What is the Iliad Intensive?
Almost two months ago, Iliad announced the Iliad Intensive and Iliad Fellowship. Fellowships are a well-understood unit, but what is an intensive? This post explains this in more detail!
Comparison. The Iliad Intensive has similarities to ARENA, but focuses more on foundational AI alignment research instead of alignment research engineering. Expect more math and less coding.
Rhythm. It’s currently four weeks long. Five days a week. 10am till 6pm every day, with lunch and an afternoon break. This makes for around 6.5 hours of learning a day, which is at the upper end of how long most people can concentrate deeply within a day. This is why we call it “Intensive”.
Content. The Iliad Intensive is broken into five clusters, with 20 total modules, one for each day. The clusters and modules in the April iteration are below. We expect to add substantially more topics and material over the coming months. There is much more material than can be covered in a single month, so different Intensives will vary in content.
Alignment Cluster
AI Alignment: an Introduction
Alignment in Practice
AI Alignment: The Field
Learning Cluster
Deep Learning 1
Deep Learning 2
Singular Learning Theory
Training Dynamics
Data Attribution
Interpretability Cluster
Intro to ML Engineering
Mechanistic Interpretability
In-Context Learning and Belief State Geometry
Abstractions and Latents
Agency Cluster
Reinforcement Learning
Idealized Agency: Coherency and AIXI
Agent Foundations
Reward Learning theory
World Models
Safety Guarantees and their Limits
Debate
Steganography and Backdoors
Worst-case Interpretability and Heuristic Arguments
We will share the entire curriculum of the April iteration at the start of May, alongside a reflection of how the program went. In the meantime, here you can find a problem sheet that formed the theoretical component for the day on reinforcement learning. The students had the choice between working through this or working through ARENA’s RL intro day.
A typical day. We do not yet have a set daily structure. We may narrow down more in the future based on student feedback, but currently, we are experimenting with the following types of sessions:
Internal lectures and expert guest lectures;
Reading sessions: Students read a paper or blogpost;
Whole-class discussions and small-group discussions, with and without discussion prompts;
Math exercise sessions, alone and in pairs;
Coding sessions, alone and in pairs.
Our impression so far is that students like exercises and coding and, broadly, a variety of different activities on a day.
Students selection. We mainly look for mathematical expertise in students, which typically comes from having a degree in maths, physics, or theoretical computer science. We also look for research experience, general competence, and the motivation for pursuing our program.
Practical Logistics. The program currently runs in-person in London, and we consider running future programs also in the Bay area. We provide a fixed stipend of $5000, which can be used by the students to pay for travel and housing. We also provide office space, and lunch and dinner for five days per week.
The team. The Iliad Intensive is organized by Iliad, an umbrella organization for applied math in AI alignment that also runs a conference series, incubates a new Alignment journal, and has the ambition to incubate new AI alignment research bets similar to Timaeus and Simplex. The materials are created by a team of around 15 internal and external researchers who have domain expertise for the relevant module. We will list their contributions in detail once we release the materials.
Apply. If the above sounds appealing to you, please apply in this form! The June Intensive runs from Saturday June 6 until Friday July 3 (both inclusive) in London. The deadline for the June Intensive is Wednesday, April 22, EoD.
Details of the August Intensive are to be determined, but you can already apply for it in the common form.
Couple questions:
Kinda embarrassing but… what do I put in a CV? I’m a college student, that has no research experience, and hasn’t ever had a real job. The most impressive things I’ve done technically look like “read some books”.
About how many people will you accept? How high is the barrier to being accepted?
What if I don’t know if I’ll attend if accepted, and don’t know if I’ll even do alignment research down the road?
I assume you’ll want me to apply anyways, so the question about what to put in a CV is the most actionable
List the things you’ve learned and courses you have attended / are attending in your CV! If you read a technical book, list it too.
In the April edition we have a cohort of 16 people among 40 applicants. If we receive sufficiently many promising applications and our office space allows, we may have cohorts of up to 30 people, or maybe eventually even more down the road.
Part of the reason of writing this post is to encourage more relevant applications, so competitiveness in future rounds may be higher than in the April round, but that’s hard to predict.
If you have uncertainties, please still apply.
I am not a mathematics major but I have completed undergraduate courses in Calculus I, Calculus II and Probability and have self-studied Linear Algebra and Real Analysis. Would this level of mathematical background be considered sufficient? I am currently studying undergraduate mathematics independently and would appreciate guidance on what topics I should focus on next in preparation for the fellowship.
In the April cohort you would probably have gotten in with this knowledge, but I can’t tell you how our bar will change over time. I definitely recommend you to apply.
We will at the start of May publish the April version of the Intensive course materials, including prerequisites (globally for the program, and more specifically for each module). So look out to when we publish this to know what knowledge is useful to have upfront.