AI: Artificial Intelligence

What do you think of artificial intelligence, Jow Forums? Pros and cons? Any concerns?

If you know any interesting articles or research papers please link them.

hooktube.com/watch?v=cQ54GDm1eL0

Attached: AI.jpg (712x400, 404K)

Pro: Allows us to improve our already improving ways of life. Translations, Scientific Research are just a couple I can think of.
Cons: Costs lots of money as it requires a lot of computational power to product reasonable results.

>Pros
It's interesting.

>Cons
Most people have no clue what AI really means and ask questions like:
-"When will AI replace programmers?"
-"How can I built a programm that talks to me like a human?"
-"When will Terminator 2 become reality?"


This is not how it works, lads.

>This is not how it works, lads.
But it is how it's supposed to work.
That it doesn't work right now just means that our implementations are shit, not that it's not part of the AI field.

We won't have a strong AI for the next couple of decades.

That's subjective. AI is already very good at image synthesis, problem solving, risk assessment and building profiles on people with thousands of data points.

Frank Herbert, the 'Dune' series

Attached: Dune-1.jpg (1920x1080, 108K)

ai is just bunch of if then blocks lmao

data scientist here, ama

do you like cheese?

But AI has already replaced programmers... Instead of hiring lots of smart people to make a good old algorithm to solve a really tough people people hire 1 AI dude who sets up this massive function approximator that produces the results you want without you even knowing how

What kind of name is ama

Why is my AI course and my ML course so different?

because the two are different things

Attached: machine_learning[1].png (371x439, 28K)

It would be cool to build a robot that detects and kills jews

yes

>What do you think of artificial intelligence?
It's just highly automated statistical analysis
>Any concerns?
Idiots thinking its anything more than that and pushing it to places where it doesn't belong

quite autistic

deep learning "experts"

Attached: 5193db0avbey.jpg (947x946, 71K)

Does adding more layers really make a neural network more accurate?

in general yes, but I would consider it as one of the last resorts cause you have to consider performance. It is better to re-evaluate your architecture as a whole and other hyperparams first, then more layers are a bonus to that. Lot of ML pajeets just dumbly stack layers and think they are the shit, cause the model performs better. That is the only solution they know. Depending on the architecture, the improvement can be marginal though.

No overfitting causes many problems

I'd just like to interject for a moment. What you're referring to as Machine Learning, is in fact, Statistical Learning, or as I've recently taken to calling it, Statistics plus Automation. Statistical Learning is does not make machines
intelligent unto themselves as the term "Machine Learning" would suggest, but rather is referring to methods that have been long since developed, which have recently been made accessible by various, easy to use free open source libraries such as Scikit Learn.

Many people have been using services that use Statistical Learning, without realizing it. Through a peculiar turn of events, there has been a marketing push to what is often called "Machine Learning", and many of its marketers are not aware that it is simply automated statistical analysis.

Attached: stallman.jpg (600x450, 45K)

AI is just heavy statistics and some advanced mathematics. Honestly more of a buzzword than anything else.
t. Big data memer

gonna shill this thread to you if you're interested

Cool, thanks.
>tfw Jow Forums is to /sci/ as /v/ is to Jow Forums
>we're their faggots

i disagree. Most pepople on /sci/ these days are pol manchildren boasting about their iq

>heavy statistics
The extent of most data science stuff I've seen in production is logistic regression or CART as they are the simplest to understand and debug. Most of the real exotic algorithms like boosted/bagged models or SVM are usually only used in competitions and academia.

The math that's used isn't that advanced either. If you understand basic linear algebra and calculus, then you have 90%+ of the knowledge needed in this area.

Big data truly is a meme and so is the 'data scientist' job title. Any kind of data analysis without an understanding of the task at hand is a complete waste of time and the results generated are complete crap. This is why all AI has resulted in these days is expert systems rather than putting everyone out of a job.