Everything You Need To Know About The Artificial Intelligence Boom23rd August 2018
If you’ve used Google Maps, you’ve experienced artificial intelligence ( AI ) firsthand. It’s a prime example of how AI technology today enables computers to take on tasks formerly reserved solely for humans — such as reading a map. In this case, Google uses historical and real-time data to visualize current traffic patterns and then applies AI to predict future traffic flow, with the objective to plot the quickest route to a destination.
Three important trends have made recent advancements in AI possible: big data collection, reduced computing costs, and improvements in algorithms.
Data these days are easy to collect and cheap to store, while the advent of cloud computing has made it much more affordable to crunch all that data. On the semiconductor side, graphics processing units (GPUs), the specialized chips used to accelerate AI, were prohibitively expensive just 15 years ago, but have now come down enough in price to drive greater adoption. Years of R&D have improved the performance of algorithms, enabling them to “learn” more from greater quantities of data.
At the heart of AI sits machine learning, the critical tool that powers computing intelligence without pre-programming. Machine learning at its core involves the ability to learn a function from data.
The first machine-learning technology dates all the way back to the early 1950s, when a data scientist built a reconfigurable maze that was used to help train a mouse to find the constantly changing exit. Over time, the mouse was programmed to follow the path it learned previously, and then adjust its path based on maze reconfigurations.
Around 1960, a cognitive scientist built the first neural network. Machine-learning algorithms use neural networks, which are self-learning programs made up of digital neurons arranged in interconnected layers that send and receive outputs to each other. Neural networks first need to be trained to recognize objects, and then are able to perform certain tasks.
By first ingesting large quantities of data, neural networks are over time able to learn without being specifically programmed. For example, a neural network would learn what an automobile looks like after recognizing repeated patterns when fed hundreds or thousands of images of cars. Eventually, the neural network is smart enough to be able to select images of autos entirely on its own.
Training is a complex procedure that requires lots of performance — that’s why GPUs are used. A trained neural network can recognize an image, detect a problem or make a suggestion.
The next stage after training is inferencing, which entails the computer acting on its own based on the results of the training. Inferencing is an important element in robotics and autonomous driving use cases. When it comes to self-driving cars, important inferences need to be made on the fly in real-time, using processors contained inside the vehicle.
The goal with machine learning is to model and predict various outcomes based on large data sets. It’s already possible to train software to recognize different patterns in data, measure the possible outcomes from the data and then make intelligent decisions based on those patterns.
In these early days of AI, the main goal is to improve the overall customer experience. Companies often deploy new technologies to retain customers and add new ones. Down the road, new revenue sources will be the primary goal of AI on the business side. IT research firm Gartner predicts the global enterprise value derived from AI in 2018 will total $1.17 trillion, up 70% year over year. By 2021, the firm estimates the value-add will total $3.3 trillion.
Machine learning is already being applied today for things such as fraud detection in banking and optimal route mapping in transportation logistics. Insurance companies deploy predictive risk modeling to gain a competitive edge, while doctors turn to machine learning to help improve patient diagnoses.
Coca-Cola (NYSE: KO ) has deployed Salesforce’s (NYSE: CRM ) AI-powered Einstein Vision technology for inventory management. Salesforce, a cloud-software giant, has applied AI technology to recognize and keep track of Coke bottles stored in cooler displays simply by analyzing photos taken with an iPhone, with no need for reps to make physical visits. Einstein Vision takes the Coke bottle stock count and combines it with seasonal sales patterns, weather data and planned promotions to automatically calculate restocking orders.
There are many potential uses for AI technology across various industries. For example, two insurance adjusters working on the same claim can come up with entirely different payout recommendations even though they both follow a set of basic guidelines. Adjusters typically work alone and don’t have constant supervision. By building an AI system based on historical data, insurance companies can create a more efficient system to monitor and direct adjusters, a move that could potentially reduce operating costs as well as payouts on claims.
AI is already being use today in call center optimization, with the goal to help people receive better customer service. First, thousands upon thousands of customer service call transcripts are loaded into the neural network to give it a rich data profile of what kinds of things people are calling in or emailing about on a daily basis. The data set is formed from various interactions across the entire customer service chain — including live chat, messaging and social media.
All the customer service data is then crunched to come up with a recommendation system. Going forward, when someone calls in or emails about a basic problem, the system can automatically respond by using an AI-based agent. For more detailed problems, human customer service reps can tap into the AI-powered system to see what the call is likely about and how to best handle the response based on historical and account-specific data.
IBM was recently awarded a patent for an AI-based traffic light management system that uses a network of processors and computer vision cameras. The adaptive system — it will change in response to traffic conditions — will monitor and control vehicle traffic and pedestrian flow in real-time (taking into account a multitude of factors) to help reduce congestion. If you’ve ever sat idling at a stop light for two minutes when there were no other cars around, you’ll be thankful once AI makes traffic management much smarter.
My #1 Pick For The AI Boom
My first AI pick for Extreme Tech Profits operates as a small vendor playing in a very big market. That combination translates into lots of growth runway.
Across its current customer base, the penetration rate is still in the low single digits, meaning there’s plenty of room for follow-on business. But what especially brought this pick to my attention was the steady climb of earnings expectations from the analysts since the company’s IPO. In fact, there have been five estimate increases since March 2017.
Raised guidance from the company and/or analysts is strongly correlated to forward price momentum, so this is a great sign going forward for this stock.
At the company’s most recent announcement on May 8, it reported a year-on-year revenue increase of 54%, over $3M higher than the best analyst estimate. The company also raised its 2019 outlook for both top and bottom line numbers. I expect to see big things from this pick, which is why we added it to our Extreme Tech Profits portfolio in our inaugural issue.
courtesy : NASDAQ