Other than machine learning, what other technology produce results that are considered black box magic?

Other than machine learning, what other technology produce results that are considered black box magic?

Other urls found in this thread:

youtube.com/watch?v=t169yNXX4oU
twitter.com/SFWRedditVideos

I mean stuff like CRISPR is pretty much like that.

How so? Isn't it just an easier way to edit genes?

Yeah but imagine what happens if you figure out how to create a human that can think on a plane hundreds of times higher than others.
Imagine suddenly being able to craft an organism for any situation.

Go to bed, Rick.

ICO's. you make a digital currency and get millions of dollars

xD

ECKS DEE

Idiot brainlet

Nanomachines, son.

A sufficiently long regular expression.

ML isn't really black box magic, neural networks are.

The most relevant, state-of-the-art ML stuff are built on neural networks, deep learning for example.

That's what I said. ML != neural networks. There are random forests, decision trees, support vector machines and other algorithms used in ML, which are not black box magic.

>mfw these summerfags starting threads about ML think K nearest neighbor, beam search, and boosting are "black box magic"

I'd probably throw genetic algorithms in there as well, but nobody cares about those since recurrent and convoluted neural networks are the hotness right now in data science.

Brainlet here. Why is machine learning a hot topic now? Why couldn't they use algorithms instead of predicting an outcome with statistics?

Predicting an outcome with statistic is using an algorithm. A machine learning algorithm. Logistic regression is probably the most simple one, look it up.

Because theres a lot of things where a rule-based algorithm would be too difficult/impossible to make, instead, you just research a general purpose algorithm that reads training data, "learns" it structure and patterns and apply it to new data, without the need for you to code the pattern detection yourself.

Example: You want to make an algorithm that tells you if an image is a cat or a dog. It would be very complex to make a conventional rule-based algorithm for that, since you would have to deal with pixel-by-pixel shit. Instead, with machine learning you just feed a lot of images correctly labeled cat or dog and the algorithm figures out by itself how to classify new, unlabeled images.

Thats why ML is getting huge nowadays, with giant datasets and computational power we can make a lot of shit that is just not viable coding manually.

Take a look at this youtube.com/watch?v=t169yNXX4oU

Thought you said 'craft an orgasm for any situation' and I was all in.

We finally hit the point in computing that you can brute force statistical maps for input/output fields w/o city sized computing equipment. It's all the rage because its a convergent algorithm whose brute forcing scales.

> Brainlets finally found a way to mimic intelligence without having to do the hard work required to understand it.

Caveat being when you expose such systems to data sets and output requirements they've never seen. Then the whole thing falls apart which is why it was later classified as weak/brittle AI.

None of this is black box magic as it just straight forward brute force statistics. The real black magic is coming.

>considered black box magic?
>considered
Please leave and never come back.

I'm still amazed at how well simanneal performs even with decently complicated "step" and "evaluation" procedures
I remember using it for finding out the most distant chromapoint from a given one in CIELab with CIE94 ΔE, mostly because color distance is a fuckhuge function and nobody wants to treat it analytically

>with machine learning you just feed a lot of images correctly labeled cat or dog and the algorithm figures out by itself
That explains Sup Forums captcha

Sup Forums captcha is from google, and machine learning explains 99% of google actually.