Drones are winning the market, and the key...
Cloud + AI = Next Tech Revolution?Today there is a lot of talk about the convergence of cloud computing and AI technologies leading to the next tech revolution. Sounds familiar? There was similar hype about mobile technologies that would revolutionize the cloud, or IoT that would bring cloud software development to a way new level. In fact we see that they have influenced the cloud, but “ revolution” would be a big word for this.
And now another buzzword – AI. Should we stay skeptical to those shifts Artificial Intelligence would bring to the cloud – or try to see the potential of this convergence and leverage it before the rest of the market does?
Top Benefits AI & Cloud Business Bring to Each OtherThe current state of AI technologies doesn’t resemble sci-fi movies yet, where Artificial Intelligence simulates human life in a form of a self-aware robot. Its capability is of a different sort: it can automate tasks which computers will accomplish much more efficiently than human beings. And tech-savvy business owners know it. The Economist Intelligence Unit’s survey showed that they see AI’s greatest benefit in improving efficiency: predictive maintenance, automation of key business processes, product design. Here are some more stats from Deloitte:
- AI early adopters are getting a 16% increase in ROI today.
- They are expecting major ramifications from AI implementation in the next 2 years.
Key AI Implementation Problems – and Ways the Cloud Helps Overcome Them
AI technologies development costsWith the rise of ML-oriented tools, libraries and frameworks, development of AI tools is getting more affordable (for example, development of, say, an insurance fraud detection tool is evaluated at $100K-300K). But for many companies these costs are still hard to bear.
How the cloud tackles it: The main cloud managed services on the market are offered by major tech companies who have budgets and human resources and can invest into grand development projects, experiment with new features and take care of their constant development and improvement. A user pays a fee for getting access to the infrastructure and its use, which considerably lowers the barrier of entry.
Lack of AI specialistsFor a company having budgets for integrating an AI-based technology into its workflow, another problem rises: who will do the tech part of the task? Typically, ML and DL algorithms, their delivery and implementation requires a high level of skills from developers. It becomes a challenge to find a reliable software development services vendor or hire tech specialists, who will face this challenge.
How the cloud tackles it: Today cloud providers offer both types of AI services:
- Off-the-shelf solutions, which can be easily implemented into a current workflow. They don’t request from in-house tech specialists learning how to work with complex ML algorithms, and provide for a quick time-to-market solution. These products make AI introduction into business processes as seamless and quick as possible.
- Reference architectures, which can be used for building a custom AI solution on top of it and help deliver unique, business-specific solutions. Working with them requires highly qualified tech resources and certain time for development and testing. In the end a client get a totally customised solution.
AI-related misunderstandingMany businesses simply don’t understand AI. No, they definitely have a notion of what Artificial Intelligence is, but they are still far from an idea of how they can make use of it.
How the cloud tackles it: Providing documentation, case studies and white papers is one of the ways cloud providers define how businesses can leverage AI in their activity. Another way is giving individual consultations or organizing special events (like Let’s Talk AI by Google Cloud).
Lack of dataBuilding an AI algorithm is only the first step. Another one is to ensure its development and evolution. Algorithm functioning is getting more exact when trained with data and learning from a variety of patterns. Small to medium sized companies usually don’t have enough data for this.
How the cloud tackles it: Peter Norvig, Research Director at Google, made a good statement: “We don’t have better algorithms than anyone else. We just have more data.” AI – and Machine Learning as part of it – needs voluminous datasets to train its algorithms –and big-name cloud companies have access to them.
Add to this continual upgrades, exceptional compute power and regularly implemented new features – and a list of benefits from cloud providers gets complete.
Examples of Combining AI & the Cloud
Image recognitionAs found in the EIU’s study of the AI market, image recognition is one of the most attractive areas for US companies within the investment in AI.
Image recognition comes in many forms:
- Image tagging: analysing images and extracting associated tags and keywords, in order to classify or search afterwards. The algorithms has multiple applications in tourism and retail.
- Face recognition & verification: the technology includes such algorithms as face detection, face tracking, face matching, and face attributes extracting (age, gender, emotions, etc.). It is applied for security and authentication, profiling / segmentation of customers, identification in physical stores and doing market research.
- Facial emotion detection: analyzing a face expression to classify, predict, and forecast the behavior. The algorithm can be used in boosting the shopping experience in physical stores. Smart car manufacturers can leverage the technology by alerting the driver when he is feeling drowsy. Some major companies (for example, Unilever) are integrating the technology into the recruitment process.
- Diagnosis of diseases: analyzing a patient’s face for signs of disease and comparison with previous diagnoses. For example, a US company called FDNA is testing its software DeepGestalt, aimed at diagnosing and predicting the development of rare genetic diseases.
- Google API Cloud Vision: boasting powerful pre-trained face recognition models, empowered with AutoML to allow developers to train these models on more personalized datasets.
- Amazon Rekognition: known for its accurate face recognition service, Amazon Rekognition adds the service to its other products. Add to this the amount of data stored in the world’s biggest cloud – and you will imagine how effective this tool can be.
- IBM Image Detection: like other tech giant’s AI tools, this one is an industry-level application, which tags, classifies and analyses images with a remarkable accuracy.
- Microsoft Face API: the service provides, among others, Face Verification as a service, i.e. it checks 2 different faces to be the same person. Besides, it identifies human’s emotions like anger, happiness or disgust.
Conversation recognition & automationA chatbot enhanced by AI is becoming quite a commonplace tool today. Though can it be called a chatbot? Now it is a human-like intelligent assistant, heavily relying on Natural Language Programming (NLP) and learning from large inputs of data.
The power of AI-enabled conversational assistants is based on these pillars:
- Smooth switchover between text and voice.
- Human level conversation & chatbot adjusting itself to a user, their speed and manner of talking.
- Context understanding & retaining, i.e. assembling different elements of the conversational context (time, location, etc.)
- Intent recognition, understanding the intention and meaning behind every phrase.
As you can see, developing such a product includes several tools:
- Speech-to-Text and Text-to-Speech conversion tools;
- Tools for working with a language (translation system, natural language classifier)
- Emotional state insights and tone analyzers.
Decision making & recommendationsDecisions made by humans, however reasonable they may seem, have one serious disadvantage: humans have biases. AI technologies are expected to make real-time predictive decisions without human subjectivity, anxiety and emotions, by following the next steps:
- Getting deep insights into historical data either from the input data or from perception AI tools (speech recognition, image tagging, etc.);
- Identifying current trends and patterns;
- Making recommendations.
In case of cloud-based AI tools and services, it is not that you are “choosing” the most powerful one. Once your infrastructure moves to the cloud, you are in some way tied to its services and APIs to integrate into your system. So far only IBM Watson has made its AI services available in any cloud, including those of their competitors. However, checking with a cloud consulting services company how efficient AI services are, can help you make the choice if you are still considering moving to the cloud or migrating to another one, once the benefits of this migration outweigh related inconveniences and expenses.