Some Ideas For Watson’s Roadmap

Yesterday, I wrote about the challenges that the IBM Watson platform is experiencing in the market. As explained then, part of the challenges Watson has been dealing with has been self-inflicted by IBM’s disproportional marketing efforts while others are due to fundamental limitations in the technology. I am afraid I can’t offer any relevant advice on the marketing front( except maybe one: tone it down!) but there are a few ideas I feel are worth considering for Watson’s short-term roadmap. Here is my top list:

1 — Support for Deep Learning Frameworks

Watson Developer Cloud(WDC) is, essentially, a series of APIs that abstract specific models in mainstream areas such as image recognition, video analytics, natural language processing and others. while APIs makes it relatively simple for developer to leverage AI models in client applications, it is completely impractical when comes to implementing and deploying new algorithms.

To address this challenge, Watson could expand its runtime to support programs written in deep learning frameworks such as Caffe, TensorFlow, Torch, Theano and others. IBM is already pretty involved with Caffe so this recommendation might not be a long stretch.

2 — New AI Research Models

complementing the previous point, Watson should find a vehicle to incorporate new AI algorithms produced by universities and research centers into the platform. Linking the AI research community to the Watson platform could give IBM a viable alternative to competitive platforms such as Google’s Kaggle.

3 — On-Premise Edition

Expanding support for on-premise and container infrastructures would certainly accelerate the adoption of Watson in the enterprise. IBM has certainly made some progress in this area but more work is necessary for mainstream adoption.

4 — Edge Edition

Watson is tightly integrated into IBM’s IOT offerings. Taking this vision a step further, some Watson capabilities are well positioned for edge computing devices. A lighter edition of Watson that works on IOT devices could be a great addition to the platform. This will provide an alternative to technologies such as Azure IOT Edge.

5 — Open Source Tools

Google open source Tensorflow, Microsoft has its cognitive Toolkit, Github is flown with innovative AI tools and frameworks and yet Watson’s open source contributions remain relatively small. Open sourcing tools and frameworks that integrate with the Watson platform could help to gain the hearts of the AI developer community.

6 — Training Tools and Frameworks

Continuing with the previous points, the Watson developer community could benefit from new tools and frameworks that help with the training and optimization of AI models. Frameworks such as OpenAI Gym or DeepMind Lab can provide some inspiration in this area.

7 — Killer Consumer Apps

Easier said than done but I believe Watson should expand its integration with popular consumer applications such as games or digital assistants. That strategy will directly expose Watson to millions of consumers and will validate some of its core capabilities.