🌞 Earlywork #48: The Simulation Arms Race
Featuring a piece by Daniel Han-Chen (Co-Founder @ Moonshot) + an exclusive offer w/ Linktree + roles from Folklore Ventures & Tutero + an interview w/ Marissa Payne (Co-Founder @ Matriarch Digital)
Ello ello Earlyworkers!
Traversing the metaverse to get to your eyes today is Earlywork #48, a careers newsletter providing free weekly career resources, news, jobs & real-world career stories for 2000+ young Aussies & Kiwis in the tech, startup & social impact landscape.
If you’re not already part of the crew, subscribe now to keep a pulse on our latest stories and conversations:
💡Weekly Cheeky Tip
We are witnessing the rise of the simulation industry, as the world’s leading tech companies and governments race to simulate the world around us.
But how the heck is this actually happening? Can it even happen?
I was lucky enough to chat to the deeply insightful Daniel Han-Chen - Founder @ Moonshot, an early-stage startup creating faster AI & machine learning algorithms to simulate stock markets, climate systems, and human lifespans -, who gives a remarkable rundown of the space.
🤿 Let’s dive in…
NVIDIA’s Earth-2 supercomputer will predict climate change accurately by 2060.
Google wants to fuse multiple AI models into one single system called Pathways.
China wants to model physics, the human body, then the whole universe.
And my mission with my brother at Moonshot is to create a public world simulation to predict everything!
💻 The simulation industry? What’s that?
How does Google Maps predict traffic congestion? How do hedge funds predict stocks and market prices? How do we predict air pollution or the weather? Each of these problems can be solved by creating a simulation to mimic the system!
🤔 But how?
Engineers use past data to predict the future! Previously scientists handcrafted equations and formulas to model natural phenomena and systems.
However, with the rise of big data, we can feed historical data into machine learning or artificial intelligence models and let the computer automatically handcraft the equations!
👾 Who’s playing the game?
The ultimate goal of companies is to feed ALL the world’s data (weather, stocks, diseases, the news, EVERYTHING) and predict the world’s future.
Google, NVIDIA, Microsoft, OpenAI, DeepMind, China, all want to achieve this, but in different ways.
Google acquired DeepMind for $600M in 2014. You might remember DeepMind beating Lee Sedol in Go a few years back. It was hailed as one of humanity’s greatest achievements. DeepMind has now gone to model the weather for the UK, solve DNA folding, reduce Google’s data centre costs and more!
Google themselves have helped model floods in India and Bangladesh, and showcased a new model called Pathways which will combine multiple models into one large model in an efficient manner to perform millions of tasks.
Elon Musk, Sam Altman and others pledged $1B to make friendly AI at OpenAI. OpenAI’s GPT3 was a breakthrough model which can mimic human language! OpenAI’s Five managed to beat Dota 2 players, and further received a $1B investment from Microsoft!
NVIDIA’s Omniverse aims to create digital twins of any system - a factory, a chemical plant, and even the whole Earth. Think of it like Google docs for 3D.
NVIDIA’s Earth 2 supercomputer will predict climate change accurately by 2060, and will be used to model climate change impacts and how to mitigate them!
🌏 Is it possible to simulate the world?
Imagine you had all the world’s data - weather, stocks, traffic, diseases, everything!
We feed this data into a model to predict the future. It should be possible right?
The Butterfly Effect says a butterfly flap can unexpectedly cause a hurricane. So a wrong prediction will affect the final result dramatically.
The 2021 Nobel Prize in Physics was awarded to those who understood this, yet dared to push on.
The key trick is to simulate averages of quantities, like the average temperature in 100 years, the average stock price in 30 days etc.
📊 How do we process all the world’s data?
Fast algorithms are the key! Current algorithms are very slow. For example, TSNE, a data visualization algorithm, takes 3 hours on CPUs!
By using some clever math tricks and a GPU, I made TSNE run in 5 seconds for NVIDIA!
Linear regression draws a straight line in high dimensional space, and previously used 700MB of memory - it’s now 10MB (99% less) and is 95% faster with it being featured in University of Washington‘s research!
Not all algorithms can be magically made faster. NVIDIA shows how they achieved a 1,000,000x speedup due to GPUs + Scaling + Machine Learning.
We want to advance this further, by focusing not only on hardware, but software.
That’s why at Moonshot, our aim is to make AI algorithms at least 50% faster. NVIDIA spent $100M USD on the Cambridge-1 supercomputer to help AstraZeneca find new drugs.
With 50% faster algorithms, the supercomputer essentially becomes a $200M one.
🤝 How can you get involved?
Many global companies like Google, Microsoft, NVIDIA are scrambling to hire artificial intelligence researchers, machine learning engineers and software engineers who can create simulations from huge amounts of data.
For example, OpenAI’s Ilya Sutskever coauthored AlexNet - the computer vision model which kickstarted the AI revolution 2.0, and coauthored AlphaGo’s paper to solve Go. He got scouted from Google with OpenAI paying him $1.9M in 2016.
Maths, computer science, algorithms, and some knowledge from all disciplines is key to creating huge simulation models. My favourite Youtube recommendations include MIT’s 6.034 Artificial Intelligence, Stanford’s CS229 Machine Learning, and Stanford’s CS231N Computer Vision courses. For maths and algorithms beginners, MIT’s 6.006 Intro to Algos, MIT’s 18.06 Linear Algebra, and MIT’s 6.041 Statistics are paramount!
There are so many problems anyone can solve if you have some data! Can you predict when the market will next crash? How about modelling COVID cases or deaths?
🏎 The simulation arms race has begun
At Moonshot, we make algorithms 50-95% faster for NVIDIA, Facebook, Microsoft and more to power a world simulation to predict the future of climate change, market crashes and more!
We got contacted by Australia’s DOD and received an OpenAI community award. I previously helped the Australian government predict Zika Virus outbreaks, traffic congestion and air pollution and was at NVIDIA making algorithms faster!
The US, UK, China and now Australia are racing to simulate the world. Whoever wins will have planetary scale power.
And we’re hoping to defeat big corporations and governments, and give access to our world simulation to everyone!
But we can’t do it alone! We need all the help we can get - so join us, and together, let’s create a world simulation! Join our waitlist for the 2022 launch here!
🌏 Earlywork Community
🐝 The Buzz
What’s been happening in the Earlywork community?
Jack Leung from Hive Empire Capital is running a free Earlywork Masterclass on GameFi, where we’re now seeing the convergence of the creator economy & “new” finance that lets people play games and make an income. Register here.
Community member Raphael Sebban recently launched the Bricks and Stones newsletter on the future of real estate
📚 Trending Topics
Our favourite reads and resources being discussed in the Earlywork community.
Think & Grow released their 2021/2022 Australian Startup Salary Guide
Blackbird, AirTree & Square Peg target major fund raises in 2022
Melbourne-based mobile ordering platform Mr Yum lands $65M led by Tiger Global
Jessy Wu (AfterWork Ventures) wrote a killer post challenging the generic usage of the term ‘diversity’ in hiring
Warwick Donald (Tractor Ventures) wrote a great roundup on alternatives to traditional venture funding
🌲 Ecosystem Exclusive: Linktree PRO
Building up your professional network in the early years of your career can be hard work – but there's one part of it that will always stay easy.
If you know about Linktree, you already know that it offers one link with a whole lot of power: a central source of truth about who you are and what you care about.
Already a fave of young startup founders, creators and professionals in tech and social impact all over the world, Linktree PRO adds enhanced features like subscriptions, analytics and a growing ecosystem of exciting integrations (a.k.a not just a list of links!).
It's all in the name of bringing together the different pieces of your digital identity to help your online presence work smarter (not just harder) for you.
For Earlyworkers, I'm excited to offer six months free in Linktree PRO (usually $6 a month). Whether you're a new or existing user, all you have to do is use the promo code LINKTREEXEARLYWORK to redeem.
- Maddie, Strategy & Ops Graduate at Linktree.
😎💼 Top Gigs
🎨 Graphic Designer (Part-Time) @ Folklore Ventures (Sydney / Remote)
Folklore are looking for a part-time designer to continue to craft and evolve Folklore’s visual identity as a long-range ambition machine for the great success stories of our time.
You will be reporting to their Head of Communications and will also work closely with their Growth Ops Manager.
💪 Growth @ Tutero (Melbourne / Remote)
Tutero are hiring a sales/marketing superstar who can come in to build systems, develop campaigns, sell solutions and scale teams!
You will be one of the first Australian hires for an EdTech startup, ready to take on the world 🌎 🚀
Check these roles out and apply at 👉 earlywork.co/gigs
If you’re hiring for early-career roles in tech, startups & social impact, you can share your role on our Gigs board for free today:
Disclaimer: Earlywork received monetary compensation for the roles featured in this section. If you’re keen to have your roles featured in our newsletter & social media, check out our premium listing options here.
1️⃣ 🕐 💪 One Minute Hustle
We are back once again with One Minute Hustle, a bite-sized interview with an emerging Australian young startup founder or operator.
This week, let’s get inside the noggin of Marissa Payne, Co-Founder @ Matriarch Digital, a young founder helping women-led & focused businesses get in the spotlight…
You can read about what she’s working on, how she got started, and why she does what she does here.
🍵 Vibed with this piece? Buy us a cheeky herbal tea!
😍 Want more FREE career resources? Be sure to check out the Earlywork Resource Hub for even more free resources on landing a role in tech/startups.
🤔 Got career questions? Join the Earlywork Slack Community to connect with the next generation of founders & operators across tech, startups & social impact.
Ciao for now,
Team Earlywork (Dan, Jono & Marina)
Email | LinkedIn | Twitter | Instagram
If you liked this post from Earlywork, why not share it?
Thanks again for having me collaborate on the Earlywork newsletter piece!!! It was super fun working with the Earlywork team!!! Keep up the fabulous work!!