What is the difference between artificial intelligence (AI), machine learning and deep learning?
AI is the ability for a machine to make a decision. The simple explanation is that AI works is that it takes data and then makes a prediction using that data. It learns to make a prediction using three ways: by learning from previous answers, by applying a mathematical formula to the answer, or typically using a combination of the two.
There are several use cases of AI which internally use either deep learning algorithms or machine learning algorithms in order to predict answers.
Computer vision detects information from an image or video. Today computer vision is used in self driving vehicles to detect option objects cars stoplights etc. Other computer vision use cases include gender detection, age detection, object detection and categorization. Companies are starting to use computer vision to identify what ads should be displayed based on your appearance or based on eye tracking movements.
Natural Language Processing (NLP)
Another type of AI is natural language processing (NLP). This is the ability to understand text-based information. Natural language processing is used to analyze information in document format and categorize information from those documents.This could be used to summarize information and is usually used to categorize a comment or statement and predict sentiment. It could be used to create answers to questions or conversational chatbots.
Natural Language Understanding (NLU)
Natural Language Understanding is the ability for a computer to understand language and respond. Common use cases for NLU are customer service bots or sales bots in which the company will try to engage you. One solution that is a great case study on NLU is X.AI which uses NLP and NLU to schedule meeting appointments.
To use X.ai, you CC your virtual assistant on the meeting request with some generic information such as “Can you schedule a meeting for next Thursday” and your AI powered assistant will contact the other party to negotiate times and automatically schedule it on your calendar. Learn more about X.ai here.
Machine Learning & Deep Learning (Detection & Prediction)
Machine Learning and Deep Learning detect patterns in your data to predict the result. Machine Learning is more math focused to predict an answer while Deep Learning is a little more abstract and is modeled on the way the human brain identifies patterns to make a prediction.
Both ML and DL can be used to solve the same types of problems but they solve them in different ways. Usually DL is used for computer vision problems and potentially for natural language problems.
An example of ML include predictive analytics or predicting churn – if a customer is going to cancel or not. Another example of machine learning is commonly used for customer segmentation. Customers are grouped into segments based on their similarities.
DL is technically a subset of ML but the key difference is when you push in the data, each data point is given a weight that determines what the prediction value. A higher weight determines a higher probability that this specific data point predicts the solution.
It’s often harder to interpret Deep Learning results because it is not transparent how the decision was made. The Deep Learning algorithm identified hundreds, thousands, or millions of permutations, and then creates algorithm behind the scenes.
Deep Learning and Machine Learning are both used for all areas of computer vision, natural language processing, natural language understanding, and detection and prediction.
Robotic Process Automation (RPA)
Robotic Process Automation is a script that automates tasks that are typically performed by a person. Today, these are generally repeatable and can be integrated into a Machine Learning or Deep Learning algorithm in order to make your bot smarter.
Robotic Process Automation is heavily used in the insurance industry to process claims and is heavily used for IT operational tasks such as account creation, password reset requests or security audits.
Today, RPA offers the most flexible way to automate digital tasks without redesigning processes and systems. They are designed to be a low-code way to automate tasks.
Levels of AI: Narrow, General & Super
There are three levels of AI: Narrow AI, General AI, and Super AI. All solutions today are considered Narrow AI solutions. They solve one specific task that they are given. Narrow AI is also called “weak” AI because they use predetermined data and algorithms to make a prediction.
Narrow AI is commonly used to automate a number routine tasks or tasks that require huge computational intelligence in order to respond. Even advanced technologies such as self-driving cars are considered Narrow AI because they chain together multiple Narrow AI solutions.
Artificial General Intelligence
The next level of AI is Artificial General Intelligence. AGI focuses on the ability to perform at the level of human intelligence and to chain together a complex series of actions. AGI is expected to be able to reason, solve problems, make judgments under uncertainty, plan, and learn over time.
Artificial Super Intelligence
Artificial Super Intelligence is expected to surpass human intelligence in all regards. Think HAL, The Terminator, The Matrix. This is the AI that people like Elon Musk believe will lead to the extinction of the human race.
While we are years away from Artificial Super Intelligence, people such as Ray Kurzweil and Elon Musk are focused on creating solutions where we can create superhuman intelligence using AI-powered brain implants.
I, for one, would love to get my super-intelligent brain implant (my husband will never be right again), but today, those technologies are in the not-too-distant future.