IBM Is Making Artificial Intelligence Easier and Renaming Solutions for AI

April 26th, 2019 IBM Is Making Artificial Intelligence Easier and Renaming Solutions for AI

Ron Gordon
Director » Power Systems

I was recently at a great AI workshop put on by IBM. IBM went over their AI strategy, directions and futures, as well as gave us hands-on labs to run the key products in the IBM AI portfolio.

AI Workshop Hands-On Labs

In the workshop hands-on labs, we ran modeling using:

  1. Watson Machine Learning Accelerated (WMLA)
  2. Power AI Vision
  3. Driverless H20

While the labs were predefined with the data elements already created in databases, the actual use of the tools was exciting and enlightening, as to the ease of use via the EUIs of the tools. In modeling and inferencing, we ingested the data into the AI modeling tool and then remodeled, while changing the model. After we achieved a high inference index, we passed new data against the optimized model. Very easy and very intuitive. Each AI product (WMLA, AI Vision, Driverless H2O) used unique use cases, demonstrating the value of different AI tools. Not that I am a data scientist now, but with the experience gained, I have a deeper appreciation for AI enablement. And, while one size does not fit all, please don’t be afraid of AI. Another very positive impression was the high level of expertise that IBM has in AI solutions and implementations, as well as the assistance they can provide, based on their broad understanding of all aspects of AI.

Renaming of the IBM AI Portfolio

One element of the IBM focus on AI is the renaming of their AI portfolio and aligning the tooling within the Watson brand. But oddly, when I received my latest copy of IBM Systems magazine, I noticed none of the new branding was included nor discussed, and references were made to the old solutions and naming. So you can properly interact with IBM and its business partners on AI, I thought I’d would also take some time to explain the new branding and support.

  • The IBM Strategy:

    Through the new simplified offerings, the objective is to make the total solution components easier to understand and easier to deliver for customer consumption. Through these offering, and additional add-ons, the solutions will enable the preparation of data, the building of AI models (both Machine Learning and Deep Learning – ML and DL based), train models easier, deploy and manage the models, leverage the CPU/GPU and FPGAs of Power Systems architecture for model training and inferencing, scale models, provide data scientists development efficiencies, and enable solutions on both on-premise and via cloud solutions.
  • The Rebranding:

    Power AI is now named Watson Machine Learning Community Edition, or WML. WML is the delivery of the key AI frameworks such as Caffe, PyTorch, Torch, TensorFlow, Chainer, etc. It is a no-charge product for Power System, for rapid installation of these frameworks, which are optimized for the POWER architecture. Power AI Enterprise is now named Watson Machine Learning Accelerated or WMLA. WMLA is the build engine for AI models. It also contains a distributed model for clusters (DDL) and large-model optimization support of AI tuneables (LMS), as well as multi-tenant support. As Power AI Vision is part of the vision brand, Power AI Vision retains its name and provides image modeling and cognition. The trite question of is that a dog or a cat is an example often used. There is also the Watson Open Scale product, which has been added to the solution to provide business KPI, along with explainability and fairness. (AI is creating answers that in many instances need to be explained, as to how the conclusion was reached, to insure non-bias, as well as how the models support requirements like GDPR. This is the objective of Watson Open Scale.)

IBM Open Source Based AI Stack

This diagram shows the current key solutions:

Other IBM portfolio elements that contribute to the AI solution, such as System Conductor and IBM Cloud Private for Data (ICP4D), Watson Studio Local for a shared model development environment utilizing Jypter Notebooks, etc., are still part of the overall AI tooling but could be considered optional from a PoC standpoint. Other solutions for model building and inferencing are also a different solution stack, such as Power AI Vision for object identification and H2O for structured data analysis.

For more information on the new IBM AI solution structuring and AI in general, please contact your Mainline Account Executive directly, or click here to contact us with any questions..

Related Information:

Watch IBM Power and AI – Your Journey Starts Now

Mainline