Financial Robotics
Care Bear stare.
Financial markets are tricky little muses. They draw you in with the promise of one golden ring to rule them all, then prove to you over and over again you are not, in fact, smarter than the average bear.
Last night, I watched a documentary that, surprisingly, has racked up 31 million views about some folks in finance that did pretty well based on a simple idea. Namely, that stock picking is for losers. Instead, invest in the index. That’s where index funds, and mutual funds were developed. Putting your money into a broad range of equities, like the S&P500, or even better, a small cap index, outperformed picking a few number of stocks.
In fact, it was so important and profitable, that the people who developed it went on to win 6 nobel prizes and changed the name of the business school after the guy who started it all (notwithstanding, he donated $300M to the school).
These were the chaps who invented the whole darn space. Things like the efficient market hypothesis, the Black-Scholes equation to price stock options, you name it. And it all happened in a small midwestern school called the University of Chicago. It’s one of the reasons I went there.
The key change that enabled it all was simple: collect lots of data, then use computers and human ingenuity to figure out the truth. As one midwestern interviewee said, “Data doesn’t lie”.
Crazy.
Robotics is at a similar precipice. While LLMs sucked up all the internet’s data, which is mostly text, some images, and a bit of video and audio, there isn’t very much of it representing robotics data.
For example, the data for how an arm, hand, and fingers move to pick up an object. Or how legs, feet, and toes can maneuver through rocky terrain while running and changing direction, without falling on your face.
The scale of that data simply doesn’t exist.
So now you have startups, AI labs, and Big Tech companies talking about world models, simulators, and virtual gyms, to create, collect, and train robotic systems with small bits of real data, and larger bits of synthetic data.
There is a problem, however, especially in light of our analogy from financial markets above.
While financial markets represent the truth about reality that the world is engaging in 24x7x365, we don’t have the same level of streaming truth data for millions or billions of robots moving around the real world in real time to use. Sure, we have people holding phones, some wearing smart glasses with cameras, and cars with sensors on them, but that’s like describing an elephant by only experience one part of its body (i.e., the trunk, tail, legs, torso, etc).
In addition, most practitioners in the space don’t truly believe that LLMs are “all you need” to drive something as complex as a make-shift biologic system.
Instead of trying to recreate the real world in a physics simulation engine, with all its Higgs fields, quantum entangled particles, gravity waves, dark energy, bosons and supernova, perhaps the better approach is to just put the thing in the real world and start collecting real world information as the robots try to engage with it.
Of course, this takes us back to our ongoing conversations on Strong Core Values.
First, what are we using the robots for? War? Or food production? Helping with housework, or dive-bombing drone swarms. Taking jobs, or becoming another member of the family?
Who’s collecting the data and making the robots? Geppetto or Stark Industries? We tried to help Amazon years ago over a Christmas holiday understand this point.
In the Jetpack Universe, our vision is much closer to Bicentennial Man (RIP, Robin).
This stuff matters.
The values you encode into the algorithm as well as the industrial design of the hardware. The materials you use. The reason you collect the data. The insights and truths you try to gain from the data you’re collecting.
Does it help, or does it hurt? Who does it help? By helping one group, does it hurt another? If so, you’ve taken a wrong turn somewhere.
Consider the reason you’re doing this in the first place. Is it seeking attention, power, money, insight? Or just the love of pure creation, stemming from a gentle soul. The Great Filter is upon us, dear friends.
Do you do it because you want a friend, or want a slave to do your bidding?
Friends don’t make friends do stuff for them. Friends do things for their friend because they care about one another.
Care. Now there’s an interesting word.
Show me the data you’ve collected that has Care at its Core.
And then we can build the efficient market, and world, we’ve all been dreaming of.
—Sean
