How the fuck do I get into AI? Machine learning specifically. I see all these fags in my Univ class who use tensorflow to make stuff but if I ask them how that shit actually works, they can't explain. I don't want that, I want to actually understand
How the fuck do I get into AI? Machine learning specifically...
Mason Martin
Other urls found in this thread:
programsandcourses.anu.edu.au
twitter.com
Jacob King
I am not expert nor in knowledgeable to give advice. it depends on what do you want to do with it,
but check this book Artificial Intelligence A Modern Approach 3e
genetic computing, back propagation A*
be specific in what you look for, you want to learn it for what? to make video game or to cure cancer or prediction then you will find that you have specific topics to search. all you need is proper books and before all find yourself a project that need AI
Asher Miller
If you actually care, major in math or stat instead a meme degree like cs
Brandon Bailey
>meme degree
meme degree? how do you think CS is less than Math?
Luis Diaz
Welcome to Jow Forums mate
Daniel Cooper
well let me guess it was programmed by white man? or racist bitch. I have degree in AI and I'd like to tell you that you can program anything to do be however you like.
Get a team of black people majority to make AI and you will see wonder(I mean counter racism) when it comes to white people
Christopher Taylor
I've got several Data Scientists on my team, they were all Stats majors. They might not be the best programmers, but they CAN explain how the ML algorithms work.
Caleb White
>well let me guess it was programmed by white man? or racist bitch.
KEK, Microsoft are one of the most SJWs corporations.
Lincoln Miller
This
>T. CSlet
Easton Moore
I can recommend 'The elements of statistical learning' by Hastie et al as well as 'Pattern recognition and machine learning' by Bishop.
Both overlap somewhat, although o prefer the former. You need some basic math understanding but it's a genuinely enjoyable read given that.
Nicholas Fisher
you are way behind the curve ad will eve catch up
Jacob Brooks
bratposters aren't allowed to use machine learning, sorry
Charles Jackson
Why is Saku mad?
Michael Morales
the dl redpill is that nobody understands it. The 'nice' mathematical algorithms that could be understood, like svm or more recent and complex vae were destroyed by a combination of student descent (ie. lots of student try many different things) and hardware power (throw hundreds of gpus at every algorithm that makes some sense).
The most recent trend is very old genetic evolution, so basically just bruteforce. On local gpu cluster and for classification problems it's worse than gradient descent, but it's main advantage is absolute parallelism and universality. Same evolutionary algorithm can learn to classify images and to play atari games.
That's how the ai future is going to look, just massive bruteforce with relatively simple evolutionary algorithms, as hardware gets more and more parallel any algorithm with sublinear scaling is going to be obsoleted.
Nobody really needs a phd to do ai, but because it started as a math-heavy field having a phd turned into pure signaling. In reality, a team of phd with 10 gpus would be absolutely destroyed by one average programmer with a 1000 gpu cluster. Once you are able to understand basic linear algebra additional math knowledge doesn't lead to any improvements, or even worse, it can lead to pointless overengineering.
The research culture is horrible, heavy overfitting to claim 'state of the art' is the norm (overfitting is done by grad student descent - meta parameters, network topology or even random seed). Because of this, most new and cool algorithms turn out to be worse than their simple predecessor.
The impact of random seed is enormous and that's when I realized the field is 90% bullshit. You can take state of the art papers and just by changing the initial random seed get variations on the order of 1-2 percentage points, while new soa often 'improves' over previous results by 0.1pp.
Aiden Garcia
Jesus Christ how horrible. I wanted to get into AI research but after reading this post am seriously reconsidering that.
Ayden Jackson
Read ben goertzel's published papers on AI.
No one here knows anything except maybe a couple people.
Isaiah Ortiz
Because you just learn how to apply math and stats without understanding the fundamentals of why.
Landon Myers
>t. failed the google brain interview and is now salty
Brody Brooks
What's the best way to git gud at ML and related things?
Lincoln Nguyen
Nothing has wasted the talent of more great programmers than the study of AI.
Evan Stewart
Same but how do I get into NOT Machine Learning AI? I literally cant find anything that isnt about writing stupid classifiation NNs: I want to learn how to repliate human behaviour
John Stewart
Invent the field
Almost nobody does this
Kevin Stewart
Unironically? Cognitive/neuroscience with a really strong background in programming
William Bennett
Read "Stuart Russell and Peter Norvig (2010) Artificial Intelligence. A Modern Approach (3rd edition)."
But seriously just look for an AI degree from a decent university, check the courses, and go through the prescribed textbooks for those courses and their prerequisite courses.
Here's one from my uni programsandcourses.anu.edu.au
Kevin Phillips
>how do i scikit
Jack Hall
I wanna sniff Bratass
Adam Morgan
Rude