It is imperative that GCP evolves even further with the addition of new features. Here is a representative list of new additions that we think will be added to GCP in the future: 

• GCP will have greater integration and availability of cross-platform products. For example, IBM Power Systems is now available on GCP. That way, the investments that have already been made by enterprises in large production systems will be utilized by migrating entire platforms onto GCP. This will result in implementation and infrastructure cost savings for enterprises.

• GCP will be enabled with ready-to-use AI and ML models. As the marketplace matures, GCP will host additional and ever-increasing numbers of AI and ML models. These models will be available to use via predefined APIs with inherent interoperability. The models will be constantly trained and tuned by GCP and produce increasingly better results with time. The marketplace will mature with increased usage. The signing up and pricing will be simplified to the extent that developers at all levels of experience (including entry level), will be able to quickly build their enterprise applications. 
• GCP will provide a drag-and-drop user interface for building entire AI pipelines, from problem classification to model deployment. At that point, the power of AI will be fully in the hands of business teams and there will be less dependence on IT and development teams. The simplification and democratization of the platform will result in even further innovation and we will experience intelligent applications that are not only used but that are built by everyone. 
• GCP will enable industry- and business-specific AI toolkits for improved profitability and innovation for enterprises of all sizes. For example, Google is already helping retailers to accelerate their digital and multichannel revenue growth. Along with this, Google is also helping retailers to become fully-data driven and make suggestions (based on data) for improvements in operational efficiency. This is possible by leveraging the AI tools on GCP, Chrome Enterprise, and Android, as well as the entire connected toolkit. 
• The AI toolkit on GCP will also facilitate research projects that require high volumes of data and computation power, as well as a process and interface for building the AI pipeline. For example, Google is helping FDA MyStudies in leveraging real-world data for biological research. Google Cloud is working with FDA on the MyStudies application with better and adaptable protection and configurable privacy policies. The goal is to provide research organizations with the ability to automatically identify and secure information that is personally identifiable. Google Cloud will continue to invest in various studies and research programs to bring general improvements to the platform, expand the number of assessments supported, and allow integration with downstream analytics and visualization tools.

• AutoML Tables enables your entire data scientist team to automatically build and deploy ML models on structured data at a highly increased speed and scale. It comes with great feature engineering model training features. When training starts, AutoML tables automatically perform some feature engineering tasks, such as normalizing the inputs. The numeric features are confined to the ranges for better model reliability, normalization of date–time input parameters, basic text processing cleanup, and stop word removal, and create one-hot encoding and embedding for the dependent variables. The AutoML Tables perform parallel testing on linear, feed-forward deep neural networks, gradient-boosted decision trees, AdaNet, and ensembles of various model architectures to determine the best model architecture for your dataset. Users will be able to view the AutoML Table structure using StackDriver Logging and will be able to export test data as well.
• AI Hub is another very useful feature that is coming to GCP. AI Hub is a one-stop facility for building even the most complex ML pipelines. It is possible to use preconfigured notebooks to build AI applications using some of the pretrained models, as well as to easily train new models. AI Hub also ensures access to relevent datasets in a consistent manner. It is also possible to collaborate on model development and leverage models built on common framework, such as TensorFlow. This greatly facilitates the training and deployment of models on GCP.
• AI Platform Notebooks will make it easy to manage JupyterLab instances through a protected, publicly available notebook instance URL. It will enable you to create and manage a virtual machine instance that is prepackaged with JupyterLab. AI Platform Notebooks instances will support PyTorch and TensorFlow frameworks. These notebooks will be protected by GCP authentication and authorization. AI Platform Notebooks will come with a lot of preinstalled commonly used software.
• AI Platform deep learning containers are a unique way in which GCP provides access to an array of pretrained models that can be quickly prototyped and used with the help of highly optimized and consistent environments on GCP. This helps in building workflows quickly and facilitates experiments with a minimal entry barrier and cost. This is a great leap toward fully democratizing AI development capabilities.

• The AI Platform data labeling service is a great way to leverage human intelligence in labeling data points at internet scale. An organization can request this service from Google to label datasets manually. This is helpful in gathering the training and evaluation data when a new use case is considered and the initial dataset is not available. There is a consistent effort from Google to crowd source the process of dataset labeling on the internet. The labeling service is also handy when we want to deal with highly secure data that needs to be labeled. The interface with the labeling service is a secure and efficient way of getting data labeled.