<img alt="" src="https://secure.tula9mari.com/150530.png" style="display:none;">
GET DEMO

LEVERTON Blog

AI, PropTech and more...

Subscribe

The 3 Deep Learning Trends For 2017

December 21, 2016 - By LEVERTON Team

The term “deep learning”, often unfairly summarized as machines able to beat humans in board games, has been on most people’s lips throughout the past year at one point or another. Although the term sometimes is used synonymously with “artificial intelligence” and “machine learning”, it is in fact a subset of the two. It is an area of machine learning research, introduced with the aim of moving machine learning closer to its original goal – true artificial intelligence.

 

the-place-of-deep-learning-in-the-space-of-artificial-intelligence

Image: The place of deep learning in the space of artificial intelligence

 

Reports estimate that by 2024, the total market for deep learning software, services and hardware will exceed $100 billion in annual revenue. The software for enterprise applications of deep learning alone is expected to rise from $109 million in 2015 to $10.4 billion in annual revenue in 2024.

Despite the fairly recent peak of interest, the first deep learning like algorithms actually date back to the mid 60’s. The use cases for those algorithms were, perhaps not surprisingly, very limited, but with ever increasing training abilities, deep learning technology is improving significantly. Now, it is starting to face serious commercial use. Below we have listed three expected trends within the area of deep learning and the application of these technologies.

 

1 - Deep learning for smart devices

Use cases for deep learning familiar to most people are to be found within B2C applications. Consider the following: you’re in China visiting a restaurant which doesn’t have an English menu. On top of this, none of the staff members speak English. So you could either close your eyes, point at something on the menu and hope it turns out to be a good choice. Or you could take out your smartphone, take a picture of the menu and let a neural network translate it for you to be able to make a better informed decision on what to enjoy for dinner.

Another example of how deep learning can simplify our every day life is if you come across a picture of a shirt you like and you would like to have a similar one. Instead of searching for terms such as “shirt”, “green” and “button down”, you simply upload the picture and use that to search companies’ products for a matching or similar item.

And with increased resources put into deep learning this is only the beginning. Automated assistance and virtual guides will improve. Self-driving cars are already reality and will become much more common. In 2017 Audi’s new A8 will incorporate fully autonomous technology and Uber CEO, Travis Kalanick, expects Uber’s fleet to be driverless by 2030.

In only a few years’ time we will probably consider Siri’s current performance as poor in comparison to what such services will be able to deliver by then.

 

2 - Deep learning for structured data creation

You might have heard the statement “data is the new oil”. However, to be able to capitalize and make use of data, solely having access to it is not enough. Without a structured set of data, quantitative analysis of it becomes, if not impossible, at the very least difficult and time-consuming.

Organizations today both generate and have access to vast amounts of data, but much of it is unstructured in the shape of emails, files, etc. Without a smart platform for gathering, processing and reviewing the data, organizations will fail to make real data driven decisions. This is where deep learning comes in. As mentioned earlier, the technology can be used to automatically review text, so called natural language processing. Deep learning is very powerful in this instance as the algorithms do not only read each word in a document, but also take the context of the surrounding words into consideration, delivering a more accurate and reliable result.

With less need for manual data aggregation and document reviewing more time can be spent on processing, analyzing and acting upon the data. By doing so, organizations can go from unstructured data to information and insights. Insights which can be used for smarter decision making and competitive advantage.

 

3 - Deep learning for business transformation

Applied in a corporate context, deep learning can be very powerful and will drive transformation of work processes – and in the long run of entire occupations. Much of this rises from the fact that deep learning technologies can be used for reviewing text and images, meaning that documents can be reviewed by a machine instead of humans.

By eliminating tasks such as manual data aggregation and document review, organizations will save cost and time. Again, spending less time on manual review of documents or contracts means more time can be spent on making sense out of the data. As a result, organizations can streamline processes.

An example is LEVERTON’s software. By applying deep learning technologies, the software recognizes and reads out terms from corporate documents. Once the relevant data is extracted, it can be synchronized and compared with the data from other systems such as SAP. Having accurate and up-to-date data helps to significantly simplify and accelerate organizational processes such as annual reviews or due diligences.

 

While deep learning and the use of it already is quite spread, it is still in its early days of development. Significant improvements is to be expected and future models will be able to learn from much less training, increasing the accuracy and speed in the areas where it is already applied and additionally infiltrate other areas such as medicine and robotics.