London.pm
20 minutes
Any
English
I attempted to train a neural network to do my job. Come to my talk and find out how I got on!
My job consists primarily of two things - talking to people, and writing code. Through the use of recurrent neural networks, I can imitate the way I talk and the way I code by feeding the network examples of what I've written before.
In this talk, I'll explain the journey I went on in order to gather examples of things I've written, feed them to a neural network, and then use it to imitate myself. I'll also go over some ways this could be expanded and *maybe* find a practical use, and the handy automation tool I built to help other people attempt to do this: https://github.com/MattChubb/torch-rnn-kit
Attendees need no previous experience of Perl or artificial intelligence. The talk includes a bit of shell, plus some Perl, and a single slide of JavaScript.
This is an extended version of the lightning talk I gave at the tech meet in May.
Silver:
Bronze and Community:
More information about all of our valued sponsors is available on our sponsors page.