🧵 View Thread
🧵 Thread (53 tweets)
Sure, you can model it using some complex mathematics.https://t.co/hFZ9Gq3RdG
But have you considered just winging it with Taguchi?https://t.co/02VLQjqdJK
YOLO Bayes oneshot ultimape optimization of how to cook and egg? https://t.co/UNmcPkeQZp
@almostlikethat @literalbanana @nsreed @zefrank Basically get concrete data into a parameter space and keep iterating until you find ones that seem to match up with lived experiences. It's a really elaborate form of guess and check. It like a intuition driven Taguchi method of paramaeter space tuning. https://t.co/aG9wcgXRTn
Have you considered calculating the risk of not testing vs the cost of testing and seeing if there is a probability manifold on this hypersurface that lets you get useful data quickly?https://t.co/c1unmILiiH
Optimizing egg cooking is a thing I do, for reason I can't explain.https://t.co/umSZpIGtIT
Do you ever try to boil an egg using your couch & wonder if you could unboil an egg using your couch, and in the process set in motion a weird idea about using quantum time crystals as way to locally reverse entropy and make a universe ending quantum bomb?https://t.co/YCmusxcWFJ
Out of chaos, order.https://t.co/LsgIBynkiO
@goblinodds @lisatomic5 People confused bayes optimal decisions and maximizing utility with there being "one true good answer" when really the answer is often a shifting mix of Schelling points and chaos when looking at it from a metagame.https://t.co/XTI9tJe7Ed
Tapping the sign. Tappy tap tap.https://t.co/eSzJXeH6Iu
Stop and think about stopping and thinking.https://t.co/TbiUdE2epB
Did you know that ants stop to think about their path integration in a similar way as humans "resting" to update their mental state. My current mental theory is the movement system and thinking system use the same 'bayes' bits in the brain.https://t.co/0CBIeF2rMb
Ideas that it seems few can see because they are mixed messages held in a quantum eigenstate of utility.https://t.co/UGjEApvllg
I don't have pedigree or language to pass any qualifications to work for Facebook's AI dept even though all I think about some days is how to solve Contextual Bandit problem by localizing inference to speed up quorum based systems.https://t.co/SU2GM9rioJhttps://t.co/gTYcQKl006
Would you like to play a game?https://t.co/5khHHBuOZV
Thank you for coming to my ted talk.https://t.co/OwCqbGVUTr
I can't help it, I speak what I see. These memes and maths are metaphors of m3https://t.co/bQ9CpE9EU4
Jerry was a race car driver. 🎶https://t.co/v1QtVjhP5q
I'm mad that rats and slimemold are better at driving than I am.https://t.co/qslyhfqxBs
Hired: putting a rat in a tiny car to see if you can use them to navigate robot dogs using drone based telerobotic systems.I'm sure they don't realize this is what is going to be done with it. But hey, at least the rats are happy and get fruitloops.https://t.co/UpqHv2dDtv
Another cool one that is similar is the concept of "Grey Code" https://t.co/Tq4Uj0eE4K and "Karnaugh map" https://t.co/wV2nkzd0GBI learned these in college as ways to understand logic circuits. Grey codes are useful for running tests.https://t.co/MXFL1gdrfJ
With grey codes, you can pump them thru a circuit using something like JTAG and it does a 'good enough' job testing all pathways. Because only one bit changes between each test, you can more easily reason about failure by tracing the paths.
In a sense Taguchi method solves a sort of 'minimum distance' between tests that maximizes usable information. Kinda like Grey Codes.When you realize that, you can then minimize tests further by using a bayes style novelty search algorithm.https://t.co/yWpQW5VU4k
Once you realize you can frame these as complex manifolds that vibrate, you can start to do really fun things with them.https://t.co/iQU23XYUVl
Learning is a game you play against the world.https://t.co/b6uFZnsv6V
The fun thing is, you aren't learning everything from scratch every time to play the game. And you can learn from other's experiences too.https://t.co/YTR43046lw
Then you can start throwing in measure theory.https://t.co/XfokpL0LYm
The funny thing about playing battleship, is even a 'miss' can help you see the playing field better. The problem with using battleship under a model of Bayesian probability of shooting in the dark is that we don't bake in accuracy. What happens if you assume noisy samples?
Noting is measurable.https://t.co/Jc9c6R4j6w
Surfaces upon surfaces."each of those is a surface, and that surface meets another surface and they have been machined to be accurate"https://t.co/KfyRS6HgeZ
Measure together.https://t.co/fv17AYCOOm
Everybody's still doing Dykstra's and occasionally stumbling upon A* when I'm over here thinking about how a decentralized collaborative hierarchical D* could be expressed as a Speed-Accuracy Trade off applied to risk/reward and novelty search using bayesian probability streams.
We are a swarm of non-identical agents mapping the state space, act as such.https://t.co/26CtBlwSeO
Langauge allows us to communicate parts of the map we haven't personally seen. Sure the map is not the territory and nothing is truly measurable, but a really good map is better than a poorly drawn one.https://t.co/1sSUMtVZ7E
> "A Karnaugh map reduces the need for extensive calculations by taking advantage of humans' pattern-recognition capability. [...] Karnaugh maps are used to simplify real-world logic requirements so that they can be implemented using the minimal number of logic gates."
You don't need to do the algebraic math if you can imagine the shape.https://t.co/yua5kNqXNi
You can imagine the state space like a 3d map you haven't fully explored. Where would you go if you want to understand the lay of the landscape?https://t.co/X5pmNp6akZ
Have you tried uncovering how ritualized ad-lib practices in improv theatre are rooted in a long history of using chaotic words as a divination ritual, and using them to power what Daniel Dennett has described as "intuition pumps", so you can more likely solve the problem?
Pick a spot at random that is far away from where you've ever been before, and aim for either a mountain, or a valley.https://t.co/8kxpeOYhxu
If the problem space is large enough, you can expect others to fill in where you haven't explored before.https://t.co/1VYFSTpUMX
Even just random wanderings will do.https://t.co/qHxA8Gw8l0
I have no idea what I'm doing... but that's the whole point."Latent learning is simply animals observing their surroundings with no particular motivation to learn the geography of it; however, at a later date, they are able to exploit this knowledge when there is motivation"
We are a swarm mapping the state space, act as if.https://t.co/VeGgwi2Ns6
@krzhang @Plinz @ricard_sole "we propose a view of the adaptive immune system as a dynamic Bayesian machinery that updates its memory repertoire by balancing evidence from new pathogen encounters against past experience of infection to predict and prepare for future threats." 🥳https://t.co/f66LL8ZWP5
Squirrels are better at mapping a state space than most humans, but only because most humans don't even try.https://t.co/I85wjoFgdN
Like that fractal vice mapping a surface, it's swarms all the way down.https://t.co/B5JvM7wPMb
Wu Wei Mapping.https://t.co/NO8XSZo6pE https://t.co/poaTFU6Wc6
Basically, this, but with a much larger set of possible options and assuming that others have already tested some subset. The trick is in ways of exploring the unknown to maximize learning.https://t.co/yWgvKcUiDW https://t.co/5Gq8t6bYGj
One you realize you can map Orthogonal Arrays to linear amounts by pretending that the discretization effect is analgous to an integration methods, you can then start to reason about Taguchi Method as a form of Discrete Bayesian Inference. https://t.co/8laM7xlx37
Have you considered just NOT DOING MATH?https://t.co/JHDOyWsO96
If you want to know the volume of a complex object, just put water in it. Calculation is for nerds.https://t.co/Yl1U4XYyup
Pythagorean Inertia."The tablet not only contains the world’s oldest trigonometric table; it is also the only completely accurate trigonometric table, because of the very different Babylonian approach to arithmetic and geometry."https://t.co/ZRqSCBoFwNhttps://t.co/mzJ2V42yO7
Just because you can represent your ideas in math and fancy charts doesn't mean you're doing your science experiments correctly.https://t.co/r4CouCddgs
Literally, you can use water to prove things.https://t.co/wwP3ORmHxwThis beats geometric or algebraic methods.
The reason Karnaugh maps work to speed up thinking is it leverages our visual pattern finding system. Same idea.https://t.co/8HC3XuigIJ
I mean, maybe i'm kinda a nerd, but cooking eggs with math is kinda weird.https://t.co/LR8wf8hRbE
Heck, if you know what you are doing, you don't even need to cook eggs with water. Have you considered that you could automate the the 32 minute long 2 minute cycle by blowing hot and cold air on an egg? https://t.co/9foN1J9hWB
What if you want to optimize for hard boiled egg open-ability and maximize PEA because your GF's hands don't work well and you want to heal her?https://t.co/mw8zkox4xo https://t.co/Ide6vqD002
What about optimizing for ease of cooking and cost of energy per batch, so you can maximize egg per day under constraints of solar panels?How does a waterless air fryer compare to a sous-vide or a pressure cooker?Suddenly you have a lot more variables. What should you do?
OTOH, if you are exploring a frontier and experimenting with your health, it is far faster to explore multiple variables at once using the Taguchi method.Get the basics down so you have a foundation to explore from. https://t.co/nUBofItKds