What is machine learning?

Machine learning (ML) is a form of artificial intelligence (AI) that enables computers to
learn without explicit programming. Instead of telling a computer everything it needs to
know to complete a task, ML can enable a computer to essentially “figure it out for itself”,
using data to learn. The more data a computer is fed, the more it learns and the smarter it
gets, improving its accuracy and ability to complete tasks over time.

How does it work?

Machine learning is driven by a cutting-edge technology called Deep Learning. Deep
Learning is inspired by the structure of the human brain and uses artificial neural networks
to process data similar to the way neurons do in our brains. It involves feeding massive
amounts of data through the neural network to “train” the system to accurately classify
the data. Today’s supercomputers and the rise of Big Data have helped make Deep
Learning, and thereby the advances we’ve seen in machine learning, a reality.

Why should you care?

Machine learning is behind many of the technologies that we use today, including online
search, natural language processing (used by digital assistants like Alexa and Siri),
marketing personalization, Netflix recommendations, and much more. It’s also powering
innovations in the works like self-driving cars, and continued research into developing full
AI robots. There’s no question that ML and its impact on our lives will continue to grow as
the technology continues to advance.

So what is it?
Machine learning is a set of algorithms that, like humans, learns from data and/or
experiences, rather than being explicitly programmed. Machine learning utilizes
techniques like neural networks and deep learning, so that it can help to create more
complex technologies like automatic speech recognition (ASR) and natural language
processing (NLP).

Deep Learning
Deep learning is a machine learning technique that teaches computers to do what
comes naturally to humans: learn by example. Deep learning is a key technology behind
driverless cars, enabling them to recognize a stop sign, or to distinguish a pedestrian
from a lamppost. It is the key to voice control in consumer devices like phones, tablets,
TVs, and hands-free speakers. Deep learning is getting lots of attention lately and for
good reason. It’s achieving results that were not possible before.
In deep learning, a computer model learns to perform classification tasks directly from
images, text, or sound. Deep learning models can achieve state-of-the-art accuracy,
sometimes exceeding human-level performance. Models are trained by using a large
set of labeled data and neural network architectures that contain many layers.

Machine Learning Examples

1. Siri and Cortana
Voice recognition systems such as Siri and Cortana use machine learning and deep
neural networks to imitate human interaction. As they progress, these apps will learn
to ‘understand’ the nuances and semantics of our language.
For example, Siri can identify the trigger phrase ‘Hey Siri’ under almost any condition
through the use of probability distributions. By selecting appropriate speech segments
from a recorded database, the software can then choose responses that closely
resemble real-life conversation.
2. Facebook
Remember when Facebook used to prompt you to tag your friends? Nowadays, the
social network’s algorithms recognise familiar faces from your contact list, using some
seriously impressive technology.
‘We closely approach human performance,’ says Yaniv Taigman, one of the
masterminds behind DeepFace, Facebook’s machine learning facial recognition
software.
DeepFace can recognise the differences in human faces with a 97.25% degree of
accuracy – only 0.28% less than an actual human being. Having studied a facial dataset
of 4 million Facebook users, DeepFace has become adept at recognising the nuances
in human countenance across 4000 separate identities.
3. Google Maps
Google introduced machine learning to Google Maps in 2017, improving the usability
of the service. These deep learning algorithms help the app extract street names and
house numbers from photos taken by Street View cars and increase the accuracy of
search results.
With over 80 billion hi-res photos collected by Street view cars, analysing these images
by hand would have been impossibly time-consuming. Machine learning frees up more
time for Google engineers, automatically extracting information from geo-located
images and achieving an accuracy rate of 84.2 percent for some of France’s most
convoluted street signs.
4. Google Search
Sticking with Google, the world’s biggest search engine now offers recommendations
and suggestions based on previous user searches. In 2015, Google introduced
RankBrain – a machine learning algorithm used to decipher the semantic content of a
search query.
Through the use of an intuitive neural network, RankBrain identifies the intent behind
a user’s search and offers them tailored information on that particular topic.
RankBrain now handles around 15 percent of Google’s daily queries, working out the
intent behind never before seen searches much faster than the previous old rules -
based system.
5. Gmail
You might have guessed by now that Google has a thing for machine learning and
artificial intelligence. In 2015, they introduced a smart reply function to Gmail to help
users tackle their inbox, with 10 percent of mobile users' emails sent using this tool
the following year.
The smart reply function is based on two recurrent neural networks: one used to
encode incoming mail, the other used to predict possible responses.
These networks work in tandem to decipher the meaning behind the incoming
message and to automatically suggest three different responses for each.
6. PayPal
PayPal uses machine learning algorithms to detect and combat fraud. By implementing
deep learning techniques, PayPal can analyse vast quantities of customer data and
evaluate risk in a far more efficient manner.
Traditionally, fraud detection algorithms have dealt with very linear results: fraud
either has or hasn’t occurred. But with machine learning and neural networks, PayPal
is able to draw upon financial, machine, and network information to provide a deeper
understanding of a customer’s activity and motives.
7. Netflix
More than 80 percent of TV shows on Netflix are found through its recommendation
engine. Machine learning is integral to this process, as the platform caters to more
than 100 million subscribers
While the finer details of Netflix’s machine learning algorithms are kept behind closed
doors, Tod Yellin, the company’s VP of product innovation states there are two things
that feed the neural network: user behaviour and programme content. Together, these
datasets create multiple ‘taste groups’, which tell the recommendation engine which
programmes to serve up.
Due to its impact on customer retention, Netflix has valued the ROI of these algorithms
at £1 billion a year.
8. Uber
Machine learning is a fundamental part of the Uber model. The tech giant uses these
algorithms to determine arrival times, pick-up locations, and UberEATS' delivery
estimations.
When you book a car, Uber’s aim is to estimate its arrival time as accurately as possible.
Machine learning enables it to do this by analysing data from millions of previous trips
and applying it to your specific situation.
The same goes for UberEATS, which takes thing such as food preparation time into
account to give you the best possible prediction of delivery time. The real-time analysis
of these datasets has improved Uber’s estimations by 26 percent and increased
customer satisfaction in the process.
9. Lyst
Lyst is an ecommerce fashion site working with a new breed of model – the machine
learning model. To match customer searches with relevant recommendations, Lyst
uses meta-data tags to make visual comparisons between items of clothing. Their
algorithms read these tags and decide on the best matches.
To do this the machine learning model must handle three distinct requirements:
Representation – this allows the models to quickly search for similar
vectors without having to process the images from scratch.
Multi-image - Lyst’s products tend to have multiple images associated
with them, so the machine learning algorithms need to map these to one
representation.
Multi-task – multiple product labels (e.g. colour, gender, size) need to be
digested by the visual models to produce an optimal representation.
10. Spotify
 You know that one cheesy pop song you listened to that triggered numerous other
cheesy pop recommendations?
That’s machine learning at work. Much like Netflix, Spotify uses machine learning to
figure out your likes and dislikes and provides you with a list of related tracks.
In its Discover Weekly promotion, Spotify rounds up 30 tracks it thinks you should
listen to and delivers them in one easy-to-navigate playlist. These songs are all ‘hand-
picked’ by machine learning algorithms, which analyse user activity and match your
tastes to music with similar meta-tags.
Accelerating machine learning in the cloud
More than 60 percent of organisations are adopting the cloud for its unrivalled
scalability. It’s this economy of scale that’s driving bigger, bolder approaches to
machine learning and providing companies with the resources to create their own
models.

//CLOSING

1. Machine learning and AI are not the same thing

While often mistaken for each other, these two technologies aren’t the same, but
rather are interconnected. It’s easiest to think of machine learning as one of the
underlying technologies of AI. While the goal of AI is to imitate and mimic human
behavior, machine learning gives us the mathematical tools that allow us to do that. AI
is able to understand languages, conduct conversations, and continually improve itself
because of machine learning algorithms.

2. It’s in the background of almost everything we do

Have you ever been writing an important text or email and notice afterwards you’ve
incorrectly spelled something? Luckily, machine learning algorithms are used to detect
when a word is spelled incorrectly and is able to suggest a new one, saving you time
and embarrassment. Similarly, Netflix uses specific machine learning algorithms to
suggest content for you based off of your preferences.From TV suggestions, to filtering
out our spam email, and much more, these algorithms are meant to radically simplify
our lives, and we encounter them a lot more than we think. According to Srinivas
Bangalore, Lead Inventive Scientist at Interactions, “good machine learning should not
 be in your face. It should be behind the scenes, tracking, and helping achieve goals
much more quickly and efficiently.”

3. Incorporating humans is the recipe for success

No human is perfect, and the same goes technology. Adding a human to the machine
learning process increases the level of accuracy and self-awareness. Humans can assist
with labeling data, and correct inaccuracies, which leads to more accurate end results.
The combination of the two makes the technology more personable, and, in-turn, less
intimidating.

Sources and References:

https://www.criteo.com/insights/machine-learning-fast-facts/
https://www.interactions.com/3-facts-need-know-machine-learning/
https://www.redpixie.com/blog/examples-of-machine-learning
https://www.wordstream.com/blog/ws/2017/07/28/machine-learning-applications
https://medium.com/app-affairs/9-applications-of-machine-learning-from-day-to-day-life-
112a47a429d0
https://cosmosmagazine.com/technology/what-is-deep-learning-and-how-does-it-work
https://www.edureka.co/blog/what-is-deep-learning