Multiverse School Newsletter - 01/28

Hey Multiverse Crew,

Last week was a whirlwind! Can you believe next week is still technically January? We kicked off 2 classes, Complex Systems and Machine Learning. Congrats to everyone who participated in the first filming of the Complex Systems and Game Theory Podcast! For those of you on the mailing list, you can watch the unedited, first-run preview here. This might not end up being released publicly ever, so you get to see something no one else has ever seen. Also, our Website has been updated!

Job Search Standup

Don't forget to join the Job Search Standup every day at 9:00 AM. It's a crucial space to focus on job hunting or lead-gen activities. $20 is the weekly rate, which is just $5 a day. Multiverse students have it included as part of their tuition. So far, 50% of the 32 respondents to our survey got jobs! Hooray for y’all! We didn’t throw out any data (we got 32 respondents), but didn’t include anyone we didn’t hear back from. To be clear, this is a much lower rate than my average- probably my lowest since beginning teaching people to get jobs- I’m willing to chalk some of it up to the market, but I’d love your feedback on what might help you so please drop me an email if you have any thoughts about what’s not working for you. (last week the link was broken)

Interviews are Back!

We're interviewing for the February and March cohorts. We do an onboarding session once a month, typically aiming for between 4-8 people per cohort, so as not to dilute the community we've built up. You should have received an email if you sent in an application, if not email tag@themultiverse.school to set up an interview.

Machine Learning

This week is the week we do the thing you wanted to be able to do with LLMs! We're training our own Mixtral LoRA on the OpenAssistant data. Make sure to keep up with the video assignments in the curriculum- those are how you walk away with a grasp on the math, when we're together, we'll focus mostly on getting lab work done.

Artificial Intelligence and Agents

This week we'll refocus on core elements of Prompt Engineering, for the crafting of agents who problem-solve.

Complex Adaptive Systems and Game Theory

If you're on the Complex Adaptive Systems and Game Theory podcast, or if you missed it and want to catch up, take a look here! We'll be reading through to the end of Part 1, or Page 27, of Complex Adaptive Systems: An Introduction to Computational Models of Social Life, and Section 2: Basic Elements and Assumptions of Game Theory in Stanford's Game Theory Course in order to catch up and be able to participate in the discussion on Tuesday.

Learn to Code

And for those just starting out, our regular Learn to Code Together session on January 25th at 5:00 PM will focus on mob programming for beginners. You will be asked to code in front of other people. It won’t just be you asked to do it though. We’ll all share our screens.

Storytelling - Detailed Character Profiles

Join us on Monday, January 29nd at 6:00 PM Pacific for a deeper dive on building character profiles, and what you really need in order to develop a plausible, sympathetic character. Francina Simone will be as usual running the program- this should help you if you're interested in writing your own character profile, playing well in TTRPGs, creating video game characters, or writing a novel or short story.

CyberSecurity - Let The War Games Begin

We're currently running an ai-enhanced wargame. See Liz to join a team- which side do you want, Blue Team (defenders) or Red Team (hackers)? This game will run all year and we'll productize, publish, or perish? at the end (and throughout). Probably only two of the three. Probably. Both teams meet together on Monday to plan a sprint on a specific layer of security, along with a friendly competitive dynamic and some collaborative efforts towards publishing this research incrementally as we make strides.

Operational Automation for Community Organizers

Tag is back teaching Operational Automation for Community Organizers again this week! There's about a trillion things you can do with no-code editors, so join us for more automations to make your life easier.

Junior Devs:

This week we're developing models for your own clicker game- on Friday we made a spreadsheet for Clicky Swamp. Clicky Swamp is a Complex Adaptive System, so this ties right into Complex Adaptive Systems! This model will drive a redesign of our whole system, which will allow us to write a much faster, well-thought-out second iteration of the project based on having developed this data model. If you missed Friday, we'll briefly re-cover the spreadsheet.

Senior Devs:

Let's do Generative Algorithms, and generate texts that need to be parsed as an abstract syntax tree, particularly relevant to your active research projects. We can also set up a series of agents behind FastAPIs, and deploy them to Google App Engine, so we have a series of endpoints up that agents can interface with.

This will help you understand when creating a LoRA vs a few-shot prompt will be relevant. If you have an active research project and it doesn't have some junior developers involved, talk to me about it.

Important Links

Apply for the Multiverse School

Tell us if you got a job!

Cool Videos

AI Tutorials and News

"Self Trained" language models are somewhat like self-supervised but slightly different - we will likely be getting into these in class

More on Self-Training, and how Synthetic Data can work

AlphaGeometry is pretty cool

David Shapiro is putting out a call to research Autonomy, which not that many folks are doing (but we are)

NVIDIA's Foundation Agent looks like Skynet

Cybernetics Principles in LLM discussions

What is a Convolution?

More leaderboard claims by Gemini

PalWorld will likely be a Landmark Case

Psychology & Critical Thinking

Most indigenous folks have been saying this about plants for a long time but it's cool that the western world is catching up. Keep at it, maybe if you study hard one day the rest of the world will take you seriously.

Every Psychological Effect, in 12 Minutes

Hierarchy is Fragile, Anarchy is AntiFragile

You might not have Free Will in a technical sense but you do have Agency and Mechanisms of Control

Meaning is the best defense against Nihilism

Economists tend to have the best bead on threats in my experience, because they do the accounting for damages

How to build a Solarpunk Future - there is no single solution, the only thing that really scales is mindset shift. Demand more ecological integration and less externalization to the environment, bit by bit.