Artificial intelligence can make your business smarter and more efficient, and cloud computing can make AI easier and cheaper to implement.
Artificial intelligence has fascinated mankind for centuries. The concept dates back to Greek myths of golden automatons crafted to assist the gods, medieval pursuits to alchemically force life into unliving matter, and countless novels of the 19th century and beyond.
It can be difficult for us to wrap our minds around the reality of AI. The historically fictional nature of AI coaxes us into a natural tendency to see it as an abstract topic—a dream of the future.
Now, though, it’s tangible. You can use AI today. Right now.
Granted, creating and using AI isn’t trivial. The process involves complex mathematics and programming. However, recent innovations in cloud computing by large companies like Amazon, Microsoft, and IBM have shifted the bulk of the work away from smaller businesses.
For companies like yours, accessing the cutting-edge power of AI is now as simple as calling an API.
So how can you implement AI to help your business? This post will help you understand the state of the AI market, why AI has historically been difficult to implement, how cloud computing is making AI accessible, what companies provide cloud AI, and some innovative applications of cloud AI.
- Understanding AI as a technology of today
- Why AI has been inaccessible to many businesses
- How cloud computing is making AI accessible
- Applying cloud-delivered AI to your business
- Choose from many cloud AI service providers
- Build your own AI applications using cloud infrastructure
Understanding AI as a technology of today
AI hasn’t quite reached sci-fi levels of complexity, but it’s no longer just a technology of the future. Cloud computing is helping to change the way we think about AI technology by making it relevant today.
One of the first challenges that AI has to overcome is public perception. Sci-fi movies and novels have taught us that AI is supposed to be nearly indistinguishable from the human mind. The AI that is available today is much simpler than that, but it is a stepping stone to that brain-level complexity.
In general, AI is software that mimics intelligent responses (i.e. human responses) to data.
This includes general-use AI like Apple’s Siri, Amazon’s Alexa, or Microsoft’s Cortana, as well as more specific applications. A mobile app might use Amazon Polly to convert a text-based interface into lifelike speech that can be spoken through a smart device. A social website might use Microsoft’s Computer Vision API to process hundreds of thousands of user-uploaded images and strategize around that data.
In other words, the current state of AI isn’t about simulating human life. Rather, its purpose is to automate tasks that computers can accomplish with a degree of efficiency that humans cannot.
There is strong demand for AI
Simply put: Businesses want AI. They want AI to improve the way they operate and help keep them ahead of their competition.
According to the Economist Intelligence Unit’s 2016 survey (PDF download), business owners believe the greatest benefit of AI is its ability to improve efficiency. This could mean using AI for tasks like predictive maintenance, product design, or streamlining logistics.
Executives also anticipate that AI will solve the current shortage of data science talent. Demand for data scientists is increasing, and will exceed supply by more than 50% by 2018. AI fills the gap by facilitating the processing of big data. A 2016 survey by Narrative Science found that business and tech executives who were using AI technologies reported higher confidence in their ability to use big data.
Businesses want AI to help them visualize, analyze, and strategize around large sets of data.
Businesses are trying to adopt AI
Optimism about using AI as a solution to business problems has encouraged quick adoption. Many companies have either deployed AI tech, or plan on doing so in the near future.
Of the 230 executives surveyed by Narrative Science, 41% said deploying AI technologies is a priority, 23% plan to do so by the end of 2017, and 56% plan to do so by the end of 2018.
Businesses are eager to start using AI. So, what has been keeping AI from seeing widespread adoption?
Why AI has been inaccessible to many businesses
It’s not too surprising that AI is a subject of excitement for many businesses. Whether you look at it from the perspective of the fantastical robots in movies or the practical technology of today, AI is exciting.
Though AI seems to be on the rise, there are a few barriers hindering its progress.
Many businesses can’t afford AI
Implementation cost is the number one thing holding back most companies from AI.
Deep learning, for example, requires a prohibitively expensive amount of computing resources. Most businesses can’t afford to build and power all of the local infrastructure necessary to train a neural network.
Many businesses are confused by AI
A more abstract barrier to AI, though still significant, is the widespread misunderstanding of what exactly AI accomplishes. Even given the definition above, it’s possible to get confused about the exact role of AI in a business.
Narrative Science helped demonstrate this in their survey. While only 38% of executives said they were using AI in the workplace, 88% said they were using technology that relies on AI (e.g. predictive analysis and voice recognition). That is a large disparity for a rather trivial semantic difference.
So, we’re able to see that there is some confusion over the definition and applications of AI. However, we can also see that there is general agreement over its importance. Given AI’s traditional place in our culture, it’s understandable that it may seem intimidating, confusing, or even worrying.
Many businesses don’t have enough data for AI
Data is essential to deep learning. A Microsoft Research experiment found that the performance of AI training algorithms increased significantly when the amount of data used was increased, to the point where a bad algorithm could become a great algorithm just by introducing more data.
Large companies like Google and Facebook are at a great advantage in AI not just because of their infrastructure, but because they generate massive amounts of data every day. Peter Norvig, Research Director at Google, sums up this advantage nicely:
“We don’t have better algorithms than anyone else. We just have more data.” (CNET)
If you own a smaller company, you might have a much harder time creating useful AI.
Fortunately, cloud computing has introduced practical ways to overcome many of the issues that have limited the growth of AI. In fact, advances in AI technology are accelerating thanks to the services now available via the cloud.
How cloud computing is making AI accessible
Cloud computing is a growing field that is creating new opportunities for businesses. Its use and acceptance is growing among companies of all sizes due to its many benefits. Currently, enterprises are transitioning from building their own IT to consuming cloud-based IT. This is all to say that cloud computing is a great companion for AI.
Cloud computing lowers the cost of accessing AI
Given that implementation cost is such a large problem for many businesses contemplating AI, cloud computing’s low cost is a huge contributing factor.
Services delivered via serverless architecture allow companies to pay for only the computing power they use. Services like AWS Lambda ensure that your IT budget get used in one of the most cost-efficient ways currently available.
This means businesses can make use of, for instance, Amazon’s AI infrastructure without having to pay for constant server uptime. Amazon is a huge company with huge infrastructure, which means it can put a relatively low price tag on the large amounts of computing power required to run AI.
Ultimately, this all means less time and money spent worrying about how to power your AI.
Cloud computing helps define AI and its capabilities
Much like cooking without a cookbook, working with AI is much more intimidating when resources and possibilities aren’t pre-defined. When you don’t know what a technology is even capable of, you’re not likely to know how to put it to work.
Fortunately, several cloud companies offer cloud-delivered services which pre-package AI’s capabilities.
Amazon, Microsoft, IBM, and Google are just the largest of the companies currently offering cloud-delivered AI services (more on their specific offerings later).
These services take the guess work out of how AI can help your business. They offer inspiring case studies, grounding limitations, and welcoming documentation.
Cloud computing delivers open data to make AI smarter
In 2012, Google conducted a deep learning experiment in which they trained a neural network to recognize both human and cat faces. To train the AI, Google used 10 million images extracted from YouTube videos. What company besides Google has the resources and computing power to access that amount of data? Facebook, maybe, but not many others.
Four years after publishing the report on their data-devouring experiment, Google released a dataset of millions of tagged photos and videos for the specific purpose of providing the machine learning community with data they could use to build and experiment. This perfectly exemplifies the concept of open data.
Open data is the idea that a particular set of data belongs to no one and everyone—it is freely available without concern for copyright or any other legal restrictions. For AI and machine learning to continue growing, more entities are going to need access to large sets of data. Open data offers a solution.
Cloud-delivered services are perhaps the greatest hope for the accessibility and adoption of AI.
Applying cloud-delivered AI to your business
The marriage of cloud computing and AI is shaping up to be a disruptive force across many industries.
Transparency Market Research predicted that the “machine learning as a service” market will increase from $1.07 billion in 2016 to $19.86 billion in 2025. This relationship not only brings a new degree of accessibility to AI, but it creates a new way of thinking about other existing technologies and methodologies.
The human element of customer service makes it a constant concern for both businesses and customers. AI’s ability to understand language presents new customer service solutions like automated personal assistants and chatbots. All of the technology needed to power these solutions are available through cloud services like Amazon Lex chatbots and Facebook Messenger’s chatbot APIs.
Microsoft CEO Satya Nadella has named this particular implementation of AI “conversation as a platform”. He believes the transition to AI conversational interfaces could be more disruptive than the role of the touchscreen in the smartphone revolution. Google CEO Sundar Pichai agrees, stating:
“In the long run, we’re evolving in computing from a ‘mobile-first’ to an ‘AI-first’ world.” (Business Insider)
If the smartphone revolution gave way to companies like Snapchat and Uber, what types of businesses might be born from AI in the cloud, and what are some applications that might apply to your business? We can already observe some of the growth taking place as a result of this relationship.
Vision and image recognition
Deep learning gives AI the ability to recognize images in a way that is similar to how our eyes and brain allow us to see. This makes it possible for AI to have vision that is equivalent to or better than human vision.
There are AI services for identifying the contents of an image, like people, animals, and objects. Similarly, AI can classify images with certain labels that make it much easier to sort through a large number of images.
This basic concept has been expanded into other, more specific services. Multiple APIs offer facial recognition, allowing programs to identify whether a certain person is present in an image. Also, when you consider that visual AI are able to process many images, it makes sense that they would also be able to process the frames of a video. This makes way for services like AI video editing and indexing.
Conversation recognition and automation
For at least a few years, there has been AI capable of passing the Turing test, which tests whether an AI is able to converse in a way that is indistinguishable from a human. While this doesn’t guarantee that any AI will be able to take the place of a company rep, it does potentially offer opportunities for your business.
Speech recognition services allow AI to identify a speaker by their voice and convert speech into text that is usable by an application. Natural language processing makes it possible for AI to understand regular, human speech rather than robotic commands. Translation services translate text and speech in real-time with AI that is optimized for conversation.
One of the greatest promises of modern AI is its ability to predict outcomes and facilitate strategizing around data. While many applications of this kind of AI will be specific to the business they’re built for, there are at least a couple of more general example available.
Recommendations AI services help you to predict which parts of your store or interface are most useful to your customers and adjust your UI/UX accordingly. Search AI services use your existing content to provide the best answers to customer queries.
These are just some general examples of some of AI’s capabilities that are easily accessible through cloud-delivered services. For a more specific view, we can take a look at some of the services presented by companies like Amazon, Microsoft and IBM.
Choose from many cloud AI service providers
Amazon has over a decade of experience managing cloud infrastructure, putting them in an excellent position to introduce their own cloud-delivered AI services.
Amazon AI allows developers to build applications with AI services like speech recognition with Amazon Lex and image recognition with Amazon Polly. Alternatively, you could use existing data to train your own AI using Amazon Machine Learning. When your AI is ready to go out into the world, Amazon EMR can help you deploy and scale your AI at an affordable cost.
Microsoft Cognitive Services
Microsoft Azure, while perhaps behind Amazon Web Services in cloud market share, seems ready to keep pace in AI services.
Microsoft Cognitive Services offers machine learning packaged as 24 user-friendly APIs. These APIs can make your product see, hear, speak, think, and otherwise provide unique value to your customers.
IBM’s jeopardy-winning Watson is the face of their AI services. IBM Watson claims to “turn unstructured data into decisions and actions.” It allows businesses to create conversational interfaces that can stand in for traditional customer service representatives. It offers 13 APIs that, at least on the surface, are comparable to Microsoft’s offerings.
Google Cloud Machine Learning
Google promises better AI training performance and accuracy than other deep learning systems, and its treasure trove of data helps. Google’s Machine Learning Services highlight the ease with which you can build your own machine learning models to bring AI to your business.
Given that the above companies command a large portion of the cloud market, it is understandable that they would have an early lead in the cloud-delivered AI market. But there are others.
For example, Facebook contributes deep learning modules to open-source computing framework Torch. Newer companies like BigML and TenPoint7 are targeting smaller business with their machine-learning services.
The novelty and momentum of the cloud-delivered AI market provide businesses many options and many opportunities to experiment.
Build your own AI applications using cloud infrastructure
Currently, the greatest value of cloud-delivered AI services lies in their ability to enhance existing products and processes. This means businesses should research, brainstorm, and experiment to find ways to use the available services to bring value to their customers.
Fortunately, the infrastructure and computing resources provided by cloud computing allows companies to build and use AI with minimal resistance.
For instance, Amazon’s machine learning service makes it possible to apply AI to your data. Essentially, this means you are able to use the same algorithms that Amazon uses in their products to enhance your applications.
Once you store data in the AWS cloud, you can build and train AI models that will, for example, forecast demand of one of your products or anticipate the behavior of your customers. It’s a technology that holds an exciting amount of power and potential. As Matt Wood of the AWS leadership team puts it,
“Once customers start to apply predictive models to their data, it becomes addictive.” (Financial Times)
Let’s take a look at how companies have used AI to improve their products and processes.
Amazon Machine Learning generates roof-age prediction API for insurance companies
BuildFax is a company that provides services around housing and construction based on aggregate data. Their core customer base is insurance companies, who rely on BuildFax’s estimates to establish policies and premiums. One common estimate BuildFax provides is that of roof age and condition.
The problem that BuildFax faced was in their method of creating their estimates. They initially created predictive models based on zip codes and other non-specific data. It was a complex, slow process that didn’t yield sufficiently accurate results.
Amazon Machine Learning provided a way for BuildFax to offer more accurate estimates in less time.
“Models that previously took six months or longer to create are now complete in four weeks or less.” (AWS)
The models were trained via Amazon ML using tens of millions of high-confidence roof ages and property characteristics from public sources and from customers. From there, BuildFax was able to offer an accurate, real-time prediction API to its customers.
Not only did Amazon ML drastically improve BuildFax’s existing prediction service, but it opened the door to new services like data analytics and cost estimates.
C-SPAN uses Amazon Rekognition to index their content
As a TV network with multiple stations covering unedited US government proceedings and other public affairs, C-SPAN has a ton of content. In order to make that content useful after it is broadcast, C-SPAN must make it all searchable by indexing who is speaking and/or on camera at any given time. That’s a big job.
Using Amazon Rekognition, C-SPAN is able to automatically tag the speakers in a video. Rekognition has more than doubled C-SPAN’s indexing speed, increasing their rate from 3500 hours per year to 7500 hours per year. Alan Cloutier, C-SPAN Archive’s Technical Manager, praised Amazon’s product:
“…it was shockingly easy to set up, even with 97,000 entities from our database.” (AWS)
Farmer uses Google’s TensorFlow machine learning to sort cucumbers
Makoto Koike was an embedded systems designer who wanted to take over his parents’ cucumber farm. He was surprised to find how laborious and potentially expensive it was to sort cucumbers.
Cucumbers can be differentiated by their color, shape, quality, and freshness. The job of identifying each of these traits originally belonged to Makoto’s mother. That process alone took her up to eight hours per day to complete. It was a system that took months to learn, making it impossible to simply hire new workers during busy periods.
Makoto decided it would be wise to automate the sorting process before taking over his parents’ business.
Using Google’s TensorFlow library, Makoto trained a neural network to recognize and sort cucumbers. By providing images and cucumber classifications, he trained an AI to recognize certain features of a cucumber. From there, he used Arduino and Raspberry Pi to control the physical operations of a conveyor belt sorting system and to send data to a Google Cloud server running TensorFlow.
After completing the first iteration of his AI project, Makoto’s results were encouraging, but far from perfect. This was due in large part to the amount of data required to create an accurate model.
When validating his models against test images, recognition accuracy exceeded 95%. However, in actual use, the accuracy was closer to 70%. Despite the fact that Makoto spent three months taking 7,000 pictures of sorted cucumbers, his model still didn’t cover a large enough variety of situations. Also, the model took around three days to train on a Windows desktop PC even though Makoto’s images were only 80 x 80 pixels in resolution.
Looking forward, Makoto considered using Google’s Cloud Machine Learning to train his models faster using higher-resolution images and more images. Over time, he could make his model more accurate and save a significant amount of time, effort, and ultimately money for his inherited business.
Your business can start using AI right now
Thanks to cloud computing, AI technology is available to your business right now. It’s not a myth, it’s not just an idea; it’s real, and it’s usable. You’ve seen the statistics, you’ve read the stories—all that’s left is to try it out for yourself.
Just to be clear, though: AI hasn’t fully arrived. We’re not yet to the point of handing off significant responsibilities to digital brains. There are still many challenges ahead for AI, but there’s only one time-tested way of overcoming them: Experimentation.
If you’re interested in taking a closer look at cloud-delivered AI services, the best way to get started is by reading documentation. Read up on how Amazon Rekognition works, or check out what you can do with Microsoft’s Computer Vision API.
The cloud and AI are transforming how we interact with our world. Taking the time to figure out how they might transform your business could be a revolutionary decision.
Which of your business processes can cloud AI help make more efficient? We’d love to hear your thoughts in the comments.