AI into 2019 - A PoV

As we move forward to 2019, Data Science and AI is getting smarter and smarter. It's always challenging to predict the "predictive analytics", "artificial intelligence" etc. considering the magnitude of complexities involved, innovations that are occurring every day, dynamism that play role in solving each and every business problems and use cases, solution approaches those evolve as we try to solve them.

Obviously, all of us know that the intent of AI and it's usage will be to make our lives easier, simpler (if we may say so), better and enable us to "do more with less". Users at ground level will prefer to see which task to be delegated to machines so that they can learn and do the work better for humans and which one they would like to do on their own. That's where lot of tricky decision making will be used and human brains will continue to dominate those.

I thought of putting together a point of view looking at twelve key points (by doing a quick math). It's always tough to jot down how many focus areas that we can cover in a thought process, hence I picked the year number which is 2019 and summed up all digits to make it (2+0+1+9=12) twelve as a number and as many points to describe about.

Below are key thoughts which could be relevant and be mainstream during next year:

1. Model usage and approaches - Appropriate permutation and combination of algorithms / approaches and it's use are going to be more helpful as ensemble based machine learning approaches will power into solving business problems, they will lead based on performance and evaluation metrics.
2. Model management - Since we are using multiple experiments to arrive at a relatively better solution for a given data, given problem at hand, it is obvious that multiple models will exist and managing them will be critical. Hence more refinement and automation will continue to be performed in this regard.
3. Augmented AI - AutoML (Automated Machine Learning) has been in play at many places and being used significantly. Various Data Science and AI platforms have made it easier for us to be able to use different processes in an automated fashion. However, there will be tasks that involve data preparation efforts, finding relevant patterns in the data, defining and selecting features etc which are still going to be challenging and complex depending on problem, variability of data and use case at hand. This is a key input to the model development process. Hence even if model developments and multiple model experiment processes can be automated, there will be a need for human intervention for decision making at every CRISP-DM step. This is where augmented AI will play a role where AutoML will continue to augment humans in powering better decision making.
4. AI-driven application development - With the help of machine learning and deep learning techniques, more complex business rules can be managed effectively to enhance the ability of a typical SDLC process or application development process. Lot of aspect of maintainability can be accomplished with the help of AI. Collecting data repositories, cleaning them, manipulating their data based on business processes/rules/needs, applying labels, analysing them and exploring them in a visualization will be fed to deep learning models for accomplishing more.
5. Digital Trust and Ethics - Focus is getting shifted to data ethics, data trust and safety aspects, data privacy etc. We can't have decent AI solution created and applied to a problem unless the underlying data is safe, has qualitative inputs. Efforts will be around ensuring data quality, data integrity to create solutions and outcomes which can be realized considering these constraints in mind.
6. Open culture and no locking - Enterprises articulate openly to solution approaches. Machine learning models and approaches will not be way different and there will be continued emphasis to make processes generic, open and accessible to all to be able to leverage and use them in specific industries to solve problems. 
7. Optimization - will always remain a key problem and we will continue to solve problems around it leveraging Data Science and AI. This will not be different in 2019 and will continue as ever.
8. Intelligent systems and environments - Systems have been getting connected thanks to different data sources, their availability, openness, usage. Environments and platforms are getting coordinated providing automated platforms to be able to perform multiple tasks in a seamless manner. Intelligent systems and environments will take both of these and benefit from it.
9. Conversational Chatbots - The role of chatbots will increase with the help of AI. Closer interlocks between other functionalities such as predictive analytics, robotics, blockchain, augmented reality etc will be leveraged in creating a combined view that will have more powerful impact the way bots are doing it today.
10. Explainability aspect - This is no brainer that if we are trying to depend heavily on AI and it's usage, we need to know how it is behaving, how will it perform, some amount of predictability and many have started doing it already. This probably will nurture more into going next year and that depends heavily on a solid foundation of the methodology and framework being used to define, create, test, deploy and maintain an AI solution.
11. Advancement in Natural Language Processing (NLP) - NLP integrates between languages and machine learning to help machines understand "human" part of the language. NLP will continue to evolve to answer analytical conversations (similar to the way a human is having conversation with the machine about the data in a way as if it is a human itself).
12. Usage of predictive and prescriptive power - Products and tools that are built using predictive and prescriptive analytics will provide answers to many questions such as what will happen, how can we make it work, how much time will it take, how can we make it happen, what are the recommended ways to make it happen and the list goes on.

Actionable data will drive AI into informed decision making contextually and help build smarter solutions. I wish all readers a very happy new year 2019!! Happy holidays!!

Disclaimer: "The postings on this site are my own from my experiences, thoughts, readings from various sources and don't necessarily represent any firm's positions, strategies or opinions.”