By: David Norfolk, Practice Leader - Development, Bloor Research
Published: 1st February 2014
Copyright Bloor Research © 2014
There is something new and big happening at IBM for what it calls the Cognitive Computing Era—basically, a new ecosystem group around Watson that could build into something as big as J2EE, and make what is fundamentally 'artificial intelligence' applied to textual and similar non-relational assets, for decision support, into a reality.
OK, Watson is clever and is becoming available more widely as affordable PaaS and SaaS solutions, but 'all' it does is understand large amounts of reasonably well-structured English text, after it has been trained on good-quality material. It doesn't (and won't ever) do mathematical analyses and maths generally, it struggles with ungrammatical tweet-speak, and it doesn't do other languages—yet. But someone might start teaching it to understand code sometime, which has interesting possibilities, perhaps around extracting business requirements...
That still leaves a huge set of problems that it could help with today (it operates in conjunction with other technologies rather than by replacing them). I think that Watson is genuinely innovative and useful technology. Some people seem to have doubts but when Rob Russell of Forbes, for example, expresses them, he talks about "the genuine emotion, sincere sympathy, and earned experience...found with a human advisor" in the context of banks and financial advisers (many of whom, or their employers, in the UK, are up before the regulators today, and paying huge compensation sums for giving insincere and misleading advice). I think that people get a bit sentimental about human advice. I still want to speak with a human but when my chemist won't sell me decongestants for my cold because they might interfere with my blood pressure control medicine and my doctor then tells me that it really doesn't matter, because my blood pressure is being well controlled anyway, I get irritated. I start wishing that my human chemist had access to a little Watson-based system telling him or her that their little "rule of thumb" isn't black and white—because Watson can make sense of more documents than my chemist can cope with, including one from a doctor explaining the circumstances that make the risk small....
I know that IBM is really taking 'Cognitive Computing' seriously when I join a Global Rational User Group webinar and hear Carson Holmes (EVP, Software Development Experts) questioning Carl Kraenzel (Director of the IBM Watson managed services group, IBM Distinguished Engineer) about using the IBM DevOps solution (and Rational Team Concert) to support continuous delivery of Watson solutions. There's a case study here.
Everybody can read the case study. The interesting bit I'd highlight is the discussion of collaborative cost models, in the context of DevOps and of delivering cost-effective innovation. To quote the case study: "Traditionally, cost models are created and reviewed exclusively by the operations teams that host and run the systems. Developers and development managers rarely see cost models and are almost never asked to help author them. As a result, development teams can unintentionally create code or prioritize features in ways that drive unintended costs, many of which manifest as deployment difficulties and delays. To prevent this issue, the IBM Watson operations team engaged lead software developers and managers as co-authors and equal owners of the cost model. This collaborative approach to cost modeling increases transparency and helps foster a culture in which everyone understands and cares about how their choices affect operations".
DevOps has more implications than some people are aware of—and one is that building systems that are operationally costly to implement and use becomes something that developers should be aware of—and do something about, before operational staff are affected. I'm starting to think that adopting DevOps is a really good metric for an organisation's real interest in improving organisational maturity!
We automatically stop accepting comments 180 days after a post is published. If you would like to know more about this subject, please contact us and we'll try to help.
Published by: electronicdawn Ltd.