Business built on AI - Google & Facebook

½ of the top 10 biggest companies by market capitalisation are built on data.

Amazon $bn 735,900

Alphabet (Google) $bn 728,360

Facebook $bn 375,890

Tencent $bn 375,110

Alibaba $bn $355,130

(Values as of Q4 2018, source – Wikipedia)

Google

Each of these companies deeply exploit machine learning, but Google (now Alphabet) was founded on a machine learning Algorithm called Page Ranking.

Figure 27. Google Page Rank

Google’s page ranking algorithm worked out the probability distribution for the likelihood that a person randomly clicking on links will arrive at any parti cular page. The initial prototype of the Google search engine was published in 1998. By 2012, Google search was handling 1.2trn enquiries a year.

Figure 28. Growth of Google searches. Image source, Internet Live Stats

Google has grown its business way beyond just search, and now uses artificial intelligence and machine learning in everything from Gmail and battery management in Android, to gathering news headlines, creating robotic voices which sound human, adding colour to century-old photos, and teaching autonomous cars to drive in the snow.

Facebook

The core business problem that Facebook tries to solve is matching buyers to sellers with targeted adverts.

Figure 29. Facebook's business model simplified

The raw material that Facebook works with to help advertisers target effectively is the massive amount of social media data that its users upload. Each piece of data uploaded reveals something about what they may be tempted to buy.

With 1.2 billion people uploading 136,000 photos and updating their status 293,000 times per minute, until recently Facebook could only hope to draw value from a tiny fraction of its unstructured data – information which isn’t easily quantified and put into rows and tables for computer analysis. Deep Learning is helping to play a part in changing that. Deep Learning techniques enables machines to learn to classify data by themselves.

A simple example is a Deep Learning image analysis tool which would learn to recognise images which contain cats, without specifically being told what a cat looks like. By analysing a large number of images, it can learn from the context of the image – what else is likely to be present in an image of a cat? What text or metadata might suggest that an image contains a cat?

Being able to extract knowledge from user content enables Facebook to offer ever more granular targeting for advertisers.

Key ways in which Facebook uses AI include:

1. Textual analysis 

Extracting meaning from words. This is used by Facebook to direct people towards products they may want to purchase. Spookily, AI listens to your conversations and accurately detects when someone is asking where to eat or buy shoes in a given area.

2. Facial recognition 

Now banned in the EU, Facebook’s image recognition tool is more successful than humans in recognizing whether two different images are of the same person or not.

3. Targeted advertising

Clustering users together to decide which adverts to show them.

4. Designing AI applications

Facebook runs simulations of 300,000 machine learning models every month, to allow engineers to test ideas and pinpoint opportunities for efficiency.

Facebook aims to combine visual and language AI engines to derive meaning from Facebook posts.

Most of Facebook’s Deep Learning is built on the open source Torch platform.

Complete and Continue