Deep learning is rapidly becoming an Enterprise AI (artificial intelligence) essential, at least according to Gartner.
The market research firm predicted that 80 percent of data scientists will include deep learning as part of their AI toolkits in 2018. Next year, the technology is poised to deliver “best-in-class” demand, fraud and failure predictions, all of which are very enticing prospects for businesses taking a data-driven approach to their operations.
But what exactly is deep learning?
In this explainer from Datamation’s Christine Taylor, deep learning is described as a subset of machine learning, which, in turn, is a subset of AI. In machine learning, supervised or unsupervised machine learning algorithms analyze data, “learn” from the results and apply their findings to generate predictions or make decisions based on the data patterns that emerge.
Deep learning, also uses algorithms, but goes a step further with artificial neural networks. Mimicking how the human brain works, deep-learning systems can ingest unstructured data (text, images, audio and video) and parse out the data to artificial neurons. While this occurs, processing layers build upon one another until a result is reached.
In practical terms, deep learning can mean the difference between marketing solutions that perform sentiment analysis based on the text-based content in emails and social media posts or identify emotional states from photos or videos in real-time. Deep learning is also putting the “autonomy” in autonomous vehicles and helping to pave the way for smart surveillance systems and medical systems that can pore through MRI scans for cancer markers without human input.
Although these scenarios seem to border on science fiction, the latest advances in deep learning are turning them into science fact for today’s businesses.
Deep Learning at Work
Tel Aviv-based Aidoc recently received the CE mark—the “Conformité Européenne” label that appears on products that conform to European standards—for a deep-learning solution that streamlines head and neck imaging workflows for radiologists. The startup’s technology can detect abnormalities in medical scans, saving time and improving patient care by helping hospitals prioritize cases that may require more immediate attention.
On Jan. 17 Babblabs announced that it had raised $4 million to further develop and commercialize its speech understanding and processing technology. Using deep neural networks, the San Jose, Calif. startup envisions faster, more precise voice-enabled interfaces and cloud services that go beyond the stilted and error-prone interactions associated with current virtual assistants.
In a Babblabs blog post, the company explored how its speech-focused deep-learning technology can help light up new services that leave today’s smart speakers and chatbots in the dust.
“Deep networks can follow not just the sounds, but the essential semantics of the audio trace, providing powerful means to overcome conflicting voices, audio impairments and confusion of meaning. The applications of improved speech understanding extend well beyond the applications we know today – smart speakers, cloud-based conversation bots and limited vocabulary device command systems,” stated the company’s staffers. “We expect to see semantic speech processing permeate into cars, industrial control systems, smart home appliances, new kinds of telephony and a vast range of new personal gadgets.”
Deep Learning for Enterprise Developers
AI startups aren’t the only ones delving into deep learning. Plenty of established IT companies are courting enterprise software developers with services that can be used to build intelligent business apps.
Amazon Web Services (AWS)
Given the amount of computing power required to make short work of large-scale AI workloads, it’s only fitting that the world’s leading cloud provider offer AI developers ways to tap into its massive infrastructure. On AWS, enterprises can use clusters of server processors and GPUs (graphics processing units) to accelerate the training of deep-learning models.
Developers can create applications with computer vision capabilities with Amazon Rekognition and speech processing with Amazon Transcribe. Amazon MXNet can be used to build recommendation engines similar to the product recommendations on the company’s ecommerce platform. Finally, Amazon Lex provides natural language processing that can detect the unspoken meanings and emotions behind words.
AWS’s cloud-based deep learning-innovations have already built up a strong following. Major brands, including Netflix, Pinterest and Edmunds, use the company’s technology.
Need a place to start?
Amazon also offers Deep Learning AMIs to help developers hit the ground running. AMIs, short for Amazon Machine Images, are preconfigured Amazon EC2 instances with popular deep learning frameworks, like Apache MXNet, Gluon, TensorFlow, Caffe and more. AWS Deep Learning AMIs are free to use, customers only pay for the resources they use.
Microsoft Azure Cognitive Services
Although its best known for its Windows PC operating system and Office suite of productivity apps, Microsoft has morphed into a cloud powerhouse in recent years, courtesy of its Azure cloud computing platform.
The Redmond, Wash. software giant offers developers an array of deep-learning APIs (application programming interfaces) dubbed Azure Cognitive Services. They include image processing, speech recognition, natural language processing, intelligent search and knowledge mapping that helps in building problem-solving apps.
The software giant also makes available an installable Microsoft Cognitive Toolkit, enabling developers to train deep learning algorithms on their local machines (64-bit Windows and Linux) and experiment before they launch full-blown AI workloads in the cloud. The toolkit is available on GitHub.
Google Cloud AI
Google has been pioneering intelligent search and applications that deliver the results users are looking for, more often than not. As another major cloud provider, Google has turned the deep learning research that powers its products into a suite of services the company calls Cloud AI.
Based on the company’s own neural net-based technologies, Cloud AI allows developers to build and train large scale machine-learning models or infuse their apps with speech recognition, translations services and more.
Businesses can also use Cloud AI to extract more value out of their video content with the Cloud Video Intelligence API, which “uses powerful deep-learning models, built using frameworks like TensorFlow and applied on large-scale media platforms like YouTube,” explained Fei-Fei Li, chief scientist at Google Cloud AI and Machine Learning, in a blog post. “The API is the first of its kind, enabling developers to easily search and discover video content by providing information about entities (nouns such as ‘dog,’ ‘flower’ or ‘human’ or verbs such as ‘run,’ ‘swim’ or ‘fly’) inside video content,” continued the Google staffer.
And as of December 2017, it’s cheaper for the company’s 10,000-plus paying Cloud AI customers to use the Google’s deep learning services thanks to reduced cloud GPU costs and other price reductions.
Of course, no discussion of modern AI is complete without IBM Watson.
After beating human opponents at Jeopardy, IBM’s AI drew mainstream attention and sparked a race to commercialize machine learning and deep learning technologies. Beyond competing on TV game shows, Watson has since been helping financial firms with their regulatory obligations and enabling healthcare organizations to detect and treat diseases faster and more accurately.
IBM Watson, available as a service on the IBM Cloud, is now home to Big Blue’s ever-evolving cognitive computing portfolio. Cognitive computing is a blanket term for AI that builds on the company’s work on deep learning and neural networks. Developers can integrate Watson into their own cognitive apps or even embed Watson into third-party apps like Salesforce.
In the coming years, businesses can expect even more Watson innovation.
IBM is pouring $240 million into the new MIT-IBM Watson AI Lab over 10 years. The lab seeks to advance deep learning hardware, software and algorithms, plus find ways of using AI to solve challenges in healthcare, cybersecurity and other industries.
Pedro Hernandez is a contributing editor at Datamation. Follow him on Twitter @ecoINSITE.