I am a CS student and I want to be developing AI interfaces for devices (like siri, alexa) in 4 years...

I am a CS student and I want to be developing AI interfaces for devices (like siri, alexa) in 4 years. What would be a good transitional job to get industry experience, and what certifications could I get to enable me?

Other urls found in this thread:

deeplearningbook.org/
twitter.com/NSFWRedditVideo

bump.bmp

Get out of CS. Become an actual scientist. For real. Do biology, get a masters or phd, and get some work experience working in the field with large amounts of data about plant or animal life in an ecosystem somewhere, but always take a programming approach to creating models from the data.

I can't imagine a better way to effectively show a company that you can create predictive models from huge amounts of relatively unstructured data.

go on coursera and/or mit opencourseware and enroll in free courses related to that

write your own projects so you can actually get hired

alright, thanks
I don't have the time to fuck around in another field for 8 years, I might start with some real-world independent projects though.

Can I get hired straight into my desired field without previous tech employment though? if not, what would be a good first experience in the business?

>Get out of CS
>to do CS
What the hell are you talking about?

>with large amounts of data about plant or animal life in an ecosystem somewhere
The kind of analysis you typically do in cases as those are quite simple.
The only complexity is interpretation, and this is nothing AI can do at this point.

start by sucking a couple dicks

>I want to be developing AI interfaces
>in 4 years
Get a job and work in machine learning research positions for the next 10 years.

The AI used in something like Siri is not actual AI but rather text recognition. A certain amount of language parsing like what is used in computer language compilers enables these text recognition programs to decipher meaning, like how MS word is able to decipher meaning of incomplete sentences. I would expect there are a lot of english and linguistics majors that are part of the Siri design team as they have to be able to interpret linguistic idioms in order to give proper responses.

>The AI used in something like Siri is not actual AI but rather AI.
Why are you talking about something you know nothing about?
Please stop, it hurts reading these posts.

Get out of CS and get into interactive systems design.

There's a massive amount of overlap at most universities, so if you have a bunch of credits you'll be most of the way there. Obviously though, there's an approach that's far more oriented towards designing systems and interfaces, as opposed to just doing back end stuff.

because I do know how machine learning works and that kind of thing is not used in chatterbots, you are one of those pretentious /sci/entists who just starts shit threads about how computers will have feelings and take over the world

Learn about everything that has Markov or Beyes in it's name.
It's the basis for most predictive models (MDPs and so on).
From there on just keep on reading papers.

Install gentoo

You apparently don't even know what AI means to engineers/computer scientists.
Have you read AI: A modern approach ?
It is pretty much considered the standard introductory work, and it has a pretty conclusive definition of AI, which is quite different from what you suggest.
If you don't even know the definition of what you are talking about, why should I take you seriously?
Protip: AI is not exclusively 'machine learning' or some other technique.
If you want to learn more, just read a fucking book.
'A modern Approach' is the best start.

Ive read a little bit of AI Modern Approach, it uses the ideas of 'agents' that are specific for performing human like tasks. Again, that kind of thing is not necessary for something as simple as a chatbot. You have to understand that there are chatbots that have passed the Turing test because the Turing test does not really define intelligence, all you have to do is create a program with real enough sounding responses. If you have access to emacs ake a look at help>emacs psychotherapist to see a simple chatbot program that can mimic human conversation, its not difficult to program these

>there are chatbots that have passed the Turing test
Haven't heard of them.
The ones I heard of were scams, which just used psychological tricks.
Most of them can't reason at all, which is the actual tricky part in human conversation.
And if you want to program an AI Interface (like siri) these programs will have to reason!
They will have to know what you want.
This is why you have to start using predictive models.
And for that there are many different techniques, of which it's not clear which are the best fit for this certain task.
This has nothing to do with computers having feelings or some shit,
but rather with them having to understand what the fuck you actually want from them.

The Turing test is furthermore not the goal or a criterium of AI, but rather a suggested test for human-like AI.

> uses the ideas of 'agents'
>at kind of thing is not necessary for something as simple as a chatbot
You are either stupid or didn't actaully fucking read a modern approach.
A chatterbot is a fucking agent by the definition given in the book.
An Agent is something that simply receives input from the outside,
does something inside of it with that information, and reacts on it/ ouputs a response.
As you can see, that is what a chatterbot does.
As engineers/cs we see an agent that acts rationally as intelligent.
That's it.
It's even flat out stated that way in 'a modern approach', so if you would have read it, then you should know.

Im not saying that Machine Learning is not used in Siri, Im sure it is. Im just saying you can get pretty far making a chatbot with realistic human responses just by using common sense and simple word categories.

And I'm telling you to read 'A modern approach' before talking about AI.

Nigga just wait till you take your first babby AI class. It's not like the movies kid, programming AI is hard and fucking boring.

Take the nanodegree courses offered in AI. Think you can take them for free.

Take Andrew Ng's Machine Learning Coursera course.

Buy some intro level Deep Learning books

Review basic Linear Algebra, Calculus, and then review Statistics hard.

Read a 'real' Deep Learning book : deeplearningbook.org/

Read 'real' Statistics / Linear Algebra text. Google for these.

I love AI programming. I think it's fun doing knowledge representation, logical inference, state search, goal planning, and learning. I love the theory and then seeing the results when I implement it.

and wait around for the guy with the cs degree to do your work for you

A PhD, and then a few influential papers as a postdoc, and then years of working towards a tenure track doing rigorous research as a professor. You are delusional if you think you can do anything with just a bachelor's in "industry".

A PhD in computational linguistics.

All you should need is a portfolio proving that you can make a computer to make rational decisions and answer questions and also the mathematical rigor to prove what you have.

You don't need a PhD. An STEM degree with MOOCs and projects is sufficient for an entry level position. Most people I talk to in industry say they rather you have 5-6 years of actual experience as opposed to 5-6 years in grad school.

Maybe for a generic code monkey stuff but if you are doing cutting edge work in AI you need proof of your research skills.

For cutting edge work yes you need a PhD.

Siri doesn't reason you prajeet cumquat. It's just a chatbot