I work primarily on AI Alignment. My main direction at the moment is to accelerate alignment work via language models and interpretability.
Yeah, I think most of the gains we've gotten from AI have been in coding and learning. Many of the big promises have yet to be met; definitely still a struggle to get it to work well for writing (in the style we'd want it to write) or getting AI agents to work well, so it limits the possible useful application.
I quickly wrote up some rough project ideas for ARENA and LASR participants, so I figured I'd share them here as well. I am happy to discuss these ideas and potentially collaborate on some of them.
MAIA (Multimodal Automated Interpretability Agent) is a system designed to help users understand AI models by combining human-like experimentation flexibility with automated scalability. It answers user queries about AI system components by iteratively generating hypotheses, designing and running experiments, observing outcomes, and updating hypotheses.
MAIA uses a vision-language model (GPT-4V, at the time) backbone equipped with an API of interpretability experiment tools. This modular system can address both "macroscopic" questions (e.g., identifying systematic biases in model predictions) and "microscopic" questions (e.g., describing individual features) with simple query modifications.
This project aims to improve MAIA's ability to either answer macroscopic questions or microscopic questions on vision models.
MAIA is focused on vision models, so this project aims to create a MAIA-like setup, but for the interpretability of LLMs.
Given that this would require creating a new setup for language models, it would make sense to come up with simple interpretability benchmark examples to test MAIA-LLM. The easiest way to do this would be to either look for existing LLM interpretability benchmarks or create one based on interpretability results we've already verified (would be ideal to have a ground truth). Ideally, the examples in the benchmark would be simple, but new enough that the LLM has not seen them in its training data.
Critique-out-Loud reward models are reward models that can reason explicitly about the quality of an input through producing Chain-of-Thought like critiques of an input before predicting a reward. In classic reward model training, the reward model is trained as a reward head initialized on top of the base LLM. Without LM capabilities, classic reward models act as encoders and must predict rewards within a single forward pass through the model, meaning reasoning must happen implicitly. In contrast, CLoud reward models are trained to both produce explicit reasoning about quality and to score based on these critique reasoning traces. CLoud reward models lead to large gains for pairwise preference modeling on RewardBench, and also lead to large gains in win rate when used as the scoring model in Best-of-N sampling on ArenaHard.
The goal for this project would be to test the robustness of CLoud reward models. For example, are the CLoud RMs (discriminators) more robust to jailbreaking attacks from the policy (generator)? Do the CLoud RMs generalize better?
From an alignment perspective, we would want RMs that generalize further out-of-distribution (and ideally, always more than the generator we are training).
Simple synthetic data reduces sycophancy in large language models by (Google) reduced sycophancy in LLMs with a fairly small number of synthetic data examples. This project would involve testing this technique for other behavioural interventions and (potentially) studying the scaling laws. Consider looking at the examples from the Model-Written Evaluations paper by Anthropic to find some behaviours to test.
Explore the effectiveness of different regularization techniques (e.g. L1 regularization, weight pruning, activation sparsity) in improving the interpretability and/or editability of language models, and assess their impact on model performance and alignment. We expect we could apply automated interpretability methods (e.g. MAIA) to this project to test how well the different regularization techniques impact the model.
In some sense, this research is similar to the work Anthropic did with SoLU activation functions. Unfortunately, they needed to add layer norms to make the SoLU models competitive, which seems to have hidden away the superposition in other parts of the network, making SoLU unhelpful in making the models more interpretable
That said, we hope to find that we can increase our ability to interpret these models through regularization techniques. A technique like L1 regularization should help because it encourages the model to learn sparse representations by penalizing non-zero weights or activations. Sparse models tend to be more interpretable as they rely on a smaller set of important features.
Methodology:
Expected Outcomes:
Investigate how misspecified reward functions influence the behavior of language models during fine-tuning and measure the extent to which the model's outputs are steered by the reward labels, even when they contradict the input context. We hope to better understand language model training dynamics. Additionally, we expect online learning to complicate things in the future, where models will be able to generate the data they may eventually be trained on. We hope that insights from this work can help us prevent catastrophic feedback loops in the future. For example, if model behavior is mostly impacted by training data, we may prefer to shape model behavior through synthetic data (it has been shown we can reduce sycophancy by doing this).
Prior works:
Methodology:
Expected Outcomes:
Understand the underlying mechanisms that lead to language models producing correct answers through flawed reasoning, and develop techniques to detect and mitigate such behavior. Essentially, we want to apply interpretability techniques to help us identify which sets of activations or token-layer pairs impact the model getting the correct answer when it has the correct reasoning versus when it has the incorrect reasoning. The hope is to uncover systematic differences as to when it is not relying on its chain-of-thought at all and when it does leverage its chain-of-thought to get the correct answer.
[EDIT Oct 2nd, 2024] This project intends to follow a similar line of reasoning as described in this post and this comment. The goal is to study chains-of-thought and improve faithfulness without suffering an alignment tax so that we can have highly interpretable systems through their token outputs and prevent loss of control. The project doesn't necessarily need to rely only on model internals.
Related work:
Methodology:
Expected Outcomes:
I'm exploring the possibility of building an alignment research organization focused on augmenting alignment researchers and progressively automating alignment research (yes, I have thought deeply about differential progress and other concerns). I intend to seek funding in the next few months, and I'd like to chat with people interested in this kind of work, especially great research engineers and full-stack engineers who might want to cofound such an organization. If you or anyone you know might want to chat, let me know! Send me a DM, and I can send you some initial details about the organization's vision.
Here are some things I'm looking for in potential co-founders:
Need
Nice-to-have
Hey everyone, in collaboration with Apart Research, I'm helping organize a hackathon this weekend to build tools for accelerating alignment research. This hackathon is very much related to my effort in building an "Alignment Research Assistant."
Here's the announcement post:
2 days until we revolutionize AI alignment research at the Research Augmentation Hackathon!
As AI safety researchers, we pour countless hours into crucial work. It's time we built tools to accelerate our efforts! Join us in creating AI assistants that could supercharge the very research we're passionate about.
Date: July 26th to 28th, online and in-person
Prizes: $2,000 in prizes
Why join?
* Build tools that matter for the future of AI
* Learn from top minds in AI alignment
* Boost your skills and portfolio
We've got a Hackbook with an exciting project to work on waiting for you! No advanced AI knowledge required - just bring your creativity!
Register now: Sign up on the website here, and don't miss this chance to shape the future of AI research!
We're doing a hackathon with Apart Research on 26th. I created a list of problem statements for people to brainstorm off of.
Pro-active insight extraction from new research
Reading papers can take a long time and is often not worthwhile. As a result, researchers might read too many papers or almost none. However, there are still valuable nuggets in papers and posts. The issue is finding them. So, how might we design an AI research assistant that proactively looks at new papers (and old) and shares valuable information with researchers in a naturally consumable way? Part of this work involves presenting individual research with what they would personally find valuable and not overwhelm them with things they are less interested in.
How can we improve the LLM experience for researchers?
Many alignment researchers will use language models much less than they would like to because they don't know how to prompt the models, it takes time to create a valuable prompt, the model doesn't have enough context for their project, the model is not up-to-date on the latest techniques, etc. How might we make LLMs more useful for researchers by relieving them of those bottlenecks?
Simple experiments can be done quickly, but turning it into a full project can take a lot of time
One key bottleneck for alignment research is transitioning from an initial 24-hour simple experiment in a notebook to a set of complete experiments tested with different models, datasets, interventions, etc. How can we help researchers move through that second research phase much faster?
How might we use AI agents to automate alignment research?
As AI agents become more capable, we can use them to automate parts of alignment research. The paper "A Multimodal Automated Interpretability Agent" serves as an initial attempt at this. How might we use AI agents to help either speed up alignment research or unlock paths that were previously inaccessible?
How can we nudge research toward better objectives (agendas or short experiments) for their research?
Even if we make researchers highly efficient, it means nothing if they are not working on the right things. Choosing the right objectives (projects and next steps) through time can be the difference between 0x to 1x to +100x. How can we ensure that researchers are working on the most valuable things?
What can be done to accelerate implementation and iteration speed?
Implementation and iteration speed on the most informative experiments matter greatly. How can we nudge them to gain the most bits of information in the shortest time? This involves helping them work on the right agendas/projects and helping them break down their projects in ways that help them make progress faster (and avoiding ending up tunnel-visioned on the wrong project for months/years).
How can we connect all of the ideas in the field?
How can we integrate the open questions/projects in the field (with their critiques) in such a way that helps the researcher come up with well-grounded research directions faster? How can we aid them in choosing better directions and adjust throughout their research? This kind of work may eventually be a precursor to guiding AI agents to help us develop better ideas for alignment research.
As an update to the Alignment Research Assistant I'm building, here is a set of shovel-ready tasks I would like people to contribute to (please DM if you'd like to contribute!):
1. Setup the Continue extension for research: https://www.continue.dev/
2. Data sourcing and management
3. Extract answers to questions across multiple papers/posts (feeds into Continue)
4. Design Autoprompts for alignment research
5. Simulated Paper Reviewer
6. Jargon and Prerequisite Explainer
7. Setup automated "suggestion-LLM"
8. Figure out if we can get a useable browser inside of VSCode (tried quickly with the Edge extension but couldn't sign into the Claude chat website)
9. "Alignment Research Codebase" integration (can add as Continue backend)
Bulk fast content extraction
Personalized Research Newsletter
Discord Bot for Project Proposals
I've created a private discord server to discuss this work. If you'd like to contribute to this project (or might want to in the future if you see a feature you'd like to contribute to) or if you are an alignment/governance researcher who would like to be a beta user so we can iterate faster, please DM me for a link!
I’m still getting the hang of it, but primarily have been using it when I want to brainstorm some project ideas that I can later pass off to an LLM for context on what I’m working on or when I want to reflect on a previous meeting I had. Will probably turn it on about ~1 time per week while I’m walking to work and ramble about a project in case I think of something good. (I also sometimes use it to explain the project spec or small adjustments I want my AI coding assistant to do.)
Sometimes I’ll use the Advanced Voice Mode or normal voice mode from ChatGPT for this instead. For example, I used it to practice for an interview after passing off a lot of the context to the model (my CV, the org, etc). I used this to just blurt out all the thoughts I have in my head in a question-answer format and then asked the AI for feedback on my answers and asked it to give a summary of the conversation (like a cheat sheet to remind myself what I want to talk about).