How AI OpenScale Overcomes AI’s Biggest Challenges
Q&A with David Kenny, Senior Vice President, IBM Cognitive Solutions
To better understand AI OpenScale, and how it helps businesses accelerate their AI strategies, we talked to David Kenny, Senior Vice President of Cognitive Solutions at IBM.
IBM is taking another big step forward in open technology with the release of AI OpenScale. How does this open approach help IBM’s customers?
For us, open means enabling our clients to move much more quickly in adopting the latest technology, wherever it comes from. We are not just committed to open source —we have been an advocate of open since the early days of Linux —we are also committed to enabling businesses to use and connect open source code and open frameworks. That's important because our clients are need to scale.
In AI, we're seeing our customers using a variety of AI and machine-learning frameworks. Our OpenScale platform works with Watson but also things like TensorFlow, Spark ML, and AWS Sagemaker. We worked hard to make this a very open, modular, service-based platform.
Many companies have used AI in pilots. We are now at the point where in order to move to scale, you need to have an orchestration layer that is safe, secure, and transparent, and also enables it to scale.
How will AI OpenScale help companies more widely deploy AI?
In the study IBM conducted, we asked why AI adoption has been limited despite the high level of interest, and it came down to two things. One, AI is still too much of a black box. If you were going to use it to make decisions that matter, whether lending, health, or safety decisions, you need to know how the decision was reached in those models. Many want to ensure that AI models are fair and unbiased before they deploy them at scale. Transparency is key.
The second thing they said is that their companies don’t have enough AI experts to do things at scale. AI OpenScale helps solve the talent shortage as well.
So, we're making big progress in solving the two main barriers to adoption.
All of this is part of our vision in enabling the tipping point for AI in business. That’s when organizations begin to embed AI and machine learning deeply into the way they make decisions.
You mentioned businesses using different AI models and frameworks. How does AI OpenScale address that?
Organizations are largely multi-cloud. They have data in different clouds and use various tools. They also find that different models are better tuned to particular use cases. For example, Watson tends to be very good with private data and things that are business-to-business oriented. Some others are more tuned to the public Internet and public data. And some of it has to do with the way organizations work. There are pilot programs within organizations, and people don't want to have to refactor all of those. So being able to integrate them is something that we do well.
How can AI OpenScale help businesses beyond orchestration?
Organizations are very concerned that when AI is being done in production at scale, it needs to support their policies. They have policies around fairness and lack of bias; many have policies around traceability to know where the data came from; and many industries are regulated. It's one thing when you're doing pilots, but when you go to production, and you're going to be audited and face regulators, you need to be compliant with policies. So having a tool that you can port these models over to address that means that you can take them out of the experiment phase and scale it up. AI OpenScale manages AI, making it explainable and addressing bias across the AI lifecycle.
How does AI OpenScale automate AI and address the skills gap?
This is a really exciting part of AI OpenScale. We call it neural network synthesis, or NeuNetS. This is a huge breakthrough. NeuNetS can create more complex deep neural networks automatically. This means that more companies will be able to adopt AI faster because they will have AI helping them use AI, instead of relying entirely on in-house skills.
What is the story behind AI OpenScale’s development?
The idea started when we did our annual Global Technology Outlook. We saw from our data that people really believe in AI and its ability to make them more productive and make better decisions. We asked ourselves: What is the future of AI? How can we make AI scale? And that’s how the idea of AI OpenScale came to be. I'm very proud of this product. It's awesome.