The article discusses:

• Challenges in AI implementations
• The emergence of new data analytics platforms for optimized AI implementations
• Plethora of available tools: Open Source platforms and commercial platforms
• The need to leverage these platforms to harness the power of AI

Ìý

Over the last decade Artificial Intelligence (AI) has been the most talked about technology and at the same time, most difficult to implement, given the broad spectrum of applications it embraces. To name a few, techniques like speech recognition, image processing, sentiment analysis, natural language generation, process automation etc. form an integral part of AI. Most of these techniques internally train various deep learning (DL) and machine learning (ML) algorithms to optimize their performance. Moreover, these can be consumed individually as an independent AI performing a specialized operation or can be sequentially connected to deliver a multifunctional AI activity.

In either scenario, implementing AI poses challenges of ingesting numerous sources of data, processing terabytes of information to gather signals, training sophisticated models, deploying and managing the models and finally, generating insights for business recommendations. Not only are these processes repetitive but also tedious and difficult, usually slowing down AI implementations. However, to the relief of organizations, the advent of various data analytics platforms and tools has simplified many of these processes so that organizations can focus on consuming the final solution rather than investing time to build a solution from scratch.

Data analytics: Open Source innovators

Open source applications have been the front-runners in transforming AI. Platforms like KNIME, H2O and Tensor Flow have evolved over time for building end-to-end AI/ML solutions. Open source libraries like Keras, Theano and Torch can be used as API with different platforms and frameworks to solve complex DL and ML problems. Tensorflow, a Google brainchild, is now widely adopted by companies like Dropbox, Uber, Twitter, eBay etc. The H2O platform has automated some of the most difficult ML workflows using AutoML functionality that automatically runs through all the algorithms and tunes their hyper parameters to produce a leader board of the best models. KNIME has gone one step further and created H2O extension so that power of both platforms can be leveraged to solve an AI problem.

Data analytics: The commercial contenders

With the open source world setting up the stage it was only a matter of time for commercially adopted platforms to begin their journey. RapidMiner is one of the leaders in commercial space. It is popular for its simple but effective GUI to build AI/ML models. Besides providing numerous in-built ML algorithms, it supports open-source integration. SAS with the execution capabilities of its new suite of products, called SAS AI,Ìý remains a leader in the market. SAS is also expanding its user base with the help of its cloud based AI platform SAS Viya. TIBCO has acquired few new age business intelligence (BI) and advanced analytics vendors (like Jaspersoft, Spotfire, Statistica, Alpine Data etc.) to build an integrated powerful analytics platform.

Alteryx, Dataiku and Datarobot are some of the other AI platforms. Alteryx provides end-to-end data science solution capabilities and is known for its intuitive workflows and drag-and-drop interface which eases the life of data analysts. Dataiku enables self-service analytics and machine learning pipelines through Data Science Studio (DSS), which is a collaborative platform for data scientists, data engineers and data analysts. DataRobot has automated ML and Time Series application along with enterprise AI deployment.

The cloud giants Google, Microsoft and Amazon have already developed their own sophisticated AI/ML platforms for existing user base. They offer pre-trained AI services that can be integrated to any application to address common AI use cases. Most of the open source and commercial AI/ML platforms are either been incorporated by the cloud providers or are available in their marketplace for easy integration to take advantage of the scalability, high availability and security of the cloud, thus streamlining deployment and management of AI pipelines.

To conclude

According to the market research firm Tractica, the global AI software market is expected to grow more than 10 times to $120 Billion by 2025¹. With investments flowing in from all directions, new platforms are launched almost every month to further democratize AI. Interesting fact to observe is that the market is more oligopolistic with the big cloud providers in the forefront and all other open source and commercial platforms fueling in their core strengths into it. Furthermore, all AI/ML platforms have built-in connectors and extensions to integrate with other platforms for seamless flow of information, giving users opportunity to harness the power of AI and ML.

With such a variety of product mix in the market, the onus is now on the users to proficiently exploit them for faster decision-making processes.

Reference

1-