The word AI has become primordial in all industries worldwide. AI is no longer a subject matter of discussion. The more relevant question that organizations have nowadays is about the area of deployment. Every CIO/CXO is thinking “How do we use AI? What are the tools that are out there which are useful? Are there readymade platforms that will help us to hit the market running?” As far as the actual consumer is concerned, he/she is already interacting with AI enabled technology without being aware of it.
However with proliferation of AI terminologies/companies there is a disproportionate increase in the number of references to AI. When we Google AI, there around 70 Million search results. The entire information technology field evolved due to the pressure of automation. But Google trends for the last year, tells us that “Artificial Intelligence” is more popular than “Automation”
This means that as far as our mindshare is concerned, AI has taken over research (industrial and academic), development, usage and applicability.
[Interesting AI statistics from Forbes]( https://www.forbes.com/sites/louiscolumbus/2018/01/12/10-charts-that-will-change-your-perspective-on-artificial-intelligences-growth/#73421d434758)
So is AI natural language processing? Or is it machine learning? What exactly does it constitute? In order to understand this we need to comprehend the evolution of artificial intelligence. AI was always around ever since the dawn of the computer age. Microsoft Word was one of the first tools to incorporate a very primitive form of AI by supporting grammatical reviews. Video games (yes even the Atari ones) learnt from our mistakes to make the game harder because of inbuilt rules. So AI is not a new term but it is important to understand how it came about.
Correspondingly the AI architecture was built up gradually
1. AI is nothing without data and it is not surprising that we need a suitably large cluster of transformed data (ETL’d) to achieve any kind success. Big data clusters are pretty widely used
2. Knowledge management is the next crucial component. Relationships need to be represented and access needs to be fast as well as decisive. Knowledge graphs are gaining popularity
3. Once we have the knowledge then Machine Learning (ML) steps in thus permitting AI to start improving beyond the initial data set. This learning can be supervised by humans, totally unsupervised, driven by rule sets or learn by mining vast data sets (structured or otherwise). TensorFlow is a good example here
4. An ML infrastructure always needs input to make it grow and the only way AI could do it was to start recognizing bits and pieces of whatever data was input to it. The ability to find patterns and report them to ML algorithms is exemplified by Natural language processing, Image/Video recognition, Speech Recognition, etc. AWS Rekognition does exactly that
5. The next level was to actually have an interactive conversation with a data source (humans too!) and probe/pester them like a good investigator resulting in a rich set of interpretive information. Chat bots are a primary example
6. Eventually all that learning has to result in the ability to make decisions which is offered by applications that run on top. E.g. a virtual assistant
7. Finally applications will need to start solving problems that could be ongoing, buried in the past or about to happen. Cognitive apps are the right hand of the AI gods. Cognitive apps are knowledge base specific and work on focused problems. E.g. Dementia training
Naturally the Artificial Intelligence landscape in the industry has evolved incrementally either addressing portions of the required infrastructure or all of it. AI infrastructure is divided along usage fault lines and some of it should be fairly familiar to most of us. Each bit of the infrastructure will address specific problems of putting a custom AI solution together. The figure below categorizes AI infrastructure which will be used to showcase available AI software.
## Will AI ensure that jobs are lost?
<br/>
I think that this is a no brainer. AI is automating everything that can possibly be done. In fact the sweet spot for AI is to automate tasks that require low level human interaction. Imagine that a tier-1 customer support executive is replaced by a standard AI program that holds a conversation until it is forced to transfer to tier-2. Another good example is a report reader who is supposed to publish basic insights or highlights from a report. This can easily be replaced by AI.
Until now automation was always limited to jobs that were not done by humans. With AI, there is an increased focus on operational efficiency which is achieved by replacing productive humans. In addition, AI could also induce reduction in compensation for workers since their portfolios will either be reduced or obliterated
# AI services
The following section will start itemizing some exemplary services in each category. The list is exhaustive and hence we cover only a specific subset which we think are representative and illustrative of AI infrastructure available today. AI in healthcare needs an independent look since it is highly specialized and is not presented in its entirety here.
## Consumer AI services
Consumers can directly use the service and typically solve a specific problem for the average consumer
* **Wazzat Fashion:** Allows users to quickly find apparels by using photographs or links. The app personalizes the experience and suggests cross matched items improving shopping experience
* **Virtual Assistants:** Learns user preferences and personalizes all interaction. Examples are Siri, Google Assistant, Alexa etc. These can be trained with supervised inputs as well
* Desktop Assistants: These are software that specifically allows you to use the desktop more productively either for business or personal use. Microsoft Cortana should rule big with enterprise users
* **Niki.ai:** helps you to eliminate apps and make online purchases
* **X.ai:** schedules meetings by integrating with well known communication apps like Gmail, outlook, slack, etc.
## AI building blocks
Building blocks are technologies around which any company can build their own solution.
*source: clairvoyant*
* **Image/Video Recognition API:** These APIs can be integrated into your app to instantly give it superior AI functionality. Examples are AWS Vision services, Google cloud vision, etc
* **Facial recognition:** Read emotions, expressions, matches biometric information and much more. Try TrueFace.ai or Ever.ai along with usual suspects like Google, AWS and Azure’s Face API
* **Natural language processors:** NLP primitives that allow you to introduce a fair amount of real world conversation capability into chats, emails, reports, etc. Service apis are available from Wit.ai, Dialogflow, Google assistant gRPC, AWS, IBM Watson conversation service, Microsoft…
* **Message Bots:** These API enables bot creation for mobile devices, web apps, and chat services such as Face book Messenger, SnatchApp, and Skype, using Slack and Twilio integration to send and receive messages. There are too many services in this space
* **Machine learning as a service (MLaaS):** Cloud API provides modern machine learning services, with pre-trained models and a service to generate your own tailored models. E.g. Google prediction API, AWS ML api, Microsoft ML, BigML, etc.
* **Behavior based security:** APIs that allow your app to dispense with password based security and instead lean on cutting edge personality/behavior based analysis. Aetna is offering this for its healthcare network.
## Tools, Utilities and Libraries
These are completely self contained software services that can be run on premise or within the cloud. All they need are access to your data to perform specific functions.
*source: techugo*
* **Cyber security:** Allows existing security vulnerabilities to be learnt, reported and constantly monitored. Companies like Paladion, Darktrace offer very comprehensively featured services. Fortinet integrates AI into their routers to underpin your cyber security infrastructure
* **Interpreters:** These are scanners that assist a business in completing a specific process. E.g.: ZocDoc has an insurance checker that can be run by doctor to ensure that the patience is covered adequately for the treatment that he/she is about to prescribe. Doc.ai will interpret test results and advise the doctor. Emotional intelligence checkers will try to extract more from your communications
* **Process Enablers:** These are independent tools that can be run on a data source just like BI apps and gives constantly learning insights into existing as well as real time processes. People.ai scratches around your sales communications to create a successful sales model for a specific sale. Legal Robot uses NLP and deep learning to create high-level legal models from contracts and acts as a legal advisor
* **Surveillance:** Tools that could monitor your video feeds, pictures or other digital media vulnerabilities. E.g. Kipod offers video/alarm monitoring services in-step with your feeds in your private cloud
* **Numerical computation:** Tensor Flow is a high performance library that can be deployed on most cloud architectures to achieve ML
## Platforms
These are essentially hosted services that provide a platform for building AI enabled apps, allow users to create custom AI algorithms, ingest existing data and offer templates that can be instantly used
* **MonkeyLearn:** Allows analysis of different textual information, creates insights that can be acted upon, and also integrates with other enterprise applications. It can closely work with with tweets, emails, chats, web pages, reviews, documents, etc. One can do sentiment analysis, correct frequent errors, automatically train users, etc
* **IBM Watson:** is an AI platform for developing any kind of intelligent business tool. It is highly modular and can ETL any data, search for insights and carry out actions
* **Microsoft Azure Machine Learning Studio:** Complete MLaaS platform that allows custom app integration and can also function as a standalone service
* **Interact:** is an AI based Tech Hiring Platform that allows enterprises to conduct assessments and Interviews accelerating the hiring process
* **Amazon SageMaker:** With a highly scalable modular architecture, AWS enables custom built AI applications coupled with training and inferences.
* **Playment:** is an AI and human intelligence platform for web-based tasks. It provides human based tagging for powering deep learning
* **Ayasdi:** serves up an SDK which gives access to their algorithms for machine learning, visualization and data science applications which can power up your enterprise app with AI functionality. You have to develop your app on top of Ayasdi
* **Infosys Mena:** is a business process automation enabler that uses ML and deep learning to provide a responsive user experience
* **Decision support systems:** These are a natural extension of AI platforms, helping in decision making for a variety of situations. Typical examples are Rainbird.ai, Absolutdata, etc
## B2B AI services
These are services or products that allow business to offer AI enabled services to their customers
*source: Aberdeen Essentials*
* **Vphrase** converts business insights into words. Insights can be financial, business, HR, etc
* **Staqu** solves cross product discovery in ecommerce using deep learning of user experiences and contextual decisions
* **Skedool** helps in scheduling appointments
* **Fluid.ai** is doing consumer smart assistant in hotels, shops …
* **Nuance** has created a virtual assistant known as Nina only for enriching customer service
* **Sigtuple** helps doctors to analyze reports better
* **Morph.ai** does conversational services over Face Book, Whatsapp, email …
* **Artifacia** offers AI based product tagging for photos
* **Infinite Analytics** develops insights on customer interactions, finds customers, targets based n their intent and continuously optimizes user engagement.
* **Boulder.ai** enables AI on any device like cameras, telephones, heat sensors, etc
## Products
These are fully AI enabled products for everyday business operations
* SalesForce offers fully integrated AI services for any product offering like CRM, sales tools and other apps using their Einstein service
* albert.ai is touted as the first hands-off digital marketer which is fully self learning
* Domo is a business management software company that integrates AI functionality and integrates with data lakes from different sources
* Siemens has MindSphere which monitors industrial equipment performance with deep insights
## Research
These are services that provide infrastructure or data for research purposes.
* **Plasticity** service provides a way to access 180+ million facts on over 20+ million entities using the Cortex Knowledge Graph. It allows you to query the graph in natural language or make your own graph pipelines
* **Plume API** is a machine learning and atmospheric science that provides air quality data and hourly forecasts. This AI powered platform is used for live access to forecast pollution data
## Development
These are platforms, tools and services that help development
* **Apache PredictionIO** is an open source machine learning server built on top of an open source stack for developers and data scientists to create predictive engines for any machine learning task
* **Eclipse Deeplearning4**j is an open-source deep-learning library for the Java Virtual Machine. It can serve as a DIY tool for Java, Scala and Clojure programmers working on Hadoop and other file systems
* **Nervana and Intel** have released Neon which is a generation of intelligent agents and applications and is an open source Python-based machine learning library
* **Apache Mahout** is for developers wanting to create scalable machine learning applications. It allows users to use its pre-formed algorithms for Apache Spark, H20 and Apache Flink
AI is here to stay and judging by the amount of investments in this space, it is evident that it will become a permanent player in all industries, products and services. Even a small company today can develop an AI enabled application with the right set of data scientists. In fact it is quite possible that no company can afford to ignore AI integration. It reduces costs, cuts operational errors and improves efficiency.
hiveblocks