpodcast

    Ko-fi: https://ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/10/04/episode-37-jaime-sevilla-forecasting-ai.html

    Topics we discuss, and timestamps:

    0:00:38 - The pace of AI progress

    0:07:49 - How Epoch AI tracks AI compute

    0:11:44 - Why does AI compute grow so smoothly?

    0:21:46 - When will we run out of computers?

    0:38:56 - Algorithmic improvement

    0:44:21 - Algorithmic improvement and scaling laws

    0:56:56 - Training data

    1:04:56 - Can scaling produce AGI?

    1:16:55 - When will AGI arrive?

    1:21:20 - Epoch AI

    1:27:06 - Open questions in AI forecasting

    1:35:21 - Epoch AI and x-risk

    1:41:34 - Following Epoch AI's research

    Links for Jaime and Epoch AI:

    Epoch AI: https://epochai.org/

    Machine Learning Trends dashboard: https://epochai.org/trends

    Epoch AI on X / Twitter: https://x.com/EpochAIResearch

    Jaime on X / Twitter: https://x.com/Jsevillamol

    Research we discuss:

    Training Compute of Frontier AI Models Grows by 4-5x per Year: https://epochai.org/blog/training-compute-of-frontier-ai-models-grows-by-4-5x-per-year

    Optimally Allocating Compute Between Inference and Training: https://epochai.org/blog/optimally-allocating-compute-between-inference-and-training

    Algorithmic Progress in Language Models [blog post]: https://epochai.org/blog/algorithmic-progress-in-language-models

    Algorithmic progress in language models [paper]: https://arxiv.org/abs/2403.05812

    Training Compute-Optimal Large Language Models [aka the Chinchilla scaling law paper]: https://arxiv.org/abs/2203.15556

    Will We Run Out of Data? Limits of LLM Scaling Based on Human-Generated Data [blog post]: https://epochai.org/blog/will-we-run-out-of-data-limits-of-llm-scaling-based-on-human-generated-data

    Will we run out of data? Limits of LLM scaling based on human-generated data [paper]: https://arxiv.org/abs/2211.04325

    The Direct Approach: https://epochai.org/blog/the-direct-approach

    Episode art by Hamish Doodles: hamishdoodles.com

  • Sometimes, people talk about transformers as having "world models" as a result of being trained to predict text data on the internet. But what does this even mean? In this episode, I talk with Adam Shai and Paul Riechers about their work applying computational mechanics, a sub-field of physics studying how to predict random processes, to neural networks.

    Patreon: https://www.patreon.com/axrpodcast

    Ko-fi: https://ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/09/29/episode-36-adam-shai-paul-riechers-computational-mechanics.html

    Topics we discuss, and timestamps:

    0:00:42 - What computational mechanics is

    0:29:49 - Computational mechanics vs other approaches

    0:36:16 - What world models are

    0:48:41 - Fractals

    0:57:43 - How the fractals are formed

    1:09:55 - Scaling computational mechanics for transformers

    1:21:52 - How Adam and Paul found computational mechanics

    1:36:16 - Computational mechanics for AI safety

    1:46:05 - Following Adam and Paul's research

    Simplex AI Safety: https://www.simplexaisafety.com/

    Research we discuss:

    Transformers represent belief state geometry in their residual stream: https://arxiv.org/abs/2405.15943

    Transformers represent belief state geometry in their residual stream [LessWrong post]: https://www.lesswrong.com/posts/gTZ2SxesbHckJ3CkF/transformers-represent-belief-state-geometry-in-their

    Why Would Belief-States Have A Fractal Structure, And Why Would That Matter For Interpretability? An Explainer: https://www.lesswrong.com/posts/mBw7nc4ipdyeeEpWs/why-would-belief-states-have-a-fractal-structure-and-why

    Episode art by Hamish Doodles: hamishdoodles.com

  • How do we figure out what large language models believe? In fact, do they even have beliefs? Do those beliefs have locations, and if so, can we edit those locations to change the beliefs? Also, how are we going to get AI to perform tasks so hard that we can't figure out if they succeeded at them? In this episode, I chat with Peter Hase about his research into these questions.

    Patreon: https://www.patreon.com/axrpodcast

    Ko-fi: https://ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/08/24/episode-35-peter-hase-llm-beliefs-easy-to-hard-generalization.html

    Topics we discuss, and timestamps:

    0:00:36 - NLP and interpretability

    0:10:20 - Interpretability lessons

    0:32:22 - Belief interpretability

    1:00:12 - Localizing and editing models' beliefs

    1:19:18 - Beliefs beyond language models

    1:27:21 - Easy-to-hard generalization

    1:47:16 - What do easy-to-hard results tell us?

    1:57:33 - Easy-to-hard vs weak-to-strong

    2:03:50 - Different notions of hardness

    2:13:01 - Easy-to-hard vs weak-to-strong, round 2

    2:15:39 - Following Peter's work

    Peter on Twitter: https://x.com/peterbhase

    Peter's papers:

    Foundational Challenges in Assuring Alignment and Safety of Large Language Models: https://arxiv.org/abs/2404.09932

    Do Language Models Have Beliefs? Methods for Detecting, Updating, and Visualizing Model Beliefs: https://arxiv.org/abs/2111.13654

    Does Localization Inform Editing? Surprising Differences in Causality-Based Localization vs. Knowledge Editing in Language Models: https://arxiv.org/abs/2301.04213

    Are Language Models Rational? The Case of Coherence Norms and Belief Revision: https://arxiv.org/abs/2406.03442

    The Unreasonable Effectiveness of Easy Training Data for Hard Tasks: https://arxiv.org/abs/2401.06751

    Other links:

    Toy Models of Superposition: https://transformer-circuits.pub/2022/toy_model/index.html

    Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV): https://arxiv.org/abs/1711.11279

    Locating and Editing Factual Associations in GPT (aka the ROME paper): https://arxiv.org/abs/2202.05262

    Of nonlinearity and commutativity in BERT: https://arxiv.org/abs/2101.04547

    Inference-Time Intervention: Eliciting Truthful Answers from a Language Model: https://arxiv.org/abs/2306.03341

    Editing a classifier by rewriting its prediction rules: https://arxiv.org/abs/2112.01008

    Discovering Latent Knowledge Without Supervision (aka the Collin Burns CCS paper): https://arxiv.org/abs/2212.03827

    Weak-to-Strong Generalization: Eliciting Strong Capabilities With Weak Supervision: https://arxiv.org/abs/2312.09390

    Concrete problems in AI safety: https://arxiv.org/abs/1606.06565

    Rissanen Data Analysis: Examining Dataset Characteristics via Description Length: https://arxiv.org/abs/2103.03872

    Episode art by Hamish Doodles: hamishdoodles.com

  • How can we figure out if AIs are capable enough to pose a threat to humans? When should we make a big effort to mitigate risks of catastrophic AI misbehaviour? In this episode, I chat with Beth Barnes, founder of and head of research at METR, about these questions and more.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/07/28/episode-34-ai-evaluations-beth-barnes.html

    Topics we discuss, and timestamps:

    0:00:37 - What is METR?

    0:02:44 - What is an "eval"?

    0:14:42 - How good are evals?

    0:37:25 - Are models showing their full capabilities?

    0:53:25 - Evaluating alignment

    1:01:38 - Existential safety methodology

    1:12:13 - Threat models and capability buffers

    1:38:25 - METR's policy work

    1:48:19 - METR's relationships with labs

    2:04:12 - Related research

    2:10:02 - Roles at METR, and following METR's work

    Links for METR:

    METR: https://metr.org

    METR Task Development Guide - Bounty: https://taskdev.metr.org/bounty/

    METR - Hiring: https://metr.org/hiring

    Autonomy evaluation resources: https://metr.org/blog/2024-03-13-autonomy-evaluation-resources/

    Other links:

    Update on ARC's recent eval efforts (contains GPT-4 taskrabbit captcha story) https://metr.org/blog/2023-03-18-update-on-recent-evals/

    Password-locked models: a stress case for capabilities evaluation: https://www.alignmentforum.org/posts/rZs6ddqNnW8LXuJqA/password-locked-models-a-stress-case-for-capabilities

    Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training: https://arxiv.org/abs/2401.05566

    Untrusted smart models and trusted dumb models: https://www.alignmentforum.org/posts/LhxHcASQwpNa3mRNk/untrusted-smart-models-and-trusted-dumb-models

    AI companies aren't really using external evaluators: https://www.lesswrong.com/posts/WjtnvndbsHxCnFNyc/ai-companies-aren-t-really-using-external-evaluators

    Nobody Knows How to Safety-Test AI (Time): https://time.com/6958868/artificial-intelligence-safety-evaluations-risks/

    ChatGPT can talk, but OpenAI employees sure can’t: https://www.vox.com/future-perfect/2024/5/17/24158478/openai-departures-sam-altman-employees-chatgpt-release

    Leaked OpenAI documents reveal aggressive tactics toward former employees: https://www.vox.com/future-perfect/351132/openai-vested-equity-nda-sam-altman-documents-employees

    Beth on her non-disparagement agreement with OpenAI: https://www.lesswrong.com/posts/yRWv5kkDD4YhzwRLq/non-disparagement-canaries-for-openai?commentId=MrJF3tWiKYMtJepgX

    Sam Altman's statement on OpenAI equity: https://x.com/sama/status/1791936857594581428

    Episode art by Hamish Doodles: hamishdoodles.com

  • Reinforcement Learning from Human Feedback, or RLHF, is one of the main ways that makers of large language models make them 'aligned'. But people have long noted that there are difficulties with this approach when the models are smarter than the humans providing feedback. In this episode, I talk with Scott Emmons about his work categorizing the problems that can show up in this setting.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    The transcript: https://axrp.net/episode/2024/06/12/episode-33-rlhf-problems-scott-emmons.html

    Topics we discuss, and timestamps:

    0:00:33 - Deceptive inflation

    0:17:56 - Overjustification

    0:32:48 - Bounded human rationality

    0:50:46 - Avoiding these problems

    1:14:13 - Dimensional analysis

    1:23:32 - RLHF problems, in theory and practice

    1:31:29 - Scott's research program

    1:39:42 - Following Scott's research

    Scott's website: https://www.scottemmons.com

    Scott's X/twitter account: https://x.com/emmons_scott

    When Your AIs Deceive You: Challenges With Partial Observability of Human Evaluators in Reward Learning: https://arxiv.org/abs/2402.17747

    Other works we discuss:

    AI Deception: A Survey of Examples, Risks, and Potential Solutions: https://arxiv.org/abs/2308.14752

    Uncertain decisions facilitate better preference learning: https://arxiv.org/abs/2106.10394

    Invariance in Policy Optimisation and Partial Identifiability in Reward Learning: https://arxiv.org/abs/2203.07475

    The Humble Gaussian Distribution (aka principal component analysis and dimensional analysis): http://www.inference.org.uk/mackay/humble.pdf

    Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!: https://arxiv.org/abs/2310.03693

    Episode art by Hamish Doodles: hamishdoodles.com

  • What's the difference between a large language model and the human brain? And what's wrong with our theories of agency? In this episode, I chat about these questions with Jan Kulveit, who leads the Alignment of Complex Systems research group.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    The transcript: axrp.net/episode/2024/05/30/episode-32-understanding-agency-jan-kulveit.html

    Topics we discuss, and timestamps:

    0:00:47 - What is active inference?

    0:15:14 - Preferences in active inference

    0:31:33 - Action vs perception in active inference

    0:46:07 - Feedback loops

    1:01:32 - Active inference vs LLMs

    1:12:04 - Hierarchical agency

    1:58:28 - The Alignment of Complex Systems group

    Website of the Alignment of Complex Systems group (ACS): acsresearch.org

    ACS on X/Twitter: x.com/acsresearchorg

    Jan on LessWrong: lesswrong.com/users/jan-kulveit

    Predictive Minds: Large Language Models as Atypical Active Inference Agents: arxiv.org/abs/2311.10215

    Other works we discuss:

    Active Inference: The Free Energy Principle in Mind, Brain, and Behavior: https://www.goodreads.com/en/book/show/58275959

    Book Review: Surfing Uncertainty: https://slatestarcodex.com/2017/09/05/book-review-surfing-uncertainty/

    The self-unalignment problem: https://www.lesswrong.com/posts/9GyniEBaN3YYTqZXn/the-self-unalignment-problem

    Mitigating generative agent social dilemmas (aka language models writing contracts for Minecraft): https://social-dilemmas.github.io/

    Episode art by Hamish Doodles: hamishdoodles.com

  • What's going on with deep learning? What sorts of models get learned, and what are the learning dynamics? Singular learning theory is a theory of Bayesian statistics broad enough in scope to encompass deep neural networks that may help answer these questions. In this episode, I speak with Daniel Murfet about this research program and what it tells us.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    Topics we discuss, and timestamps:

    0:00:26 - What is singular learning theory?

    0:16:00 - Phase transitions

    0:35:12 - Estimating the local learning coefficient

    0:44:37 - Singular learning theory and generalization

    1:00:39 - Singular learning theory vs other deep learning theory

    1:17:06 - How singular learning theory hit AI alignment

    1:33:12 - Payoffs of singular learning theory for AI alignment

    1:59:36 - Does singular learning theory advance AI capabilities?

    2:13:02 - Open problems in singular learning theory for AI alignment

    2:20:53 - What is the singular fluctuation?

    2:25:33 - How geometry relates to information

    2:30:13 - Following Daniel Murfet's work

    The transcript: https://axrp.net/episode/2024/05/07/episode-31-singular-learning-theory-dan-murfet.html

    Daniel Murfet's twitter/X account: https://twitter.com/danielmurfet

    Developmental interpretability website: https://devinterp.com

    Developmental interpretability YouTube channel: https://www.youtube.com/@Devinterp

    Main research discussed in this episode:

    - Developmental Landscape of In-Context Learning: https://arxiv.org/abs/2402.02364

    - Estimating the Local Learning Coefficient at Scale: https://arxiv.org/abs/2402.03698

    - Simple versus Short: Higher-order degeneracy and error-correction: https://www.lesswrong.com/posts/nWRj6Ey8e5siAEXbK/simple-versus-short-higher-order-degeneracy-and-error-1

    Other links:

    - Algebraic Geometry and Statistical Learning Theory (the grey book): https://www.cambridge.org/core/books/algebraic-geometry-and-statistical-learning-theory/9C8FD1BDC817E2FC79117C7F41544A3A

    - Mathematical Theory of Bayesian Statistics (the green book): https://www.routledge.com/Mathematical-Theory-of-Bayesian-Statistics/Watanabe/p/book/9780367734817
    In-context learning and induction heads: https://transformer-circuits.pub/2022/in-context-learning-and-induction-heads/index.html

    - Saddle-to-Saddle Dynamics in Deep Linear Networks: Small Initialization Training, Symmetry, and Sparsity: https://arxiv.org/abs/2106.15933

    - A mathematical theory of semantic development in deep neural networks: https://www.pnas.org/doi/abs/10.1073/pnas.1820226116

    - Consideration on the Learning Efficiency Of Multiple-Layered Neural Networks with Linear Units: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4404877

    - Neural Tangent Kernel: Convergence and Generalization in Neural Networks: https://arxiv.org/abs/1806.07572

    - The Interpolating Information Criterion for Overparameterized Models: https://arxiv.org/abs/2307.07785

    - Feature Learning in Infinite-Width Neural Networks: https://arxiv.org/abs/2011.14522

    - A central AI alignment problem: capabilities generalization, and the sharp left turn: https://www.lesswrong.com/posts/GNhMPAWcfBCASy8e6/a-central-ai-alignment-problem-capabilities-generalization

    - Quantifying degeneracy in singular models via the learning coefficient: https://arxiv.org/abs/2308.12108

    Episode art by Hamish Doodles: hamishdoodles.com

  • Top labs use various forms of "safety training" on models before their release to make sure they don't do nasty stuff - but how robust is that? How can we ensure that the weights of powerful AIs don't get leaked or stolen? And what can AI even do these days? In this episode, I speak with Jeffrey Ladish about security and AI.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    Topics we discuss, and timestamps:

    0:00:38 - Fine-tuning away safety training

    0:13:50 - Dangers of open LLMs vs internet search

    0:19:52 - What we learn by undoing safety filters

    0:27:34 - What can you do with jailbroken AI?

    0:35:28 - Security of AI model weights

    0:49:21 - Securing against attackers vs AI exfiltration

    1:08:43 - The state of computer security

    1:23:08 - How AI labs could be more secure

    1:33:13 - What does Palisade do?

    1:44:40 - AI phishing

    1:53:32 - More on Palisade's work

    1:59:56 - Red lines in AI development

    2:09:56 - Making AI legible

    2:14:08 - Following Jeffrey's research

    The transcript: axrp.net/episode/2024/04/30/episode-30-ai-security-jeffrey-ladish.html

    Palisade Research: palisaderesearch.org

    Jeffrey's Twitter/X account: twitter.com/JeffLadish

    Main papers we discussed:

    - LoRA Fine-tuning Efficiently Undoes Safety Training in Llama 2-Chat 70B: arxiv.org/abs/2310.20624

    - BadLLaMa: Cheaply Removing Safety Fine-tuning From LLaMa 2-Chat 13B: arxiv.org/abs/2311.00117

    - Securing Artificial Intelligence Model Weights: rand.org/pubs/working_papers/WRA2849-1.html

    Other links:

    - Llama 2: Open Foundation and Fine-Tuned Chat Models: https://arxiv.org/abs/2307.09288

    - Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!: https://arxiv.org/abs/2310.03693

    - Shadow Alignment: The Ease of Subverting Safely-Aligned Language Models: https://arxiv.org/abs/2310.02949

    - On the Societal Impact of Open Foundation Models (Stanford paper on marginal harms from open-weight models): https://crfm.stanford.edu/open-fms/

    - The Operational Risks of AI in Large-Scale Biological Attacks (RAND): https://www.rand.org/pubs/research_reports/RRA2977-2.html

    - Preventing model exfiltration with upload limits: https://www.alignmentforum.org/posts/rf66R4YsrCHgWx9RG/preventing-model-exfiltration-with-upload-limits

    - A deep dive into an NSO zero-click iMessage exploit: Remote Code Execution: https://googleprojectzero.blogspot.com/2021/12/a-deep-dive-into-nso-zero-click.html

    - In-browser transformer inference: https://aiserv.cloud/

    - Anatomy of a rental phishing scam: https://jeffreyladish.com/anatomy-of-a-rental-phishing-scam/

    - Causal Scrubbing: a method for rigorously testing interpretability hypotheses: https://www.alignmentforum.org/posts/JvZhhzycHu2Yd57RN/causal-scrubbing-a-method-for-rigorously-testing

    Episode art by Hamish Doodles: hamishdoodles.com

  • In 2022, it was announced that a fairly simple method can be used to extract the true beliefs of a language model on any given topic, without having to actually understand the topic at hand. Earlier, in 2021, it was announced that neural networks sometimes 'grok': that is, when training them on certain tasks, they initially memorize their training data (achieving their training goal in a way that doesn't generalize), but then suddenly switch to understanding the 'real' solution in a way that generalizes. What's going on with these discoveries? Are they all they're cracked up to be, and if so, how are they working? In this episode, I talk to Vikrant Varma about his research getting to the bottom of these questions.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    Topics we discuss, and timestamps:

    0:00:36 - Challenges with unsupervised LLM knowledge discovery, aka contra CCS

    0:00:36 - What is CCS?

    0:09:54 - Consistent and contrastive features other than model beliefs

    0:20:34 - Understanding the banana/shed mystery

    0:41:59 - Future CCS-like approaches

    0:53:29 - CCS as principal component analysis

    0:56:21 - Explaining grokking through circuit efficiency

    0:57:44 - Why research science of deep learning?

    1:12:07 - Summary of the paper's hypothesis

    1:14:05 - What are 'circuits'?

    1:20:48 - The role of complexity

    1:24:07 - Many kinds of circuits

    1:28:10 - How circuits are learned

    1:38:24 - Semi-grokking and ungrokking

    1:50:53 - Generalizing the results

    1:58:51 - Vikrant's research approach

    2:06:36 - The DeepMind alignment team

    2:09:06 - Follow-up work

    The transcript: axrp.net/episode/2024/04/25/episode-29-science-of-deep-learning-vikrant-varma.html

    Vikrant's Twitter/X account: twitter.com/vikrantvarma_

    Main papers:

    - Challenges with unsupervised LLM knowledge discovery: arxiv.org/abs/2312.10029

    - Explaining grokking through circuit efficiency: arxiv.org/abs/2309.02390

    Other works discussed:

    - Discovering latent knowledge in language models without supervision (CCS): arxiv.org/abs/2212.03827

    - Eliciting Latent Knowledge: How to Tell if your Eyes Deceive You: https://docs.google.com/document/d/1WwsnJQstPq91_Yh-Ch2XRL8H_EpsnjrC1dwZXR37PC8/edit

    - Discussion: Challenges with unsupervised LLM knowledge discovery: lesswrong.com/posts/wtfvbsYjNHYYBmT3k/discussion-challenges-with-unsupervised-llm-knowledge-1

    - Comment thread on the banana/shed results: lesswrong.com/posts/wtfvbsYjNHYYBmT3k/discussion-challenges-with-unsupervised-llm-knowledge-1?commentId=hPZfgA3BdXieNfFuY

    - Fabien Roger, What discovering latent knowledge did and did not find: lesswrong.com/posts/bWxNPMy5MhPnQTzKz/what-discovering-latent-knowledge-did-and-did-not-find-4

    - Scott Emmons, Contrast Pairs Drive the Performance of Contrast Consistent Search (CCS): lesswrong.com/posts/9vwekjD6xyuePX7Zr/contrast-pairs-drive-the-empirical-performance-of-contrast

    - Grokking: Generalizing Beyond Overfitting on Small Algorithmic Datasets: arxiv.org/abs/2201.02177

    - Keeping Neural Networks Simple by Minimizing the Minimum Description Length of the Weights (Hinton 1993 L2): dl.acm.org/doi/pdf/10.1145/168304.168306

    - Progress measures for grokking via mechanistic interpretability: arxiv.org/abs/2301.0521

    Episode art by Hamish Doodles: hamishdoodles.com

  • How should the law govern AI? Those concerned about existential risks often push either for bans or for regulations meant to ensure that AI is developed safely - but another approach is possible. In this episode, Gabriel Weil talks about his proposal to modify tort law to enable people to sue AI companies for disasters that are "nearly catastrophic".

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    Topics we discuss, and timestamps:

    0:00:35 - The basic idea

    0:20:36 - Tort law vs regulation

    0:29:10 - Weil's proposal vs Hanson's proposal

    0:37:00 - Tort law vs Pigouvian taxation

    0:41:16 - Does disagreement on AI risk make this proposal less effective?

    0:49:53 - Warning shots - their prevalence and character

    0:59:17 - Feasibility of big changes to liability law

    1:29:17 - Interactions with other areas of law

    1:38:59 - How Gabriel encountered the AI x-risk field

    1:42:41 - AI x-risk and the legal field

    1:47:44 - Technical research to help with this proposal

    1:50:47 - Decisions this proposal could influence

    1:55:34 - Following Gabriel's research

    The transcript: axrp.net/episode/2024/04/17/episode-28-tort-law-for-ai-risk-gabriel-weil.html

    Links for Gabriel:

    - SSRN page: papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=1648032

    - Twitter/X account: twitter.com/gabriel_weil

    Tort Law as a Tool for Mitigating Catastrophic Risk from Artificial Intelligence: papers.ssrn.com/sol3/papers.cfm?abstract_id=4694006

    Other links:

    - Foom liability: overcomingbias.com/p/foom-liability

    - Punitive Damages: An Economic Analysis: law.harvard.edu/faculty/shavell/pdf/111_Harvard_Law_Rev_869.pdf

    - Efficiency, Fairness, and the Externalization of Reasonable Risks: The Problem With the Learned Hand Formula: papers.ssrn.com/sol3/papers.cfm?abstract_id=4466197

    - Tort Law Can Play an Important Role in Mitigating AI Risk: forum.effectivealtruism.org/posts/epKBmiyLpZWWFEYDb/tort-law-can-play-an-important-role-in-mitigating-ai-risk

    - How Technical AI Safety Researchers Can Help Implement Punitive Damages to Mitigate Catastrophic AI Risk: forum.effectivealtruism.org/posts/yWKaBdBygecE42hFZ/how-technical-ai-safety-researchers-can-help-implement

    - Can the courts save us from dangerous AI? [Vox]: vox.com/future-perfect/2024/2/7/24062374/ai-openai-anthropic-deepmind-legal-liability-gabriel-weil

    Episode art by Hamish Doodles: hamishdoodles.com

  • A lot of work to prevent AI existential risk takes the form of ensuring that AIs don't want to cause harm or take over the world---or in other words, ensuring that they're aligned. In this episode, I talk with Buck Shlegeris and Ryan Greenblatt about a different approach, called "AI control": ensuring that AI systems couldn't take over the world, even if they were trying to.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    Topics we discuss, and timestamps:

    0:00:31 - What is AI control?

    0:16:16 - Protocols for AI control

    0:22:43 - Which AIs are controllable?

    0:29:56 - Preventing dangerous coded AI communication

    0:40:42 - Unpredictably uncontrollable AI

    0:58:01 - What control looks like

    1:08:45 - Is AI control evil?

    1:24:42 - Can red teams match misaligned AI?

    1:36:51 - How expensive is AI monitoring?

    1:52:32 - AI control experiments

    2:03:50 - GPT-4's aptitude at inserting backdoors

    2:14:50 - How AI control relates to the AI safety field

    2:39:25 - How AI control relates to previous Redwood Research work

    2:49:16 - How people can work on AI control

    2:54:07 - Following Buck and Ryan's research

    The transcript: axrp.net/episode/2024/04/11/episode-27-ai-control-buck-shlegeris-ryan-greenblatt.html

    Links for Buck and Ryan:

    - Buck's twitter/X account: twitter.com/bshlgrs

    - Ryan on LessWrong: lesswrong.com/users/ryan_greenblatt

    - You can contact both Buck and Ryan by electronic mail at [firstname] [at-sign] rdwrs.com

    Main research works we talk about:

    - The case for ensuring that powerful AIs are controlled: lesswrong.com/posts/kcKrE9mzEHrdqtDpE/the-case-for-ensuring-that-powerful-ais-are-controlled

    - AI Control: Improving Safety Despite Intentional Subversion: arxiv.org/abs/2312.06942

    Other things we mention:

    - The prototypical catastrophic AI action is getting root access to its datacenter (aka "Hacking the SSH server"): lesswrong.com/posts/BAzCGCys4BkzGDCWR/the-prototypical-catastrophic-ai-action-is-getting-root

    - Preventing language models from hiding their reasoning: arxiv.org/abs/2310.18512

    - Improving the Welfare of AIs: A Nearcasted Proposal: lesswrong.com/posts/F6HSHzKezkh6aoTr2/improving-the-welfare-of-ais-a-nearcasted-proposal

    - Measuring coding challenge competence with APPS: arxiv.org/abs/2105.09938

    - Causal Scrubbing: a method for rigorously testing interpretability hypotheses lesswrong.com/posts/JvZhhzycHu2Yd57RN/causal-scrubbing-a-method-for-rigorously-testing

    Episode art by Hamish Doodles: hamishdoodles.com

  • The events of this year have highlighted important questions about the governance of artificial intelligence. For instance, what does it mean to democratize AI? And how should we balance benefits and dangers of open-sourcing powerful AI systems such as large language models? In this episode, I speak with Elizabeth Seger about her research on these questions.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    Topics we discuss, and timestamps:

    - 0:00:40 - What kinds of AI?

    - 0:01:30 - Democratizing AI

    - 0:04:44 - How people talk about democratizing AI

    - 0:09:34 - Is democratizing AI important?

    - 0:13:31 - Links between types of democratization

    - 0:22:43 - Democratizing profits from AI

    - 0:27:06 - Democratizing AI governance

    - 0:29:45 - Normative underpinnings of democratization

    - 0:44:19 - Open-sourcing AI

    - 0:50:47 - Risks from open-sourcing

    - 0:56:07 - Should we make AI too dangerous to open source?

    - 1:00:33 - Offense-defense balance

    - 1:03:13 - KataGo as a case study

    - 1:09:03 - Openness for interpretability research

    - 1:15:47 - Effectiveness of substitutes for open sourcing

    - 1:20:49 - Offense-defense balance, part 2

    - 1:29:49 - Making open-sourcing safer?

    - 1:40:37 - AI governance research

    - 1:41:05 - The state of the field

    - 1:43:33 - Open questions

    - 1:49:58 - Distinctive governance issues of x-risk

    - 1:53:04 - Technical research to help governance

    - 1:55:23 - Following Elizabeth's research

    The transcript: https://axrp.net/episode/2023/11/26/episode-26-ai-governance-elizabeth-seger.html

    Links for Elizabeth:

    - Personal website: elizabethseger.com

    - Centre for the Governance of AI (AKA GovAI): governance.ai

    Main papers:

    - Democratizing AI: Multiple Meanings, Goals, and Methods: arxiv.org/abs/2303.12642

    - Open-sourcing highly capable foundation models: an evaluation of risks, benefits, and alternative methods for pursuing open source objectives: papers.ssrn.com/sol3/papers.cfm?abstract_id=4596436

    Other research we discuss:

    - What Do We Mean When We Talk About "AI democratisation"? (blog post): governance.ai/post/what-do-we-mean-when-we-talk-about-ai-democratisation

    - Democratic Inputs to AI (OpenAI): openai.com/blog/democratic-inputs-to-ai

    - Collective Constitutional AI: Aligning a Language Model with Public Input (Anthropic): anthropic.com/index/collective-constitutional-ai-aligning-a-language-model-with-public-input

    - Against "Democratizing AI": johanneshimmelreich.net/papers/against-democratizing-AI.pdf

    - Adversarial Policies Beat Superhuman Go AIs: goattack.far.ai

    - Structured access: an emerging paradigm for safe AI deployment: arxiv.org/abs/2201.05159

    - Universal and Transferable Adversarial Attacks on Aligned Language Models (aka Adversarial Suffixes): arxiv.org/abs/2307.15043

    Episode art by Hamish Doodles: hamishdoodles.com

  • Imagine a world where there are many powerful AI systems, working at cross purposes. You could suppose that different governments use AIs to manage their militaries, or simply that many powerful AIs have their own wills. At any rate, it seems valuable for them to be able to cooperatively work together and minimize pointless conflict. How do we ensure that AIs behave this way - and what do we need to learn about how rational agents interact to make that more clear? In this episode, I'll be speaking with Caspar Oesterheld about some of his research on this very topic.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    Episode art by Hamish Doodles: hamishdoodles.com

    Topics we discuss, and timestamps:

    - 0:00:34 - Cooperative AI

    - 0:06:21 - Cooperative AI vs standard game theory

    - 0:19:45 - Do we need cooperative AI if we get alignment?

    - 0:29:29 - Cooperative AI and agent foundations

    - 0:34:59 - A Theory of Bounded Inductive Rationality

    - 0:50:05 - Why it matters

    - 0:53:55 - How the theory works

    - 1:01:38 - Relationship to logical inductors

    - 1:15:56 - How fast does it converge?

    - 1:19:46 - Non-myopic bounded rational inductive agents?

    - 1:24:25 - Relationship to game theory

    - 1:30:39 - Safe Pareto Improvements

    - 1:30:39 - What they try to solve

    - 1:36:15 - Alternative solutions

    - 1:40:46 - How safe Pareto improvements work

    - 1:51:19 - Will players fight over which safe Pareto improvement to adopt?

    - 2:06:02 - Relationship to program equilibrium

    - 2:11:25 - Do safe Pareto improvements break themselves?

    - 2:15:52 - Similarity-based Cooperation

    - 2:23:07 - Are similarity-based cooperators overly cliqueish?

    - 2:27:12 - Sensitivity to noise

    - 2:29:41 - Training neural nets to do similarity-based cooperation

    - 2:50:25 - FOCAL, Caspar's research lab

    - 2:52:52 - How the papers all relate

    - 2:57:49 - Relationship to functional decision theory

    - 2:59:45 - Following Caspar's research

    The transcript: axrp.net/episode/2023/10/03/episode-25-cooperative-ai-caspar-oesterheld.html

    Links for Caspar:

    - FOCAL at CMU: www.cs.cmu.edu/~focal/

    - Caspar on X, formerly known as Twitter: twitter.com/C_Oesterheld

    - Caspar's blog: casparoesterheld.com/

    - Caspar on Google Scholar: scholar.google.com/citations?user=xeEcRjkAAAAJ&hl=en&oi=ao

    Research we discuss:

    - A Theory of Bounded Inductive Rationality: arxiv.org/abs/2307.05068

    - Safe Pareto improvements for delegated game playing: link.springer.com/article/10.1007/s10458-022-09574-6

    - Similarity-based Cooperation: arxiv.org/abs/2211.14468

    - Logical Induction: arxiv.org/abs/1609.03543

    - Program Equilibrium: citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=e1a060cda74e0e3493d0d81901a5a796158c8410

    - Formalizing Objections against Surrogate Goals: www.alignmentforum.org/posts/K4FrKRTrmyxrw5Dip/formalizing-objections-against-surrogate-goals

    - Learning with Opponent-Learning Awareness: arxiv.org/abs/1709.04326

  • Recently, OpenAI made a splash by announcing a new "Superalignment" team. Lead by Jan Leike and Ilya Sutskever, the team would consist of top researchers, attempting to solve alignment for superintelligent AIs in four years by figuring out how to build a trustworthy human-level AI alignment researcher, and then using it to solve the rest of the problem. But what does this plan actually involve? In this episode, I talk to Jan Leike about the plan and the challenges it faces.

    Patreon: patreon.com/axrpodcast

    Ko-fi: ko-fi.com/axrpodcast

    Episode art by Hamish Doodles: hamishdoodles.com/

    Topics we discuss, and timestamps:

    - 0:00:37 - The superalignment team

    - 0:02:10 - What's a human-level automated alignment researcher?

    - 0:06:59 - The gap between human-level automated alignment researchers and superintelligence

    - 0:18:39 - What does it do?

    - 0:24:13 - Recursive self-improvement

    - 0:26:14 - How to make the AI AI alignment researcher

    - 0:30:09 - Scalable oversight

    - 0:44:38 - Searching for bad behaviors and internals

    - 0:54:14 - Deliberately training misaligned models

    - 1:02:34 - Four year deadline

    - 1:07:06 - What if it takes longer?

    - 1:11:38 - The superalignment team and...

    - 1:11:38 - ... governance

    - 1:14:37 - ... other OpenAI teams

    - 1:18:17 - ... other labs

    - 1:26:10 - Superalignment team logistics

    - 1:29:17 - Generalization

    - 1:43:44 - Complementary research

    - 1:48:29 - Why is Jan optimistic?

    - 1:58:32 - Long-term agency in LLMs?

    - 2:02:44 - Do LLMs understand alignment?

    - 2:06:01 - Following Jan's research

    The transcript: axrp.net/episode/2023/07/27/episode-24-superalignment-jan-leike.html

    Links for Jan and OpenAI:

    - OpenAI jobs: openai.com/careers

    - Jan's substack: aligned.substack.com

    - Jan's twitter: twitter.com/janleike

    Links to research and other writings we discuss:

    - Introducing Superalignment: openai.com/blog/introducing-superalignment

    - Let's Verify Step by Step (process-based feedback on math): arxiv.org/abs/2305.20050

    - Planning for AGI and beyond:
    openai.com/blog/planning-for-agi-and-beyond

    - Self-critiquing models for assisting human evaluators: arxiv.org/abs/2206.05802

    - An Interpretability Illusion for BERT: arxiv.org/abs/2104.07143

    - Language models can explain neurons in language models https://openaipublic.blob.core.windows.net/neuron-explainer/paper/index.html

    - Our approach to alignment research: openai.com/blog/our-approach-to-alignment-research

    - Training language models to follow instructions with human feedback (aka the Instruct-GPT paper): arxiv.org/abs/2203.02155