| |June 201919make predictions with the help of labeled data or data for which the desired outcome is already known. It is an increasingly used ML tool in risk assessment for financial services, fraud detection, and visual recognition/identifying places or people in images. In 2019, auto ML will improve such that any supervised learning task, one where a label is involved, will be able to have an algorithmic selection and hyperparameter optimization confidently automated by a computer. Though supervised learning is currently the dominant approach in AI, it often requires the labor-intensive and time-consuming process of getting humans to annotate data manually. There are also limits on what it can do. Businesses need a better way to bridge the gap between representational learning and causal reasoning. Accordingly, in 2019, research will focus increasingly on unsupervised learning (the machine is trained on unlabeled data, and the algorithm acts on information minus supervision) and reinforcement learning (agent/algorithm learns by interacting with its environment). There will be more popularization of RL techniques, and a larger number of RL libraries will be incorporated into an enterprise environment in 2019. Explainable AI to have More Takers in 2019 Machine Learning models are increasingly being used in uncharted areas and complex/critical AI applications such as healthcare, defense, transportation, and security, among others, which can offer tremendous benefits. However, AI/ML predictions are typically not accompanied by a justification for that specific prediction or any insight on the model that was learned, which could leave users frustrated, particularly if they are responsible for critical applications. The industry has made a lot of progress in AI usage without ascertaining how to explain it, and most often, no one has a clue as to how accurate the prediction is or how it was arrived at. This has created a demand for AI systems ­ called explainable AI or ex-AI ­ which can be used by humans to understand the decisions made by the machine learning model. It is an explainable model that gives an understanding as to why a specific prediction was generated. Accordingly, in 2019, enterprises/research organizations will focus on creating a new generation of AI systems that are ML enabled, in which users will understand why the predictions have been generated, ascertain when they can trust the model and work with it effectively to manage emerging AI systems or critical applications. The focus will be on creating a suite of ML techniques that produce trustworthy, reliable, and explainable models, without sacrificing the high level of learning performance. Explainable AI is critical for the acceptance of ML models into mainstream use cases. For organizations and practitioners, it is imperative to start focusing on accountable and transparent AI, which will make it more explainable. This will become particularly necessary this year for regulated sectors making critical decisions, who will have no choice but to ensure that the model is explainable. AI and ML at the EdgeThe importance of taking local and faster decisions has become more critical than ever before, and AI and ML are moving to the edge, facilitating intelligence and smart decisions where it is needed. The amount of data is growing exponentially, which requires analyzing and transforming them into actionable insights at an unprecedented scale, without compromising on efficiency or performance. This is necessitating a move from connectivity to intelligence and local/real-time processing. The reason: In the absence of useful insights from the data generated by multiple connected devices, there is no return on investment. Hence, to enable this evolution at the edge and tackle big data challenges, more number of organizations this year will migrate to edge computing this year, which will give them speed and latency, locality of computing, security and privacy, scalability, and enable a business to push intelligence to the edge of the network. Edge computing will allow millions of devices to process data locally along with AI and ML algorithms, facilitating smart, real-time decision-making. Companies like Amazon, Google, and Microsoft, have already introduced technologies/platforms that can help businesses implement AI/ML capabilities to the edge of the network cost-effectively, respond quickly to local demands/challenges, and make real-time predictions. AI and ML are moving to the edge, facilitating intelligence and smart decisions where it is needed
< Page 9 | Page 11 >