Posts

Machine learning : digging into recommendation system

Gotta build on of these for a complex and content rich elearning site.

So did some research.

Here are some of my findings.


Here is a good start :

How to Build a Recommender System by Martin Kihn

The Components of a Recommender System on a site with quite some articles on the topic – like this one on Impression Discounting

Or this one :

Beginners Guide to learn about Content Based Recommender Engines which is from 2015 but still cover the general basis.

Then you have some actual building – for example here in R : A recommendation system in R, applied with respect to the movielens database by Nelson Manohar who seems to specialised in machine learning based on his online profile.


Then we have an old plugin writen for WooCommerce Perso Recommendation Engine Plugin for Woocommerce which is nice to see how it is integrated in WordPress.


Then on youtube, we have quite some tutorials providing some interesting insights with some practical applications – usually movie recommendation engines.


Here we have a good intro from Standford university:

Overview of Recommender Systems | Stanford University


Then finally we can do one :

A really short tutorial (40 min total) to make a python movies recommendation system :

Recommendation Systems – Learn Python for Data Science #3


And the same guy but different stack :

Build a Movie Recommender – Machine Learning for Hackers #4


Then a different approach, with more algorithm into it :

Coding Challenge #70.1: Nearest Neighbors Recommendation Engine – Part 1

Coding Challenge #70.2: Nearest Neighbors Recommendation Engine – Part 2

Coding Challenge #70.3: Nearest Neighbors Recommendation Engine – Part 3


Then a long one of an hour+ integrating with tensor flow (the google library for deep learning) : Recommendation Systems / Engines with TensorFlow – Google Cloud Platform User Group Singapore


Probably more to come, but that’s a good start for now.


Do not hesitate to suggest stuff – either in the comment or on social somewhere


WordPress plugins – AI & machine learning

Just a list of plugin I am considering using as a base for further developments:


Kindred Posts

Kindred Posts uses artificial intelligence to learn how your website visitors use your site and recommends content based on their interests.

(not maintained anymore but developper actively asked for people to join and keep maintaining the project… might do that… we’ll see)


WordLift – AI powered SEO

WordLift helps you organize posts and pages adding facts, links and media to build websites. Lets you create, own and publish your own knowledge graph. WordLift publishes your content as Linked Open Data following Tim Berners-Lee‘s Linked Data Principles.

Doesn’t seem very much self-contained – but probably rely on some external API – i’ll have to see.

Company behind this is italian: https://wordlift.io/


AFS Analytics

I am excited about this plugin because it does 90% of one of the tasks I have to deploy on some LMS and might actually be of great use… will have to dig into that…


Primal for WordPress

Primal for WordPress uses a patented artificial intelligence to
deliver highly relevant and up-to-date information to your readers,
tailored to each page you create.

not thrilled about this one.

First – the applied functionality of the learning process is to fetch external content from the rest of the web – so not very much what I want to do. But maybe the learning algorithm is worth watching. If it is accessible… which is not sure yet.


Watsonfinds

Watsonfinds analyzes your content to provide insights about the emotions that your audience may perceive as they read it.


well… here not much to learn about the AI part but interesting to see how Watson perform on your content.


Contextly Recommends

Contextly helps you build a loyal audience through engagement tools that keep readers reading. Contextly does far more than just related post recommendations. We also get readers to explore your site by identifying your evergreen posts (think oldies, but goodies) and your most popular posts. For larger sites, we can also personalize recommendations for return readers.

Also has a massive amount of features:

Multiple Recommendation Modules / Related Content / Personalized Content / Popular Content / Evergreen Content / Video Recommendations / Product Recommendations / Editorial Control/Curation / SEO / Promotional Links / Responsive Designs / Analytics / Performance / Custom Post Types / Channel Notifications / FollowUp Notifications

but, there is a catch… as always – that none of these elements are a part of the wordpress infrastructure – just some external API – and cost anywhere from 9$ to 99$ per month (more if you go for an enterprise version).

I want to state that while this is not a solution I would recommend on the long run – it does look very appetising – and is on point for most of the features I’d need.

Also – all these cloud solution are raising a lot of privacy questions – what do they do with your users data and who is it sold to in the end… so there is also that.



Bottom line:

seems to have some interesting plugins but most are going to rely on external engine – which is also not open source and not accessible – so that’s a lot of lost data and not very much useful.

Probably next step will be to explore the world of open source engines for machine learning.


WordPress and Machine learning

I have been working on an idea for a LMS project I am involved with.

 

The general problem we are trying to solve is to provide a customise learning path for users

 

Here are some of the videos I found relevant and quick research notes.

 

#Machine Learning with WordPress

(not as much about wordpress as I thought it would be but a nice intro)

 

#Building a Machine Learning Platform at Quora

Very good consideration – must relisten to it again.

 

IDEAS/DIRECTIONS:

  • scoring engine
  • gamification / scoring engine

 

Reference articles

Predicting Customer Behavior: How we use Machine Learning to Identify Paying Customers before they Subscribe

Lasso (statistics) [WIKI]

 

Track User Behavior in Real-Time to Offer Relevant Website Content to Auto Shoppers – Mini Webinar

That last one got like just wow!!!

 

Ok… gotta think about that..

Slaughterbots – AI & Military

 

Watch this video:

 

 

None of the technologies involved in this video are too far away from what is available today.

 

When you think about all the data available from face recognition, and geo-localisation – available from all major corporation – from Google to Apple including Facebook or Amazon – just to name the big ones – it does make you think.

 

And mistakes in targeting happen everyday in advertising – but it’d be a very different topic to have it happen in a military operation – just like it is happening in this film.

 

 

I honestly don’t know yet what to think of it … beside of course that’s a pretty terrible future that’s coming if this would become a reality.

 

I made a post earlier – on dystopian future – go check it out – it *will* make you think…

 

The only thing I am sure, is that whatever evolutions will come from all these technologies, education and preparation are the only things can help us overcome the darkest aspects of human nature.

Are you a robot?

Automation is coming.

It is affecting nearly all industries.

And it would be stupid to ignore it.

But aren’t we taking the frenzy a tad too far?

(Disclaimer:

This article is a bit messy – I have tried my best to give it a decent shape – but if we are to be honnest, it needs some serious re-writing – which I  will probably do – but I wanted to get this out now anyway – so take it as a work in progress with maybe a few unfinished thoughts / sentence here and there…)

Human Need Not Apply

(Watch this one if you haven’t!)

 

AI will replace the human race

If we are to believe headlines :

AI will replace HR (MIT Tech Review)

AI will replace project managers (LPMO)

AI will replace the customer services (The Telegraph)

AI will replace CEO (Jack MA, Alibaba founder)

 

So basically, it sounds like the only one that AI will not replace are the customers (unless for B2B I guess… I mean if all the companies are made of AI employees and AI boss, then who run the purchasing department…)

And actually if we are to belive Elon Musk, Tesla CEO artificial intelligence poses an “existential threat” to human civilization. (NPR)

So, while we can clearly invision AI create serious redundancies in some industries – like for instance in logistics with self-driving trucks or containers management system like Amazon is using, but like many ports around the world as well – to claim AI will take over humanity like Elon Musk is kind of trying to say – is a bit far fetched.

 

Fear is a tool to influence public opinion

Actually, if we were to believe everything that is published, every working person will be a bot, no job is safe and we will all end up unemployed and therefore we have to introduce Universal Basic Income as a solution because we cannot work with each other anymore – all job have been taken – so now – your job is to be a customer – so here is your check and go buy what the AI has produced for you.

Yeah..

ok maybe I am over dramatizing the situation but honnestly, if you read what is being published, this is exactly the kind of future the press is  sort of painting for us.

The mantra seems to be:

All job will be replaced, therefore there will be mass unemployement, therefore we have to subsidize your existence to keep the economy going (i.e. so you can keep being a customer to a system that has replaced you).

 

Let’s pause to think for a moment

First, if machines can do these jobs for us, great, why not after all…

I mean – why would it be a bad thing inherently to have better means of production, cheaper and more efficient?

That would probably limit the amount of accidents in the workplace – especially in jobs like logistics.

So yeah, it will generate unemployement.

Well, clearly we will have the reassess a few things.

Seems like carreer management and personal developpment will be booming.

Some says these people cannot find a job elsewhere.

 

I dunno.

 

Why do we go to work on the first place?

Why do we need these specific jobs?

What will we do with our lives if most of todays’ jobs are gone? (Which is a much more frightening questions that I don’t think is discussed very often)

If this is such a bad thing, why did we introduced AI on the first place?

And it is good, why is everybody so worried?

 

I think we should consider a couple of things.

  1. Let’s make the difference between a “job” and “work”
  2. Let’s try to see what still make us human once machines are doing everything for us

Job vs Work

We all get the difference between value production and just getting busy.

At least on paper.

Because sometime even the most efficient people are caught into “getting busy” and are not accomplishing anything.

Just being employed – or being useful to those around you – to society – is how I see the major difference between jobs and work.

In a 12 min TED talk Rudy Karson make the point that in the future what matter is not to have a job but to be useful, to make an impact:

And I would argue that somehow this is connected to the fundamental element that we have and that machine cannot replace us.

The quest for purpose.

What makes us human?

 

 

 

When we have jobs we don’t like, we are looking for ways to improve them.

So we created Artificial Intelligence (among other things) so they would free us from the daunting task of doing boring shit.

In a nutshell, that’s how we got started with AI.

But then we started to realise it makes some (or all if you believe headlines) our fellow humans’ activities less relevant.

 

So there a couples of points we could and should think about:

  • These robots are tools in the hand of humans – so let’s not forget that AI makes anybody miserable in the end, that is still the result of an other human being making use of AI in a specific way that apparently doesn’t improve the life of everybody…
  • If the work we do is not efficient and we are wasting our lives – getting busy with a fucked up job – maybe mass unemployement – somehow is way to structure change how we occupy our lives – what are filling up our days with…
  • Headlines from 20’s or 30s or 60’s have always been fascinated by technology taking over humanity and so far, it hasn’t really happen.

 

All these are open points.

The conclusion I want to make here is maybe less down-to-earth – but not less important:

Robots exist because they work.

We work because exist …

Since, we are alive we work and create things around us, through this work.

We can create the smile on the face of customer or create a piece of engineering that will aleviate somebody else situation.

We work because deep down we want to feel part of society and of the world, we want to feel connect to them in a meaningfull way.

Robots just work – at best.

They can do many thing instead of us.

But they cannot live our lives instead of us.

As long as we exist, we will do things for one another.

The fact that we have quantified these interaction with money and categorised them it in jobs, doesn’t take from the fact that our activities are a way to express our connection with others.

That’s why we feel more alive in a job where we feel meaningful, useful, where we can contribute and our opinion can be taken in account, etc.

So,  yeah… this whole article is all over the place and not very well structure – I am totally aware of that – but I needed to put get it out so that’s kind of a first draft to be refined…