For some of us, there is confusion between Deep Learning, Machine Learning, and Artificial Intelligence. All these three words meant disruptive innovation in the computer science world. But what is the difference? In this article, we’ll understand specifically Machine Learning and Deep Learning, as well as their applications in business.
Artificial Intelligence is a study to enhance computers to mimic a human’s mind. The exploration of AI began in World War II with Alan Turing and his Enigma machine.
We can generally categorize AI into three types: Artificial Narrow Intelligence, Artificial General Intelligence, and Artificial Super Intelligence. ANI is called Weak AI, while the other two are Strong AIs. While Weak AI performs particular tasks, it will not surpass a human’s mind. Strong AI, on the other hand, imitates human intelligence.
Machine Learning is the study of training computers to learn with statistical data. It is a subset of Artificial Intelligence. Tony and Stephanie created this visualization of how machine learning works.
As ACADGILD’s YouTube video mentioned, there are supervised and unsupervised learnings in Machine Learning. In supervised learning, machines predict outcomes with the help of a data scientist. Meanwhile, unsupervised learning is where machines predict outcomes by themselves using patterns from the data they analyze.
Finally, Deep Learning is the study of imitating a human's brain into computers using multi-layered analysis to conclude to solve a problem. Deep Learning is a subset of Machine Learning.
Companies like Google, Amazon, Apple, and Microsoft have invested in Machine Learning AI technologies. Deep Mind, which Google acquired, mentioned how they wanted to give back to the community with Artificial Intelligence through real-world impacts.
Elon Musk, CEO of SpaceX, created Neuralink in 2016. One of its products is a chip that can be planted into human brains. One use of such a product is to help people with brain damage regain control. For example, a video by Real Engineering depicts how one can grab a can of soda without moving her hands.
We can also look at the ‘smaller’ impacts of Machine Learning. Personal Assistants like Siri, Google Now, and Alexa use Machine Learning to better serve humans. Facebook and LinkedIn use “People You Might Now” with algorithms based on learning: they learn whose profiles you look at, whose friends you connect with, your interests and activities, and more. Regarding security, there is also Face Recognition and spam/malware detection. All of these use Machine Learning to build better customer products and services.
Deep learning requires a larger data set and a longer work time than machine learning. In its methodology, deep learning constructs layers with increments in information abstractions. It uses neural networks that try to mimic how the brain works. Information is transferred from one layer to the other through connecting channels.
Serokell mentioned that Deep learning had been used in multiple fields, such as Natural Language Processing, portfolio management, drug discovery, self-driving cars, and robotics. Geeks for Geeks added some more: automatic text generation, automated machine translation, and earthquake prediction.
In implementing machine learning and deep learning, we can look at two types of apps: customer-facing and employee-facing. The key here is that we train computers to learn, which will assist us in making decisions.
Customer-facing apps benefit users with easy-to-access information/reports they can use in purchasing decisions.
Bonlook is one of the eyewear companies that use virtual try-on features. A customer can record a video of their face and later use it as a try-on for different eyeglasses. The purpose is to help customers decide which glasses they would buy without the need to visit stores. In this customer-facing app, the computers first recognize the customer’s faces. With smart technology, the app fits the glasses with the face as it moves around.
Apple Health reports health statistics based on the user’s movement captured by the iPhone’s sensors. They even went further with Apple Watch, where the data can help predict the next seizure the user most likely experience. Users or their loved ones can contact hospitals and prepare for the event. The company can better serve the healthcare industry with the amalgamation of Apple’s data science data.
Employee-facing apps benefit companies by enhancing their capabilities and increasing productivity.
With the help of machine learning, Google and Facebook can now target ads to different customer segments. The algorithms essentially learned about customers and their behaviors: sites they visit, friends they have online, interests, etc. When a business owner wants to advertise his products and services, Google and Facebook promote the news to targeted segments.
Persist IQ is one of the many email automation tools sales teams can use. One of its features is Intelligence, where Persist IQ can identify already existing prospects. It can avoid misformatted emails and also identifies and categorizes response emails.
As you launch your business, you begin with the “Launch” stage, establishing your brand. The next stage is “Mature,” where many customers purchase your products and services. Finally, you are challenged to the next stage, “Growth,” where you experiment in a new field to cater to the customers’ needs. Learn more about which team should own the product experience here.
When you start, you can start by creating the learning algorithm. You can set up computers to categorize activities and recognize speech and text patterns. For example, the machine can categorize a repeat customer versus a one-time buyer or show related results when a customer uses a search bar. It is important at this stage to build your IT infrastructure. If you don’t have the right data set for your algorithm, you can borrow some free data that is publicly available. In this case, you will train the computers with free data before putting in your customer data later as you collect them.
In the Mature stage, you might want to use more advanced algorithms. The goal is to enhance customer experience with new features like chatbots and personalized recommendations. You can also offer an image recognition feature. Chase and Bank of America are examples of banks that allow check deposits through mobile apps by capturing an image of the check. You can try text recognition as well. For example, Google launched a mobile app called Voice Search to help ease search experiences. You can read an overview by Wordstream here.
Lastly, the Growth stage is the turning point for your business. Here, you are challenged whether you want to grow or remain where you are. Companies like Blackberry and Nokia decided to stay on the course and failed to capture the new trends of touch-screen phones. But if you are susceptible to changes, think ahead of the game. In the Growth stage, you want to incorporate Machine Learning and Deep Learning in your experimentation. You would look at a customer segment you want to focus on and offer new products and services to them.
A quick example would be Generation Z, who can navigate the online world. Here, you want to incorporate fun and useful activities into your apps. An idea would be to launch a social media-integrated app for shopping. So, for example, they can take a selfie or record a video of themselves hanging in a store and quickly get price tags of the products around them.
Downturns are inevitable. We live in a world where change is the only constant variable. In that case, we need to be always prepared for downturns.
A good case study about facing changes is a publishing company the Designial team worked with. The challenge is the ever-growing demand for eBooks that has portability and cheaper prices. Designial’s solution is to embrace instead of reject the digital revolution. The publishing company specializing in medical, business, and engineering reference books then offers an app for customers to ‘try’ the books. This helps customers decide whether to invest in their learning and buy books.
Some practical thing you can do in times of downturn is to invest in IT infrastructure. Technology will advance, and starting early will help you save more in the future. You can rent computers and research several Cloud Computing platforms, such as Google Cloud, Amazon AWS, Microsoft Azure, and IBM Cloud Pak & Data.
For Machine Learning and Deep Learning, you can start building your algorithms and later use a data source to train the computers.
As we know, analytics are very useful for predicting an event's likelihood to occur based on given data. However, it does not necessarily assist us in making decisions.
For example, we know how Tesla is leading with self-driving cars. Regarding autonomy, their cars are at level 2 on a scale of 0 to 5. With a forward-facing camera, the car can see two cars in front of it. Should the one in the front hit a brake, it can predict whether the second one (the one right in front of the user) would turn left, right, or hit a brake. Knowing that humans have a gap in reaction time, Tesla cars used that time to activate an algorithm for safety. Deep learning helps us decide, while the analytics of traffic accidents would only inform us to be aware.
There is a study about the relation between visiting primary care doctors often and the less likely the person to be hospitalized. Taking a step further, deep learning can study a user's behavior: how many visits we take to primary care doctors. With every visit, we get to know our body better, knowing if there’s a disease we need to recover from. Deep learning can change human behavior and decision-making.
Both cases show how deep learning is useful for customer experience. Previously we only had analytics and user testing. User testing here means that we ask for information from 8-10 users for their experience. This is also called qualitative testing. But with deep learning, we can scale the effect on many users.
Machine learning and deep learning are both parts of Artificial Intelligence. While machine learning studies training computers with statistical data, deep learning uses neural networks or multi-layered filters to process information. To implement both, you can start small in your organization by building the learning algorithms and later leverage them as you mine more data.
Designial offers app-delivery service with a fixed-price solution. Check our mobile-app development, UX research, and digital transformation. Contact us here for more information.
Editor's note: This post was originally published in July 2020 and has been updated for comprehensiveness.