Life as a multivariate optimisation game:

The older I get, the clearer it gets that I have scarcity of time and energy. Life is about figuring out how that allocate that energy to solve the multivariate optimisation game of life. There are a range of variables/levers one can allocate energy to: Health, Wealth, Learning, Exploration, Relationships, Pleasure, Impact, Status, Creation, Fame, Legacy. These variables can be positively or negatively correlated. There is also a key decision conflict in optimising for the short term versus the long term. Ask yourself, do I make this decision to optimise the quarter, the game, the season, history?

Choose the rules:

Understanding the relative importance of these variables is crucial. Think about this deeply and honestly. What is important to you, what kind of person do you want to be. I rank the variables on least to most important and also on long term to short term importance. Success often comes from trading short term pain for long term gain. These are the rules I am going to play by. These are the filters that decisions should be judged against. You can make the decision to be optimistic, to be positive, to be happy, to enjoy the little things. These are active choices. You define the rules.

Choose your game:

After choosing your priorities and rules its time to choose your game. Assuming you are going to spend a significant amount of your working ours working, especially in your younger years, picking the right game is key. It will determine your lifestyle, the availability of time and energy and influence your variables in meaningful ways. When you are young it makes sense to play a lot of games to gather more data. The game I have chosen to play so far in life is investing. It meets a lot of the principles that align with the priorities I had when I made the decision ten years ago. The lens I made the decision through was the ability to participate in the upside, constantly learn, improve. Seek out infinite games.

Choose your playing field:

After choosing your game it’s time to choose your playing field. Both physically and thematically. The place you stay matters. It determines the people around you, the climate, your tax situation, the healthcare, the education of your children, your safety and the standards you set in your life. For me, being in a big, fast paced city was the right choice when I made it. If you are an investor there are many playing fields one could choose. From public to private, across asset classes, across styles. The pond you fish in matters, and I arguably made a suboptimal decision in choosing my pond in the past. Be aware that you might be playing the right game but on the wrong pitch. Think about this deeply and honestly. Remember that your life is long. You can always switch.

Choose your bets:

Poker has taught me so much about life as it is a game of imperfect information, you have to make decisions under pressure, luck matters in the short term, managing your emotions is key. It also does not help you to live in the past, hold onto mistakes. Learn and move on. I like to frame decisions as bets. A bet is a calculated, informed, sized decision that you make based on probabilities derived from imperfect information. You are constantly making bets in your life. It could be the company you join, the investments you make, the books you read, skills you learn, the side project you work on, the friendships you keep, the hobbies you pursue. Fundamentally I believe that optimal bets to take have asymmetric risk reward and I believe in a bias towards action. This is also something where I also have been too passive, too timid in my life so far.

Key Principles:

Here are the key principles for the next decades of my life:

• Seek infinite games.

• Be clear what variables I am optimising for.

• Be aware of long-term consequences.

• Align my game with what I am optimising for.

• I can always change my playing field. Don’t be afraid.

• Make attractive, asymmetric bets.

• Be bold.

• It’s a game. Have fun.

Note:

This is purely my personal opinion. You are playing your own game. All love.

Welcome to the second edition of the Minds in Machines Weekly Roundup. Its fantastic to have you here. In this letter I explore where the physical world and artificial intelligence intersect. Minds in Machines. You can expect stories, news, research and random musings. Any feedback is appreciated.

In the Spotlight:

Helix does the dishes.

We have seen a lot of robots do our laundry as Chris Paxton pointed out it his article: Why is Everyone's Robot Folding Clothes?. In short: It’s repeatable, relatively easy, requires precision but not too much precision and is quite forgivable. Figure has upped the ante releasing its newest video of Helix loading a dishwasher with a range of plates, glasses and bowls. I would probably be in trouble for loading the dishwasher this shoddy, that aside, it’s extremely impressive as loading the dishwasher provides several challenges in comparison with folding clothes:

• Requires significantly more precision than folding clothes as items are fragile, different weight, shape, break and need to fit into specific slots.

• Requires more precise force control.

• The action space is far more complex and sequential dependencies create additional planning challenges.

It is encouraging to see the progress of humanoids towards more and more useful tasks and to see the same robot and model tackle two complex tasks.

What do you think will be the next task Helix will demonstrate?

Research that Caught my Attention:

Meet EMMA

Researchers at the Georgia Institute of Technology have published a new approach to tackle data scarcity and cost of mobile robot teleoperation: EMMA (Egocentric Mobile MAnipulation).

The core problem: Large-scale data collection using whole-body robot teleoperation is extremely expensive, resource intensive and requires tethering human operators to robots. This is a fundamental bottleneck for scaling data collection which is vital for improvements in robotics.

The core idea: Cut out mobile teleoperation completely by co-training on human full-body motion data (from egocentric video) paired together with static robot manipulation data.

The results: EMMA achieves comparable or superior performance to mobile ALOHA across three real world tasks (Grocery Shopping, Handover Wine, Table Service). This could represent an unlock in scaling data collection.

All credit: EMMA

Random Musing of the Week:

When listening to an episode of BG2 I came across the launch story of Xiaomi Auto. It is a remarkable story of that it takes to engineer greatness.

The day its founder Lei Jun decided to build a car, he swapped seats with his driver and obessively fell in love with cars. He drove over 170 cars himself. Whenever he saw a new one in the Xiamoi parking lot he would ask to borrow it to get a feeling for the car. He set himself the goal to become “the best driver among all the automotive company bosses in China”. Over 70 days he and his team visited 10 cities and talked to over 200 industry insiders. In 3 years he achieved what Apple failed to do. The SU7 sold 10,000 units in just 4 minutes.

Its an amazing inside in the mindset, the grid and passion it takes to build a great machine.

Funding News:

• Robotic farming startup Orchard AI has raised $22 million to improve data availability and quality in farming. Orchard offers camera systems, data management systems and is working on decision engines to empower farmers.

• Shenzen-based UBTECH raised $1 billion credit to drive manufacturing of humanoid robot. The company has over 2,700 patents and announced the first batch purchase for its intelligent humanoid robots.

• HappyRobot AI has raised $44 million to develops its AI operating system for the physical world. The company has won over 70 enterprise customers including electrification champion Schneider Electric.

Launches, Unleashed and spotted in the Wild:

• Innate Bot has launched MARS, a personal AI robot starting at $1,999 and shipping in Q4. What I find exciting, is the idea that you can train it yourself using your phone and share it with the community. Looks very cool, especially in Orange/Black. Check out the launch video here.

• Simulation software developer RoboDK has launched a free academy to learn basic robot programming skills. The course is self paced and allows users to learn the fundamentals of manipulation. The ambition is to become the global hub for robotic learning.

Thank you for reading. If you like this, please share with one person you think could get value from it. Means the world.

Keep constructing.

Dominic

Welcome to the inaugural edition of the Minds in Machines Weekly Roundup. Its fantastic to have you here. In this letter I explore where the physical world and artificial intelligence intersect. Minds in Machines. You can expect stories, news, research and random musings. Any feedback is appreciated.

In the Spotlight:

Brain update.

Image credit: (NVIDIA)

NVIDIA has launched its $3,499 developer kit Jetson Thor, powered by Infineon microcontrollors, sensors and actuator technology. It delivers 7.5x compute with a 3.5x the energy efficiency improvement over Orin. In raw numbers Jetson Thor delivers up to 2070 FP4 TFLOPS of AI compute and 128 GB of memory.

Early adopters include industry leaders Agility Robotics, Amazon Robotics, Boston Dynamics, Caterpillar, Figure, Hexagon, Medtronic and Meta, while 1X, John Deere, OpenAI and Physical Intelligence are evaluating Jetson Thor.

Out in the wild Galbot has already integrated Jetson Thor into its G1 humanoid robot. Professor Wang He, Founder & CTO stated: “Our G1 Premium, now running on NVIDIA Jetson Thor, delivers significant gains in speed and real-time reasoning, enabling our proprietary VLA models to achieve enhanced real-world performance”. Galbot recently won a Gold Medal at the World Humanoid Robot Games and has deployed its robors in 10 pharmacies.

You can get your hands on Jetson Thor at a hackathon organised by Seeed Studio and HuggingFace. Its going to be exciting to see what is about to be built with this new mind.

Research that Caught my Attention:

Researchers at Tsinghua University, together with industry partners have published a paper on Memory in Vision-Language-Action Models for Robotic Manipulation: MemoryVLA.

The core problem: Robotic manipulation is non-markovian and therefore temporal context matters. Current approaches ignore this and this causes issues. The image frame before, during and after pressing a button looks remarkably similar.

The core idea: Add a dual memory system to VLAs: short-term “working” memory tokens + a long-term Perceptual-Cognitive Memory Bank then condition a diffusion action head on what the robot retrieves.

The results: Dual-memory VLA hits 84% across 12 real-world tasks (+26 pp long horizon vs SOTA) and 71.9/72.7/96.5% on SimplerEnv-Bridge/Fractal/LIBERO-5 (including +14.6 pp on Bridge)

All credit: MemoryVLA

Random Musing of the Week:

I listened to an older podcast with Eben Upton, Founder of Raspberry Pi this week. Couple of things got me thinking:

• Features get added to the board through the following formula: (Cost of feature / Percent of people finding it useful). This means a $1 feature everyone finds useful gets same priority or 10c feature 10% of users find useful get same priority. This pushes lesser used features onto external hardware.

• The Raspberry Pi Zero is his favourite Pi because it explores a different application of Moores Law. Most companies (including RPI) apply Moores Law in the following way: Pick a price point and then fit ever more compute every 12-18 months. The Pi Zero picks an amount of compute (2000era PC) and sell it as cheap as possible.

• Incredible that this company was bootstrapped with just £100,000 investment.

Funding News:

Thematically, defence applications continues to draw funding both non-dilutive and venture.

• Unmanned Ship Startup Blue Water Autonomy raises $50 million Series A (Fortune)

• Reliable Robotics wins $17.4 million U.S. Air Force funding for autonomous aircrafts (Reliable Robotics)

• Reframe Systems raised $20 million to build affordable homes powered by robots achieving a 35% cost reduction (Businesswire)

Launches, Unleashed and spotted in the Wild:

• Robomart launches delivery robot with innovative $3 flat charging structure. The company is looking to roll-out in Austin later this year.

• Harvard scientists have demonstrated a smart, wearable robot helping patients with upper-limb disabilities.

• Researchers at the University of Surrey have developed new drones based on the agility of birds of prey to navigate urban environments.

Thank you for reading. If you like this, please share with one person you think could get value from it. Means the world.

Keep constructing.

Dominic