Levelling Up in AI Safety Research Engineering

Summary: A level-based guide for independently up-skilling in AI Safety Research Engineering that aims to give concrete objectives, goals, and resources to help anyone go from zero to hero.

Cross-posted to LessWrong. View a pretty Google Docs version here.

Introduction

I think great career guides are really useful for guiding and structuring the learning journey of people new to a technical field like AI Safety. I also like role-playing games. Here’s my attempt to use levelling frameworks and break up one possible path from zero to hero in Research Engineering for AI Safety (e.g. jobs with the “Research Engineer” title) through objectives, concrete goals, and resources. I hope this kind of framework makes it easier to see where one is on this journey, how far they have to go, and some options to get there.

I’m mostly making this to sort out my own thoughts about my career development and how I’ll support other students through Stanford AI Alignment, but hopefully, this is also useful to others! Note that I assume some interest in AI Safety Research Engineering—this guide is about how to up-skill in Research Engineering, not why (though working through it should be a great way to test your fit). Also note that there isn’t much abstract advice in this guide (see the end for links to guides with advice), and the goal is more to lay out concrete steps you can take to improve.

For each level, I describe the general capabilities of someone at the end of that level, some object-level goals to measure that capability, and some resources to choose from that would help get there. The categories of resources within a level are listed in the order you should progress, and resources within a category are roughly ordered by quality. There’s some redundancy, so I would recommend picking and choosing between the resources rather than doing all of them. Also, if you are a student and your university has a good class on one of the below topics, consider taking that instead of one of the online courses I listed.

As a very rough estimate, I think each level should take at least 100-200 hours of focused work, for a total of 700-1400 hours. At 10 hours/​week (quarter-time), that comes to around 16-32 months of study but can definitely be shorter (e.g. if you already have some experience) or longer (if you dive more deeply into some topics)! I think each level is about evenly split between time spent reading/​watching and time spent building/​testing, with more reading earlier on and more building later.

Confidence: mid-to-high. I am not yet an AI Safety Research Engineer (but I plan to be)—this is mostly a distillation of what I’ve read from other career guides (linked at the end) and talked about with people working on AI Safety. I definitely haven’t done all these things, just seen them recommended. I don’t expect this to be the “perfect” way to prepare for a career in AI Safety Research Engineering, but I do think it’s a very solid way.

Level 1: AI Safety Fundamentals

Objective‏‏‎ ‎

You are familiar with the basic arguments for existential risks due to advanced AI, models for forecasting AI advancements, and some of the past and current research directions within AI alignment/​safety. You have an opinion on how much you buy these arguments and whether you want to keep exploring AI Safety Research Engineering.

Why this first? Exposing yourself to these fundamental arguments and ideas is useful for testing your fit for AI Safety generally, but that isn’t to say you should “finish” this Level first and move on. Rather, you should be coming back to these readings and keeping up to date with the latest work in AI Safety throughout your learning journey. It’s okay if you don’t understand everything on your first try—Level 1 kind of happens all the time.

Goals‏‏‎ ‎

  • Complete an AI Safety introductory reading group fellowship.

  • Write a reflection distilling, recontextualizing, or expanding upon some AI Safety topic and share it with someone for feedback.

  • Figure out how convinced you are of the arguments for AI risk.

  • Decide if you want to continue learning about AI Safety Research Engineering, Theoretical AI Alignment, AI Policy and Strategy, or another field.

Resources‏‏‎ ‎

  1. AI Safety Reading Curriculum (Choose 1)

    1. 2022 AGI Safety Fundamentals alignment curriculum—EA Cambridge

    2. CUEA Standard AI Safety Reading Group Syllabus—Columbia EA

    3. Intro to ML Safety Course—CAIS

    4. AGI safety from first principles—Richard Ngo

  2. Additional Resources

    1. AI Alignment Forum

    2. Alignment Newsletter—Rohin Shah

    3. The Alignment Problem—Brian Christian

Level 2: Software Engineering

Objective‏‏‎ ‎

You can program in Python at the level of an introductory university course. You also know some other general software engineering tools/​skills like the command line, Git/​GitHub, documentation, and unit testing.

Why Python? Modern Machine Learning work, and thus AI Safety work, is almost entirely written in Python. Python is also an easier language for beginners to pick up, and there are plenty of resources for learning it.

Goals‏‏‎ ‎

  • Solve basic algorithmic programming problems with Python.

  • Know the basics of scientific computing with Python, including NumPy, and Jupyter/​Colab/​iPython Notebooks.

  • Create a new Git repository on GitHub, clone it, and add/​commit/​push changes to it for a personal project.

  • Know other software engineering skills like how to use the command line, write documentation, or make unit tests.

Resources‏‏‎ ‎

  1. Python Programming (Choose 1-2)

    1. www.learnpython.org

    2. Learn Python 3 - Codecademy

    3. Scientific Computing with Python Certification—freeCodeCamp

    4. CS50′s Introduction to Programming with Python—Harvard University

  2. Scientific Python (Choose 1-2)

    1. Data Analysis with Python Certification—freeCodeCamp

    2. Learn Python for Data Science, Structures, Algorithms, Interviews—Udemy

  3. Command Line (Choose 1-3)

    1. Learning the shell—LinuxCommand.org

    2. Linux Command Line Basics—Udacity

    3. The Unix Shell—software carpentries

  4. Git/​GitHub (Choose 2+)

    1. GitHub Tutorial—Beginner’s Training Guide—Anson Alexander

    2. Git Immersion

    3. GitHub Skills

    4. Version Control with Git—software carpentries

    5. first-contributions—contribute to open source projects

  5. Documentation (Choose 1-2)

    1. Documenting Python Code: A Complete Guide—Real Python

    2. Documenting Python Code: How to Guide—DataCamp

  6. Unit Testing (Choose 1-3)

    1. Getting Started With Testing in Python—Real Python

    2. A Gentle Introduction to Unit Testing in Python—Machine Learning Mastery

    3. Unit Testing in Python Tutorial—DataCamp

  7. Additional Resources

    1. Things I Wish Someone Had Told Me When I Was Learning How to Code

    2. The Hitchhiker’s Guide to Python!

    3. Cracking the Coding Interview—Gayle Laakmann McDowell

    4. Online Programming Learning Platform—LeetCode

    5. Challenging mathematical/​computer programming problems—Project Euler

    6. 100_Numpy_exercises—rougier

    7. The Good Research Code Handbook — Good Research Code

Level 3: Machine Learning

Objective‏‏‎ ‎

You have the mathematical context necessary for understanding Machine Learning (ML). You know the differences between supervised and unsupervised learning and between classification and regression. You understand common models like linear regression, logistic regression, neural networks, decision trees, and clustering, and you can code some of them in a library like PyTorch or JAX. You grasp core ML concepts like loss functions, regularization, bias/​variance, optimizers, metrics, and error analysis.

Why so much math? Machine learning at its core is basically applied statistics and multivariable calculus. It used to be that you needed to know this kind of math really well, but now with techniques like automatic differentiation, you can train neural networks without knowing much of what’s happening under the hood. These foundational resources are included for completeness, but you can probably spend a lot less time on math (e.g. the first few sections of each course) depending on what kind of engineering work you intend to do. You might want to come back and improve you math skills for understanding certain work in Levels 6-7, though, and if you find this math really interesting, you might be a good fit for theoretical AI alignment research.

Goals‏‏‎ ‎

  • Understand the mathematical basis of Machine Learning, especially linear algebra and multivariable calculus.

  • Write out the differences between supervised and unsupervised learning and between classification and regression.

  • Train and evaluate a simple neural network on a standard classification task like MNIST or a standard regression task like a Housing Dataset.

Resources‏‏‎ ‎

  1. Basic Calculus (Choose 1)

    1. Essence of calculus - 3Blue1Brown

  2. Probability (Choose 1)

    1. Probability—The Science of Uncertainty and Data—MIT

    2. Introduction to Probability—Harvard

    3. Part I: The Fundamentals | Introduction to Probability—MIT OpenCourseWare

  3. Linear Algebra (Choose 1)

    1. Essence of linear algebra − 3Blue1Brown

    2. Linear Algebra—MIT OpenCourseWare

    3. Georgia Tech’s course (parts 1, 2, 3, 4)

    4. Linear Algebra—Foundations to Frontiers—edX

    5. Linear Algebra Done Right—Sheldon Axler

  4. Multivariable Calculus (Choose 1)

    1. Multivariable Calculus—MIT OpenCourseWare

    2. Multivariable Calculus—Khan Academy

    3. The Matrix Calculus You Need For Deep Learning—explained.ai

    4. Mathematics for Machine Learning—Imperial College London

  5. Introductory Machine Learning (Choose 1-2)

    1. Course 6.036 - MIT Open Learning Library

    2. Machine Learning by Stanford University—Coursera

    3. Introduction to Machine Learning—Udacity

    4. Introduction to Machine Learning—Google

    5. Advanced Introduction to Machine Learning—CMU

    6. Supervised Machine Learning: Regression and Classification- DeepLearning.AI

  6. Additional Resources

    1. pytorch_exercises—Kyubyong

    2. Writing better code with pytorch+einops—einops

    3. Contemporary ML for Physicists—Jared Kaplan

Level 4: Deep Learning

Objective‏‏‎ ‎

You’ve dived deeper into Deep Learning (DL) through the lens of at least one subfield such as Natural Language Processing (NLP), Computer Vision (CV), or Reinforcement Learning (RL). You now have a better understanding of ML fundamentals, and you’ve reimplemented some core ML algorithms “from scratch.” You’ve started to build a portfolio of DL projects you can show others.

Goals‏‏‎ ‎

  • Be able to describe in moderate detail a wide range of modern deep learning architectures, techniques, and applications such as long short-term memory networks (LSTM) or convolutional neural networks (CNN).

  • Gain a more advanced understanding of machine learning by implementing autograd, backpropagation, and stochastic gradient descent “from scratch.”

  • Complete 1-3 deep learning projects, taking 10–20 hours each, in 1 or more sub-fields like NLP, CV, or RL.

Resources‏‏‎ ‎

  1. General Deep Learning (Choose 1)

    1. Practical Deep Learning for Coders—fast.ai

    2. Deep Learning by deeplearning.ai—Coursera

    3. Deep Learning—NYU

    4. PyTorch Tutorials

    5. Deep Learning for Computer Vision—UMich (Lectures 1-13 only)

    6. Deep Learning Online Training Course—Udacity

    7. Neural networks and deep learning—Michael Nielsen

  2. Advanced Machine Learning

    1. Studying (Choose 1-2)

      1. Backpropagation—CS231n Convolutional Neural Networks for Visual Recognition

      2. A Recipe for Training Neural Networks—Andrej Karpathy

    2. Implementing (Choose 1)

      1. MiniTorch (reimplement the core of PyTorch, self-study tips here)

      2. building micrograd—Andrej Karpathy

      3. Autodidax: JAX core from scratch

  3. Natural Language Processing (Choose 1 Or Another Sub-Field)

    1. Stanford CS 224N | Natural Language Processing with Deep Learning (lecture videos here)

    2. CS224U: Natural Language Understanding—Stanford University

    3. Week 12 Lecture: Deep Learning for Natural Language Processing (NLP) - NYU

    4. A Code-First Introduction to Natural Language Processing—fast.ai

  4. Computer Vision (Choose 1 Or Another Sub-Field)

    1. Deep Learning for Computer Vision—UMich

    2. CS231n: Deep Learning for Computer Vision—Stanford University (lecture videos here)

  5. Reinforcement Learning (Choose 1 Or Another Sub-Field)

    1. Spinning Up in Deep RL—OpenAI

    2. Deep Reinforcement Learning Class—Hugging Face

    3. Deep Reinforcement Learning: CS 285 Fall 2021 - UC Berkeley

    4. Deep Reinforcement Learning: Pong from Pixels—Andrej Karpathy

    5. DeepMind x UCL | Deep Learning Lecture Series 2021 - DeepMind

    6. Reinforcement Learning—University of Alberta

  6. Additional Resources

    1. Neural networks − 3Blue1Brown

    2. Cloud Computing Basics (Cloud 101) - Coursera

    3. Learn Cloud Computing with Online Courses, Classes, & Lessons—edX

Level 5: Understanding Transformers

Objective‏‏‎ ‎

You have a principled understanding of self-attention, cross-attention, and the general transformer architecture along with some of its variants. You are able to write a transformer like BERT or GPT-2 “from scratch” in PyTorch or JAX (a skill I believe Redwood Research looks for), and you can use resources like 🤗 Transformers to work with pre-trained transformer models. Through experimenting with deployed transformer models, you have a decent sense of what transformer-based language and vision models can and cannot do.

Why transformers? The transformer architecture is currently the foundation for State of the Art (SOTA) results on most deep learning benchmarks, and it doesn’t look like it’s going away soon. Much of the newest ML research involves transformers, so AI Safety organizations focused on prosaic AI alignment or conducting research on current models practically all focus on transformers for their research.

Goals‏‏‎ ‎

  • Play around with deployed transformer models and write up some things you notice about what they can and cannot do. See if you can get them to do unexpected or interesting behaviors.

  • Read and take notes about how transformers work.

  • Use 🤗 Transformers to import, load the pre-trained weights of, and fine-tune a transformer model on a standard NLP or CV task.

  • Implement basic transformer models like BERT or GPT-2 from scratch and test that they work by loading pre-trained weights and checking that they produce the same results as the reference model or generate interesting outputs.

Resources‏‏‎ ‎

  1. Experiment With Deployed Transformers (Choose 1-3)

    1. OpenAI Playground (GPT-3) - OpenAI

    2. Elicit: The AI Research Assistant—Ought

    3. DALL·E 2 - OpenAI, Stable Diffusion, or Craiyon (see The DALL·E 2 prompt book)

    4. Codex—OpenAI

  2. Study The Transformer Architecture (Choose 2-3)

    1. Attention Is All You Need—Vaswani et al. (Sections 1-3)

    2. Lectures 8, 9, and optionally 10 from CS224N—Stanford University

    3. The Illustrated Transformer—Jay Alammar

    4. Formal Algorithms for Transformers—DeepMind

    5. The Illustrated GPT-2 - Jay Alammar

    6. the transformer … “explained”? - nostalgebraist

    7. The Annotated Transformer—Harvard

  3. Using 🤗 Transformers (Choose 1-2)

    1. Hugging Face Course

    2. CS224U: Natural Language Understanding—Stanford University (Supervised Sentiment Analysis unit only)

  4. Implement Transformers From Scratch (Choose 1-2)

    1. MLAB-Transformers-From-Scratch—Redwood Research (refactored by Gabriel Mukobi)

    2. deep_learning_curriculum/​1-Transformers—Jacob Hilton

  5. Compare Your Code With Other Implementations

    1. BERT (Choose 1-3)

      1. pytorchic-bert/​models.py—dhlee347 (PyTorch)

      2. BERT—Google Research (TensorFlow)

      3. How to Code BERT Using PyTorch—neptune.ai (PyTorch)

      4. nlp-tutorial/​BERT.py—graykode (PyTorch)

      5. Transformer-Architectures-From-Scratch/​BERT.py - ShivamRajSharma (PyTorch)

    2. GPT-2 (Choose 1-3)

      1. Transformer-Architectures-From-Scratch/​GPT_2.py—ShivamRajSharma (PyTorch)

      2. gpt-2/​model.py—openai (TensorFlow)

      3. minGPT/​model.py—Andrej Karpathy (PyTorch)

      4. The Annotated GPT-2 - Aman Arora (PyTorch)

  6. Additional Resources

    1. Study Transformers More

      1. How to sample from language models—Ben Mann

      2. Neural Scaling Laws and GPT-3 - Jared Kaplan—OpenAI

      3. Transformer-Models-from-Scratch—Hongbin Chen (PyTorch)

    2. Other Transformer Models You Could Implement

      1. Original Encoder-Decoder Transformer (impl. Transformers from scratch—Peter Bloem)

      2. ViT, PERFORMER (impl. Transformer-Architectures-From-Scratch—ShivamRajSharma)

      3. CLIP: Connecting Text and Images—OpenAI (impl. openai/​CLIP)

Level 6: Reimplementing Papers

Objective‏‏‎ ‎

You can read a recently published AI research paper and efficiently implement the core technique they present to validate their results or build upon their research. You also have a good sense of the latest ML/​DL/​AI Safety research. You’re pretty damn employable now—if you haven’t started applying for Research Engineering jobs/​internships, consider getting on that!

Why papers? I talked with research scientists or engineers from most of the empirical AI Safety organizations (i.e. Redwood Research, Anthropic, Conjecture, Ought, CAIS, Encultured AI, DeepMind), and they all said that being able to read a recent ML/​AI research paper and efficiently implement it is both a signal of a strong engineering candidate and a good way to build useful skills for actual AI Safety work.

Goals‏‏‎ ‎

  • Learn how to efficiently read Computer Science research papers.

  • Learn tips on how to implement papers and learn efficiently by doing so.

  • Reimplement the key contribution and evaluate the key results of 5+ AI research papers in topics of your choosing.

Resources‏‏‎ ‎

  1. How to Read Computer Science Papers (Choose 1-3)

    1. How to Read a Paper—S. Keshav

    2. How to Read Research Papers: A Pragmatic Approach for ML Practitioners - NVIDIA

    3. Career Advice /​ Reading Research Papers—Stanford CS230: Deep Learning—Andrew Ng

  2. How to Implement Papers (Choose 2-4)

    1. Lessons Learned Reproducing a Deep Reinforcement Learning Paper—Amid Fish

    2. Advice on paper replication—Richard Ngo

    3. ML engineering for AI Safety & robustness: a Google Brain engineer’s guide to entering the field − 80,000 Hours

    4. A Recipe for Training Neural Networks—Andrej Karpathy

  3. Implement Papers (Choose 5+, look beyond these)

    1. General Lists

      1. Machine Learning Reading List—Ought

      2. Some fun machine learning engineering projects that I would think are cool—Buck Shlegeris

    2. Interpretability

      1. Thread: Circuits—Olah et al.

      2. A Survey on Neural Network Interpretability—Zhang et al.

      3. Post-hoc Interpretability for Neural NLP: A Survey—Madsen et al.

      4. Locating and Editing Factual Associations in GPT (ROME) - Meng et al.

    3. Robustness/​Anomaly Detection

      1. Agreement-on-the-Line: Predicting the Performance of Neural Networks under Distribution Shift—Baek et al.

    4. Value/​Preference Learning

      1. Deep Learning from Human Preferences—OpenAI

      2. Fine-Tuning Language Models from Human Preferences—OpenAI

    5. Reinforcement Learning

      1. Key Papers in Deep RL — OpenAI

Level 7: Original Experiments

Objective‏‏‎ ‎

You can now efficiently grasp the results of AI research papers and come up with novel research questions to ask as well as empirical ways to answer them. You might already have a job at an AI Safety organization and have picked up these skills as you got more Research Engineering experience. If you can generate and test these original experiments particularly well, you might consider Research Scientist roles, too. You might also want to apply for AI residencies or Ph.D. programs to explore some research directions further in a more structured academic setting.

Goals‏‏‎ ‎

  • Write an explanation of what research directions fit your tastes.

  • Create 5+ concrete research questions you might want to explore. These can be from lists like those below, the future research sections at the ends of ML papers, or your own brainstorming.

  • Conduct AI Safety research and publish or share your results.

Resources‏‏‎ ‎

  1. Research Advice

    1. Research Taste Exercises—Chris Olah

    2. How I Formed My Own Views About AI Safety—Neel Nanda

    3. An Opinionated Guide to ML Research—John Schulman

    4. Personal Rules of Productive Research - Eugene Vinitsky

  2. General Lists of Open Questions to Start Researching

    1. AI Safety Ideas—Apart Research

    2. Open Problems in AI X-Risk—CAIS

    3. Random, Assorted AI Safety Ideas—Evan Hubinger

    4. Some conceptual alignment research projects—Richard Ngo

  3. Open Questions in Interpretability

    1. Ten experiments in modularity, which we’d like you to run! - TheMcDouglas, Lucius Bushnaq, Avery

    2. A Mechanistic Interpretability Analysis of Grokking#Future Directions—Neel Nanda, Tom Lieberum

  4. Open Questions in Robustness/​Anomaly Detection

    1. Benchmark for successful concept extrapolation/​avoiding goal misgeneralization—AlignedAI

    2. Neural Trojan Attacks and How You Can Help—Sidney Hough

  5. Open Questions in Adversarial Training

    1. Final Project Guidelines- MLSS

Epilogue: Risks

Embarking on this quest brings with it a few risks. By keeping these in mind, you may be less likely to fail in these ways:

Capabilities

  • There is a real risk of novel AI Safety research advancing AI capabilities, reducing AI timelines, and giving us less time to solve alignment. If you go through this or other career guides, you may produce work that inadvertently leads to capabilities externalities.

  • That said, I expect most of the work from following this guide to be directly harmless, at least until you get to Level 6. Before then, it’s probably fine (and helpful motivation) to share what you are working on with others! After then, if you think you might have produced research that could be dangerous, err on the side of secrecy and ask for advice from a trusted and willing AI Safety researcher.

  • As a general guideline, consider building the habit of marking how hazardous your work is, evaluating other researchers’ work, and reading other AI Safety researchers’ evaluations of other AI work (e.g. in the comments of the AI Alignment Forum) in order to grow your sense of what various forms of dangerous work “smells” like.

Difficulty

  • Following this career guide alone will be extremely difficult, and there’s a high chance you could fail not because of your own lack of skill, but just because you don’t have the right support structures behind you.

  • To stay supported, find a mentor. This could be a willing AI Safety researcher or anyone else you trust who is skilled in machine learning with whom you could meet every so often to ask conceptual questions, get feedback, or discover new learning resources.

  • To stay supported, find coworkers. These could be friends who are also up-skilling in AI Safety with whom you could meet once or a few times per week to discuss things you are learning, collaborate on projects, or set accountable goals (e.g. “If I don’t read and take notes on this paper by next Monday, I owe you $20.”).

  • Also, keep in mind that this guide is kind of fake: breaking up the challenge of developing Research Engineering skills into a series of levels is especially reductive of the reality of learning. You probably shouldn’t do every suggested thing in a level once, move on to the next level, and never look back—rather, I think a better approach involves frequently revisiting previous levels, exploring other resources not listed here, and moving on from things if the marginal benefit you’d get from immediately spending extra time on those things doesn’t outweigh the opportunity costs of learning new things.

  • With all that said, this guide can be a way to test your fit. If you give programming or machine learning an honest effort and find you really don’t like any of it, then maybe Research Engineering just isn’t for you, and that’s okay! If you still think AI Safety is important, consider exploring AI Policy and Strategy.

Early Over-Specialization

  • Don’t be afraid to try new things! One failure mode I imagine here is diving deep into a small niche sub-field of AI Safety early in your journey and then either never making any useful contributions or getting bored and quitting. You actually might have been better at or derived more enjoyment from a different sub-field.

  • This could be important for getting jobs in AI Safety. Unless you get real lucky and happen to choose right, you might over-specialize in a narrow band of skills and lack knowledge in other domains that are important for the jobs you are interested in.

  • AI in general and even AI Safety in specific are very diverse fields with too many different things for any one person to specialize in, but you can still aim for a breadth of knowledge early on. When choosing projects in Levels 3-5, consider trying things you haven’t tried yet. It might feel scary, but breaking out of your comfort zone in this way is probably better for efficient learning and exploration.

  • Somewhat relatedly, AI moves super fast, and soon many of the resources here might become outdated or new promising research areas might emerge. I intend to keep this guide updated, but I encourage you to look beyond what’s here to explore new and interesting things in the future!

  • On a meta-level, it’s also possible to over-specialize in Research Engineering. Consider exploring other impactful career paths—many of the skills here could be really useful for Theoretical AI Alignment, AI Policy and Strategy, Information Security, and even Operations Management.

Late Over-Generalization

  • That said, you should eventually find some areas you particularly like and dig deeper into them (between Levels 6-7). I just think it’s worth exploring broadly before then so you can have a good taste of areas to choose from. Specialization is good; early over-specialization is bad.

  • This is also important if you intend to work in an AI Safety research lab: A common failure mode here is skimming across many different areas at a surface level but never diving into any enough to gain deeper insights and produce tangible results.

  • This failure mode seems to be pretty common with Ph.D. students, many of whom repeatedly hear from their advisors to make their projects even narrower in scope.

Sources

Here are some of the other great career guides and resources I used in the making of this. Most of the guides here also have good general advice that would be useful to read even if you don’t do the other things they suggest. Consider checking them out!

Many thanks to Jakub Nowak, Peter Chatain, Thomas Woodside, Erik Jenner, Jacy Reese Anthis, and Konstantin Pilz for review and suggestions!