These are chat archives for FreeCodeCamp/DataScience

15th
May 2018
evaristoc
@evaristoc
May 15 2018 16:09

PEOPLE

Where ML is doing ok and where is failing. Hardest? Natural Language Processing. Easy? Images.

http://www.bbc.com/news/av/technology-44128537/facebook-reveals-scale-of-abuse

evaristoc
@evaristoc
May 15 2018 16:26

But...

Don't run to fast with Image processing yet...
http://www.bbc.com/news/technology-44089161

And then there are a couple of driverless car accidents...

Hèlen Grives
@mesmoiron
May 15 2018 20:44
@evaristoc yes that’s precisely why I am doing all this sh*t with image analysis and building my data base. AI is very interesting but I trust my own brain more. Nice article. fB has its own language that is I and maybe others too refer to this phenomenon because it is very Facebookish. That’s funny about humans they can recognize it before it becomes official; written in algorithms. I am jealous about those labs. I still have thousands of files to process. Well I know why I want my AI startup.
Anyway my YouTube rmarkdown file crashed after I rearranged the files. Suddenly read.csv pulled in rubbish. So I am setting up a new directory from scratch. It is such a moment where you want to pull out your hair.
Josh Goldberg
@GoldbergData
May 15 2018 20:47
@mesmoiron can you elaborate on “doing all this image analysis and building my data base?” It sounds interesting. I have to check out the articles. @evaristoc. Love this community by the way. Always great shares on here.
@mesmoiron if you upload your working directory with all the files to github, I can try to trouble shoot the issue for you. You can make it a private repo and invite me as a collaborator. My name on github is the same as here
@mesmoiron I always work with project directories. It’s clean and transferable
Hèlen Grives
@mesmoiron
May 15 2018 20:58
@GoldbergData I have this crazy idea; that is capturing my own analysis process on human level. So I invented a temporary database that captures my thought process. In the mean time I will slowly learn ML to try to automate it later. It is a bit 360 degrees of holistic analysis. I use data from multiple sources. Text, sound, smell, images etc. I build it because IBM has this awesome software they use for crime analysis. Well I could not get my hands on it; thus I decided to think of something better. Like Darwin was obsessed with species; I am obsessed with human interaction and making the invisible visible.
Thanks for your offer; I will push it to Git as it is.
Josh Goldberg
@GoldbergData
May 15 2018 21:29
@mesmoiron that sounds like an awesome project. You should make that your opening lines of your abstract. Haha. I’m hooked. And you’re welcome.