followmeon

> But there's no computation happening to it. It's static.

There are three possible answers to this: The binary string itself is a form of computation, with 1's being upwards fluctuation and 0's being downwards fluctuations. Something like a Boltzmann Brain. Second, Zuse-style, the universe itself could be a computer: A Turing-complete cellular automata, acting on the input formed by a random process. Third, our universe may be mental and the binary forms mental states.

> This is true whether or not a human is there to observe it and count them.

This may be true, only because human brains made it true. Objects, like apples, or apple trees, are somewhat arbitrary selections of parts of the universe (The human chose to wall off a certain subset of everything).

With a little imagination you could see the apple on the ground and the tree it fell from as one object. Why is a hair on my head part of the object "me", but the moment I pull it out, it becomes its own object? What if the apple on the ground rots away and turns to mush? Is it still its own countable object? What if a cow eats the four apples? Is there now no object, one object, or still four objects?

Your view is much closer to reasonable than mine though, because to prove that maths is a mental construct, I have to invoke scientific unpopular views, like intuitionism (maths is an advanced form of Wittgenstein's language games) or Indian/Asian mysticism (make me one with everything).

But the reason that intuitionism is unpopular, is not because we found it to be objectively untrue, but because of the very human element of Hilbert hating Brouwer's guts. If Brouwer had more clout then we may see the Law of excluded middle as untrue or unprovable, as it stands, we entered Hilbert's timeline where this law is deemed objectively true even outside of minds.

Then again, my view has no problem with "Spooky action at a distance", because I am not required to view two electrons as separate objects, and distance between arbitrary subsets is relative (owes its value) to the chosen metric, location and measurement equipment of observers.

If the sun rises I can capture this with maths and prove to you that the sun is not going down. But I forgot to add "the sun rises to me", I forgot the creative power of observation: Because for someone on the other side of earth the sun is going down. For someone in space the sun may not appear to move at all.

If our reality consists of mathematical programs and the multiverse theory holds, then the theory of everything is a simple incrementing binary counter. Every possible binary string is ran on a Turing machine in a different universe.

The halting problem (endless loops) is sidestepped by being contained to a single universe. Also, apparently our universe got lucky as we get to live in an ordered universe, and not dissolve into nothingness due to a bug in our universe's binary code.

Of course, like "normal" mathematics, "reality" mathematics should also be incomplete. Furthermore, there is no way to know everything of the universe one is a part of, as this requires at least 1 bit more than there are available bits in the universe. But this still allows for the possibility of the universe itself being an inference machine.

Edit: Then again, I believe that mathematics is a mental construct, mathematics does not exist without brains, or if it really exists external to human minds, it exists in all possible forms: Humans do not discover maths as a whole, they cast a local net of subjective observation over it, and keep the things that make sense to our brains and are consistent within the current framework. Again, this does not preclude the possibility that our mathematical reality exists inside the mind of a giant supercomputer.

Dirty data is not as much as a problem for me than human-biased data. Dirty data engineering, like modeling, will soon be largely automated.

Let's say you are predicting store sales. You create a feature that holds the store sales of one year back. The feature works really well and you are happy with your evaluation. But you captured bias: The previous model the store used was "predict today's sales by looking at last year's sales". Store managers fitted their sales tactics to this model (when the model predicted too much sales, the store managers do their best to get rid of the surplus inventory, for instance: by adding discounts or moving the products to a more prominent spot in the store).

So in the end you end up with a model with good evaluation, but you actually have (over)fitted to previous policies/models. You have not created the best possible sales predictor. How to ever find this out, without a costly intimate deep-dive in the data and data generation processes?

Next it'd be interesting to see Python 2k vs. Python 3+. My own experience tells me that the majority of top Kagglers still use Python 2k, despite Kaggle Kernels being Python 3+ exclusively.

I also am quite amazed with the predominant use of Logistic Regression. I wonder if that is less about interpretability / ease of engineering, and more about the barriers that data scientists face when using more complex methods: lack of data science talent, lack of management support, results not used by decision makers, limitations of tools.

If Kaggle results are anything to go by, all businesses that care about best performance on structured data, should be using a form of gradient boosting.