- IBM and HP formalized service and support packages around Unix, providing enterprise organizations the confidence to move that OS out of the sandbox and into wide usage.
- IBM’s integration of AI into a business application shows it is ahead of the curve in the deep learning space, and that lends credence to their claim to focus on enterprise.
IBM today made a major announcement about new products and services aimed at helping alleviate the roadblocks to AI adoption in the enterprise. It is only a beginning, but it’s very interesting in the breadth and comprehensiveness of IBM’s plan.
There are three key areas of concern that IBM addressed:
- Infrastructure with limited I/O
- Data bottlenecks
- AI is not enterprise ready
On the infrastructure front, they announced enhancements to the AC922 Server, primarily be improved integration of NVIDIA V100 GPUs and NVLink for faster system communications. This server is for the heavy lifting, training of AI models and processing in HPC systems.
They also announced the LC921 and LC922 servers aimed at data-intensive applications. I will let others describe the detail of the hardware, what’s interesting to note is the focus on data is aimed at two areas: data stores and AI. While many companies have talked about the importance of information sources for AI, this is a clear message that IBM understands it needs to support efficient access to data stores in order to support responsive AI inference.
The Enterprise Wants More Than “Cool”
The final point is the one most interesting to me. I’m not surprised to see IBM take a leadership position in understanding and addressing productization of machine learning, which is still primarily in the R&D phase. There are very few applications using the technology, and those are usually being created by startups. Enterprise organizations understand that AI can change most things, but they must be careful in their adoption of any new technology.
IBM has always been focused on the enterprise. Without going all the way back, I’ll quickly mention Unix. When I worked with it, in the 1980s and early 90s there were a myriad of flavors based on open source ATT & BSD Unix versions. Many companies toyed with Unix but were hesitant to incorporate its use widely or in mission-critical applications. Then Sun, IBM and HP formalized service and support packages around Unix, providing enterprise organizations the confidence to move that OS out of the sandbox and into wide usage.
One of the key announcements today was IBM stating that Power AI now supports TensorFlow on Red Hat Enterprise Linux. Just as the three companies formalized provision of Unix, Red Hat is the major company to formalize provisioning of Linux, a slimmed version of Unix great for faster services. The organizational infrastructure of IBM and Red Hat will combine to provide enterprise IT organizations the confidence to move forward in deep learning model development and deployment.
I’ve previously commented on frameworks not being enough for AI to move forward. IBM has announced an interesting step to move past frameworks in video and images analytics. IBM Intelligent Video Analytics is an existing product that helps business use video for security, safety, and other functions. Power AI Vision is IBM’s DL application set for doing what the name implies, using DL for vision processing. Today’s announcements included IBM integrating the two applications.
That is important for wider DL adoption because IBM IVA has a GUI interface. It has point-and-click, drag-and-drop, and other features that allow the non-technical, end-user, to optimize the video environment. Integration PowerAI Vision into IBM IVA means allowing the same non-technical people to quickly leverage the power of AI within their existing environment. The users don’t have to understand neural networking, they don’t need to know a thing about GPUs or acceleration. They know they can quickly gain benefit to their own workflow and business.
In the business intelligence world, I’ve regularly said that the technical people need to stop claiming the end user needs to think as they think. The technical people need to learn to think as the users think. BI is more mature than AI, but still has worked around the problem.
IBM’s integration of AI into a business application shows it is ahead of the curve in the deep learning space, and that lends credence to their claim to focus on enterprise.
The author and members of the TIRIAS Research staff do not hold equity positions in any of the companies mentioned. TIRIAS Research tracks and consults for companies throughout the electronics ecosystem from semiconductors to systems and sensors to the cloud.
Date: May 08, 2018