Projections for AI in 2019

Projections for AI in 2019

AI is out there ready to be consumed by startups and corporations alike to solve almost any problem from commuting to visualizing, replacing many mundane human tasks with efficient machines and leaving us humans to make more complex decisions.

When Turing proposed the concept of the thinking machine, this ability of a machine to think for itself was too farfetched and crazy. As a result, the project titled 'Artificial Intelligence' (AI) kept getting shelved. But if we were to learn from history machines would also become smarter than humans once they get the drift. So, we should ask ourselves, 'How close will we be to that stage in 2019?' Only that can summarize any projections for 2019 because 'projections' are towards an inevitable future, otherwise they're merely wishful thoughts or prophesies.

AI could impact every aspect of our lives but due to the limitations of space and time I will restrict myself to AI in text processing which we've been working on for the last five years.

1. Data Integration: Natural Language Processing (NLP) has played an important role in analyzing textual information and this continues to be the case. Up until late 80s there wasn't much digital data which meant machines didn't have enough training data or it had to be entered manually. In fact, IBM's Watson had engineers feeding data like 'water makes you wet', 'milk is nourishing', etc. In 2019, the availability of data is pretty decent and structured. Almost every organization has its data sources; CRM, HRMS, ERP, LMS, etc. Coming to unstructured data, which has a lot of intelligence still untapped and growing at a phenomenal rate, thanks to e-mail and collaborative platforms like MS Teams, Slack, Chat assistance, WhatsApp for business, Skype for business, etc. There is now a huge possibility of combining these databases to derive meaning from them, automate conversations, processes, and bring enterprise conversations a lot closer to human parity.

2. Omnipresence of Chatbots and Virtual Assistants: By 2019, at least 25 percent of employees at all large corporations will communicate with a bot for information. More than half of organizations have invested in VCAs for customer service, as they realize the advantages of automated self-service and the ability to escalate to a human in complex situations. Across industry verticals, business functions that are seeing most demand with customers span across sales, marketing and HR.

60 percent of all hires would have been either screened or shortlisted or interviewed by some NLP AI engine. Any bot can be deployed on any integrated channel in a few clicks, so there is only one bot overall, saving time and effort. An AI assistant semantically understands job descriptions you feed in and finds relevant matches for the requirement from available job portals and databases. A hiring assistant can also reach out to identified candidates and engage in a chat to pre-qualify them as per company requirements.

3. Reinforcement learning: When I discussed structured and unstructured data, the AI learns from categorised and uncategorised information and forms an output. But what happens when the AI must make an unbiased decision? This is where reinforcement learning comes into place. The framework does not use data recognition as above but takes into consideration its previous experience and outcomes that resulted into rewards. Reinforcement learning is mostly used in computer games. The actions taken by the computer and the player to decide the winner of the game. This type of learning is still unchartered territory and can be useful in several ways like determining treatment methods for chronic illnesses like Alzheimer's or schizophrenia. It can also help in higher education or career choices.

4. AI & DevOps: Modern day applications constantly collect information on how a user interacts with an application, as well as on how the application is being delivered. There is a large amount of data that can be used for indexing and analytical purposes. Add to this machine learning and this data can be processed at a remarkable pace. By integrating machine learning into the delivery system of the application, organizations would be able to generate insights into all bottlenecks and relevant patterns that make the user experience and app delivery seamless and avoid similar blockages in the future.

AI is out there ready to be consumed by startups and corporations alike to solve almost any problem from commuting to visualizing, replacing many mundane human tasks with efficient machines and leaving us humans to make more complex decisions. According to O'Reilly data, 51 percent of surveyed organizations already use data science teams to develop AI solutions for internal purposes. There is no doubt that adoption of AI tools would be one the most important AI trends in 2019. Exciting times ahead, indeed!

(The author is Co-Founder & Chief Evangelist, Light Information Systems)