Deep learning implementation in enterprise
Introduction to DL
Add bookmark
What is Deep Learning?
The term “deep learning” has firmly worked its way into the enterprise lexicon. Deep learning is a class of machine learning algorithms that uses a deep neural network in order to learn. A deep neural network is a collection of artificial neurons, which are simply trainable mathematical units. These neurons collectively “learn” complex mathematical functions to map raw input to an output. Neural networks have existed in machine learning since the 1950s, but today’s “deep” networks have many more stacked layers of neurons than older networks. This enables them to learn much more complex problems requiring computation at a scale that was previously not possible. A typical deep neural network has about eight layers and 60 million parameters, but there are examples of deep neural networks today that have 200 to 400 million parameters.
In this report we explore the true business possibilities and challenges of Deep Learning and the best way to understand how enterprises can leverage deep learning is through use cases. First, let’s jump into how Deep Learning actually works.
How does Deep Learning work?
Deep Learning involves feeding a computer system a lot of data, which it can use to make decisions about other data. This data is fed through neural networks. These networks are made up of logical constructions which ask a series of binary true/false questions or extract a numerical value of every bit of data which pass through them in order to classify it according to the answers received.
Deep Learning work is focused on developing these networks, they become what are known as Deep Neural Networks – logic networks of the complexity needed to deal with classifying datasets as large as, say, Google’s image library, or Twitter’s feed of tweets.
With datasets as comprehensive as these and logical networks sophisticated enough to handle their classification, it becomes simple for a computer to take an image and state with a high probability of accuracy what it represents to humans.
Pictures present a great example of how this works, because they contain a lot of different elements. It isn’t easy for a human to grasp how a computer, with its one-track, calculation-focused mind, can learn to interpret them in the same way as a human. But Deep Learning can be applied to any form of data – machine signals, audio, video, speech, written words – to produce conclusions that seem as if they have been arrived at by humans – very, very fast ones.
ALSO READ: DX Talent
Let's look at a practical example on the next page.
The key for global corporate enterprise is to benefit from the collective intelligence presented by RPA and cognitive technologies along with human workers. Only by having technology combine with human talent can global corporate enterprise achieve scalable intelligent automation. And only with scalable intelligent automation enterprise resiliency be realized. Join that community for lessons learned at SRIA Live.Scalable RPA & Intelligent Automation Live
A Deep Learning example
Take a system designed to automatically record and report how many vehicles of a particular make and model pass along a public road. First, it would be given access to a huge database of car types, including their shape, size and even engine sound. This could be manually compiled or, in more advanced use cases, automatically gathered by the system if it is programmed to search the internet and ingest the data it finds there.
Next it would take the data that needs to be processed – real-world data which contains the insights, in this case captured by roadside cameras and microphones. By comparing the data from its sensors with the data it has “learned,” it can classify, with a certain probability of accuracy, passing vehicles by their make and model.
So far, this is all relatively straightforward. Where the “deep” part comes in, is that the system, as time goes on and it gains more experience- can increase its probability of a correct classification, by “training” itself on the new data it receives. In other words, it can learn from its mistakes -just like us. For example, it may incorrectly decide that a particular vehicle was a certain make and model, based on their similar size and engine noise, overlooking another differentiator which it determined had a low probability of being important to the decision. By learning that this differentiator is, in fact, vital to understanding the difference between two vehicles, it improves the probability of a correct outcome next time.
Key Insights of Deep Learning in Industries
Overall adoption of artificial intelligence is increasing year over year among businesses. According to a RELX report, adoption of AI in business increased from 48 percent in 2018 to 72 percent in 2019. Senior executives know that AI isn’t just hype. Organizations across sectors are looking closely at the technology to see what it can do for their business. McKInsey’s Frontier Study estimates that 40% of all the potential value that can created by analytics today comes from the AI techniques that fall under the umbrella “deep learning”. In total, McKinsey estimates deep learning could account for between $3.5 trillion and $5.8 trillion in annual value.
Enterprise AI Adoption
However, many business leaders are still not exactly sure where they should apply AI to reap the biggest rewards. After all, embedding AI across the business requires significant investment in talent and upgrades to the tech stack as well as sweeping change initiatives to ensure AI drives meaningful value, whether it be through powering better decision-making or enhancing consumer-facing applications.
Enterprise AI Investment
Following an examination of more than 400 actual AI use cases across 19 industries and 9 business functions, it seems AI has the most potential in the areas that have traditionally brought in the most amount of money for businesses. The business areas that traditionally provide the most value to companies tend to be the areas where AI can have the biggest impact. In retail organizations, for example, marketing and sales has often provided significant value. McKinsey notes that using AI and Deep Learning on customer data to personalize promotions can lead to a 1-2 percent increase in incremental sales for brick-and-mortar retailers alone. In advanced manufacturing, by contrast, operations often drive the most value. Here, AI can enable forecasting based on underlying causal drivers of demand rather than prior outcomes, improving forecasting accuracy by 10-20 percent. This translates into a potential 5% reduction in inventory costs and revenue increases of 2-3 percent.
Areas Where Deep Learning Provides the Most Bang for Buck
After surveying 200+ digital transformation leaders on the matter, we compiled their input into our latest, in-depth report: DX Talent Reimagined. Download now and discover:DX Talent Reimagined: 2022 Industry Report
Download Now
While applications of AI cover a full range of functional areas, it is in fact in these two cross-cutting ones where AI can have the biggest impact, at least for now, in several industries: supply-chain management/manufacturing and marketing/sales— Combined, it’s estimated that these use cases make up more than two-thirds of the entire AI opportunity.
In manufacturing, the greatest value from AI can be created by using it for predictive maintenance, reducing costs by 12 percent, per PWC’s Beyond the Hype report. AI’s ability to process massive amounts of data including audio and video means it can quickly identify anomalies to prevent breakdowns, whether that be an odd sound in an aircraft engine or a malfunction on an assembly line detected by a sensor. AI can create $1.4-$2.6 trillion of value in marketing and sales across the world’s businesses and $1.2-$2 trillion in supply chain management and manufacturing per McKinsey’s Two Areas Report.
Another way business leaders can hone in on where to apply AI is to simply look at the functions that are already taking advantage of traditional analytics techniques. It has been found that the greatest potential for AI to create value is in use cases where neural network techniques could either provide higher performance than established analytical techniques or generate additional insights and applications. This is true for 69 percent of the AI use cases identified in the McKinsey Frontier Study. Only 16 percent of use cases found a “greenfield” AI solution that was applicable where other analytics methods would not be effective.
Also read: Enterprise AI
Deep Learning Models (With real life examples)
There are 2 major models that have proven to be most promising for businesses.
The first: Convolutional neural networks (CNN’s)
Per Toward Data Science’s Guide, a multilayered neural network with a special architecture designed to extract increasingly complex features of the data at each layer to determine the output.
- Cisco
CNN model can be used if you have unstructured data set (e.g., images) and you need to infer information from it. One of the most promising uses of CNN powered – ‘visual recognition’ – has been in the automotive industries, particularly for the adoption of self-driving cars. The reason behind these machines includes various reasons like decreasing the rate of road accidents; follow traffic rules and regulations in order, etc.
- Hyundai
For example, Cisco announced its partnership with traditional car manufacturing company, Hyundai to help enable over the air updates for autonomous cars with an estimated savings of $3.5 billion over the four years post-implementation per CB Insights.
- HSBC
Per Find Biometrics, banks like HSBC are leveraging deep learning in the form of ‘image recognition’ to allow customers to log in to their apps with minimum effort. These banks also use image recognition to allow customers to simply take pictures of checks to be able to deposit money.
Deep Learning Models (With real life examples cont.)
There are 2 major models that have proven to be most promising for businesses.
The second: Recurrent Neural Networks (RNC’s)
A multi-layered neural network that can store information in context nodes, allowing it to learn data sequences and output a number or another sequence. The type of data that is necessary for RNC’s to work are time-series data or sequences (e.g., audio recordings or text). The 2 most common sub-divisions of deep learning that use RNC model are Natural Language Processing & Speech Recognition. Many companies use chatbots that use natural language processing to help with their customer service department.
- JP Morgan Chase launched COIN, a chatbot to manage its back-office operations. It can analyse complex contracts quicker and more proficient than human lawyers. The Chabot use cases in the legal industry are discovered in unimaginable ways. Legal jargon is a language of its own and piles up every day, across multiple document structures. Analysing these documents is a time-consuming process for humans.
- Shell, a leading British-Dutch oil and gas company has introduced an AI chatbot – LubeChat for its B2B Lubricant customers. The bot provides real-time customer support by providing information about the company products, services lubricants data etc. RNC models are also used for, yet another sub-branch of deep learning called ‘Predictive Analytics’.
- Coca-Cola launched a data strategy by building a digital-led loyalty program. In fact, big data analytics is strongly behind customer retention at Coca-Cola. Per Stefan Leadbetter, the company counts wins in vending machines through AI-inspired location-based recommendations, social media through analysis of 12,000 posts and proof of purchase through image recognition.
Conclusion
Research institutions and tech companies have made massive progress in certain areas of machine learning, including computer vision, speech recognition, and natural language processing. Still, this technology is not a silver bullet. For now, though, most of the news is coming from the suppliers of ML technologies. And many new uses are only in the experimental phase. Few products are on the market or are likely to arrive there soon to drive immediate and widespread adoption. As a result, analysts remain divided as to the potential of ML: some have formed a rosy consensus about ML’s potential while others remain cautious about its true economic benefit. This lack of agreement is visible in the large variance of current market forecasts, which range from the hundreds of millions to couples of billions. Given the size of investment being poured into ML, acting quickly is important but preparing to act is paramount.
