The_Who-_What-_and_Why_of_Machine_Learning_in_your_Business-3

Machine learning is a hot topic, and it’s changing the world around us. Yet many of us still don’t know much about this emerging field. It's crucial to understand machine learning, as it redefines and shapes our daily life. It's time to consider it as part of a solution in your business - if you haven’t already.

Believe it or not, machine learning is nothing new. It was conceived 50 years ago by Arthur Samuel. It “evolved from the study of pattern recognition and computational learning theory in artificial intelligence". Machine learning uses statistical techniques to give computer systems the ability to “learn” and adapt. Then, they are able to make predictions based on that data. For example, a machine learning algorithm can tell if a person is likely to buy cushions because they recently purchased a new couch. The algorithm can make this prediction due to previous data where other people buying couches, bought cushions as well.

It evolved from the science of recognizing patterns and the use of computational learning theory in Artificial Intelligence (AI). Over the past 5 years, machine learning has become increasingly common. Now, it's an integral part of software for many large scale and high profile applications. Why the hype? And why now?

Data is everywhere

With the popularity of apps like Netflix, Facebook, Twitter ,and LinkedIn, user data is in abundance. When data is very big like this, it becomes difficult to make sense of and leverage in order to bring value to customers. By using algorithms and models to do work humans aren’t able to do, machine learning is able to tailor customized solutions for the end users’ needs, mostly with very high accuracy.

Today, healthcare, marketing and advertising greatly benefit from machine learning.

Andrew Burt and Samuel Volchenboum highlight the importance of machine learning in healthcare. The technology offers doctors insight on the status of critical patient cases in emergency rooms. In one example, they showcase a study conducted for emergency departments to predict the odds of a patient dying after spending 24 hours in the ER. The algorithm performed with a whooping 90% accuracy!

John Jersin explains how LinkedIn uses machine learning to tailor job searches for users on the platform. Not only is machine learning detecting trends and making predictions, it's adapting to the changing resumes of users. It can keep up with different job postings as well as new fields of expertise added to the platform daily. With that constant evolution of data, it's still able to tailor targeted content for the user. Whether it’s the job seeker, HR/recruiter or the company posting the job role, LinkedIn can match users with roles according to the job seeker’s skills.

Using machine learning algorithms at LinkedIn removes the need for paperwork and minimizes wasted time. It also makes matching jobs and job seekers seamless, fast and more accurate.

Xiyang Chen, a Rangle colleague and specialist in machine learning, says: "We’re at a tipping point where we have devices around us constantly collecting data. We have a very fertile ground for machine learning and algorithms that can scale, learn and adapt better to trends."

"We’re at a tipping point where we have devices around us constantly collecting data. We have a very fertile ground for machine learning and algorithms that can scale, learn, and adapt better to trends."

Storage is Much Cheaper

One of the biggest challenges of machine learning in the past was the availability of the data sets. Today, holding a large amount of data has become a non-issue with the reduced cost of cloud storage. So, introducing machine learning as part of business solutions has become easier. This is thanks to the growth of analytics, metrics, and CRM datasets available.

Recently, Google announced its Google one making cloud storage cheaper than ever! It’s only a matter of time before others in the space, such as Amazon and Microsoft, follow suit.

The Giants in the Field are Encouraging Machine Learning and Making it Accessible

At Google I/O 2018, Google announced its ML kit. It’s very easy to use, downloadable from firebase, and great on mobile. They’ve provided developers with a basic algorithm or model that offers a good degree of prediction, enough for basic machine learning tasks that can run locally (offline) on the device. The model will learn and train itself subsequently without needing to connect back to the cloud. This makes ML very easy to use and accessible for those that are not experts in the matter.

Amazon’s (AWS) machine learning use is up by 250%, more than double in one year. They have created an easy-to-use ‘playground’ for developers to test their machine learning theories. This provides great infrastructure, is setup to consume large amounts of data, and can scale as data becomes larger.

Apple provided Core ML for developers, giving them a useful API that can help developers that don't have intensive machine learning knowledge. It supports many of the standard ML/AI features such as the vision for image analysis, foundation for natural language processing, along with many others.

Government Initiatives

So far several countries have started national initiatives to support and encourage AI under which Machine Learning falls as a subset. World leaders are recognizing how this field will reshape governments. Canada has committed 125 million dollars with its Pan-Canadian Artificial strategy to support the research and training, in hopes of establishing Canada as a leader in the field. This massive support from the Canadian government is driving investments in this domain to over $1B dollars in the last year alone.

Culture is More Accepting

Cultural change is a big part of why machine learning is picking up and scaling so fast. People are more used to hearing about it and becoming increasingly comfortable with using it in daily life. We are relying more on this technology (knowingly or unknowingly) to go about day to day tasks.

Machine learning is making people’s lives easier by enabling computers to resolve tasks that have, until now, been carried out by humans.

Machine learning is making people’s lives easier by enabling computers to resolve tasks that have, until now, been carried out by humans.

The big question asked by technology companies is not ‘why use machine learning’, but:

  • ‘how do we apply machine learning’
  • ‘what are the use cases in which our company will benefit’
  • ‘how do we leverage it in our solution’
  • ‘how do we achieve our goals using ML’

These are loaded questions, but if approached with the correct mindset and right strategy, it will prove to be rewarding for a company keeping up with technology and the future.

Here are a few starting points to consider if you are considering using machine learning as a part of your solution:

Collect Data Early and Focus on Quality

According to Xiyang, data quality is the most important part of achieving a good solution. He says, “Data quality must to be good, therefore we need to educate the client early. Data collection has to begin during the early phases of product design and implementation, so that we have a substantial amount of data for better algorithms. Sometimes this comes as an afterthought and opportunities are lost.”

This will probably be the hardest step, having enough data to train algorithms on, as well as providing clean data with reduced noise to get better accuracy in models.

Keep Privacy in Check

With the latest Facebook & Cambridge Analytica controversy and data breaches on the rise, privacy and security of data has become a hot topic. Scientists and technology professionals are putting effort in using different techniques to mitigate risks. Companies have a bigger responsibility for ensuring the privacy of the data they collect about their users. There are many approaches to take from, including:

Anonymizing the data so that it becomes very difficult to figure out identities. One way of approaching this is to collect data, aggregate them, then have two separate servers, one for identity and another one for the anonymous data. The aim is to make it extremely hard to single out identities from the data used by the machine learning. Another example is using nicknames instead of the user’s real name.

Using Differential Privacy Rf* is based on learning as much as possible about a group while learning as little as possible about any individual in it. This is achieved by adding enough random noise to the aggregate data, although the application of it will depend on the country’s laws and regulations and make sure the data conforms to them. Now there’s the question of how this complies or compliments the GDPR, it is shown to be a very powerful and promising approach. More related work on this approach is found here. For ‘an overview of a system architecture combining differential privacy and privacy best practices’, a well documented article was written by the Differential Privacy Team in order to learn from a user population.

Google's sidewalk labs project in Toronto will be using cutting edge technologies and making the most of machine learning, which will provide lots of efficiencies, but at what cost to privacy? Only time will tell what approaches will be taken to ensure the privacy and security of people’s data as well as their right to take part and withdraw from this project altogether.

Hire the right team

This might sound like a no-brainer, but most of the time, initiatives to use machine learning as part of the solution fail because the right team was not built to succeed.

A good team to start a pilot will have at least two developers that know Python, R or any other language of preference that can achieve machine learning goals efficiently. You will need a strategy leader on the team that can help guide the project from the pilot stage, to production ready, as well as the long term objectives of the software. A data analyst is required for learning and adapting in house data. This is very important because they truly understand data, they know what to look for, what can be done with data and what to expect, bringing great support for the entire project.

This combination of the right skills, will get you started and set you up for success. To learn more about what machine learning is or what your team will be applying, Priyadharshini wrote a nice article that explains learning in machine learning.

Now, if you have trouble hiring people with those skills, which most companies find challenging, Xiyang suggests increasing your company’s capacity with machine learning by creating a study group, which is a team of few professionals chosen internally within the company. That team will already be made up of engineers that are domain experts who will know different aspects of the project.

Set Your Company Apart From Competition With Machine Learning

With the very strong competition in the market nowadays, companies are trying different ways to set themselves apart from their rivals, and machine learning seems to be a top choice to shine.

Megan Beck and Barry Libert state this fact really nicely: “Capability with machine learning is what distinguishes a good platform from a bad one”. They gave the example of monster.com, which didn’t implement machine learning, vs LinkedIn or Glassdoor which realized that the piles of resumes became too large for users to process and find jobs that match. They provided the same solution way using machine learning.

Another example is how powerful machine learning was for older companies is Netflix, which reinvented itself and delivers tailored content to the users unique preferences.

Research, Education and Seeking the Experts

One of my favorite action items is researching the market and what is being accomplished in this field. It’s important to educate oneself about the different techniques, and how this applies to your business so you can implement the correct solution while having a solid strategy and long term vision. The internet is full of supporting documentation, these are two excellent pieces showcasing the right questions to ask when considering machine learning:

Conclusion

In conclusion, the machine learning hype is real. In fact, it is so well justified that Forbes magazine predicts businesses to double their use of machine learning by 2018! With the availability of enormous amounts of data, cheap storage, machine learning kits provided by the industry’s leaders and funding from governments as well as investors, machine learning has all the ingredients to be adopted by companies. If you are not thinking about it yet, this is where the future is going. What better way to be ready than start now!

elevate banner