July

@july

1755 Following

127978 Followers

July pfp

So: - If it is rewarding, do it -> Someone will game the system - Game the system -> Follow the gradient - Gradient reaches the local minima - > collapse the system because proxy minima reached - The optimizer figured out the 'surrogate intent' even if the designer's true intent was something else Then: - Essentially tokenomics or economics or even AI alignment in general works (unless value extraction before system collapse is your objective) only works when the 'intent' is opaque or does not have a uniquely global minima - You need to setup a non-convex problem, where the intent is explicitly unlearnable So the goal should be: - Don't make the intent fully learnable, ever

3 replies

2 recasts

12 reactions

July pfp

I mean it makes sense when you start thinking about lasting human institutions (religions, cultures, legal systems etc) rely so much on norms, rituals, and meta-structures rather than purely economic or technical incentives Specifically because it is non-convex, and the intent is explicitly unlearnable

2 replies

0 recast

7 reactions

July pfp

Gaming the system (overfitting to the incomplete, imperfect, or proxy signals that the system exposes as rules) collapses the system because you're optimizing a problem that has a small local minima -- or maybe the way to think about it is, gaming the system is 'overfitting'. Because a robust system requires a much better understanding of the intent of the problem. It's about making sure that the optimizer cannot ever find a profitable strategy without also implicitly learning the designer's true objective, i.e., the intent I feel like this is why AI alignment, or markets are hard

2 replies

1 recast

9 reactions

July pfp

Sometimes when the world is quiet I think about how BJ Penn's entry song was Hawai'i '78

1 reply

0 recast

9 reactions

July pfp

Steps are the bicep curls of the mind

1 reply

0 recast

29 reactions

July pfp

Thinking about the problem beyond the surface level, second order effects, subtext, changing of the frame of reference, how it affects others, changing your frame of reference, disagreeing with your hypothesis, disagreeing with their hypothesis - while being open to any outcome Take any problem - you’re probably doing more than what 90%+ of what most folks would do

3 replies

0 recast

25 reactions

July pfp

Taking a walk clears the mind

7 replies

4 recasts

51 reactions

July pfp

Imagine not thinking the world is amazing I can’t

8 replies

3 recasts

76 reactions

July pfp

Pools are just domesticated oceans and lakes

5 replies

1 recast

27 reactions

July pfp

Also Phalanx CIWS has existed for ages. This is maybe a cheaper version https://en.wikipedia.org/wiki/Phalanx_CIWS#:~:text=The%20basis%20of%20the%20system,from%20the%20Star%20Wars%20films.

1 reply

0 recast

5 reactions

July pfp

some thoughts - honestly this whole "rethought the drone thing" is complete BS - this is a known form factor - as with vehicle / dynamics / control design -- its all design trade offs - what is your mission and what are you maximizing for? sure you get: - benefits of with coaxial rotors, you have a shorter moment arm for pitch, roll, yaw - better thrust to weight per area - but that means you need a weirder more precise (is my guess) control strategy - multirotor drones have well known control strategies - control-tuning standpoint means you gotta do your own thing which is cool but just less proven, means more testing - lack of redundancy - if you have more rotors, it means better redundancy. with 2 coaxial rotors, you lost one rotor you lost 50% thrust - not the best again - it's all A) what is your mission B) what tradeoffs are you making C) there is no single bullet that solves all missions

4 replies

1 recast

22 reactions

July pfp

If you think of it in this frame of mind, and extend it further: - Engineering is the art of lying to reality to make it solvable - Art is the structure to lie about reality even when it’s unsolvable

1 reply

0 recast

8 reactions

July pfp

Honestly I think this is fundamentally the essence of engineering in general. You’re constantly trading fidelity, optimality, and expressivity — for tractability and that’s wild — because it works Essentially you are trading what is ontologically true for solvability, you’re lying about reality so you can solve it - it’s sort of the opposite of art

3 replies

0 recast

22 reactions

July pfp

Here’s the deal: - Reality isn't linear but systems behave linearly around operating points - it’s turtles all the way down and the turtles all have Jacobians on their backs - Convexity is a hammer, everything is a nail - Linear Transformations Rule Everything Around Me - Everything is a convexity problem if you squint hard enough - Yes, everything

1 reply

0 recast

30 reactions

July pfp

no one: absolutely no one: applied mathematicians: maybe... its a linear map?

1 reply

0 recast

16 reactions

July pfp

No one: Mathematicians: smiling politely *under their breath* heathens

0 reply

1 recast

6 reactions

July pfp

0 reply

1 recast

12 reactions

July pfp

Many things are really hard to actually do If it's easy -- you are probably not doing it

7 replies

3 recasts

63 reactions

July pfp

Hello I would like a flying blender pls

8 replies

3 recasts

54 reactions

July pfp

The audacity it takes to throw yourself onto the big green grass and roll around recklessly because you love life and it’s golden hour and sun is smiling softly before going to bed absolutely does not only have to be reserved for children

6 replies

6 recasts

64 reactions