In defense of IBM Watson

IBM takes a lot of criticism about its Watson business, including from me, but what if Watson ends up
In defense of IBM Watson
By ARCHITECHT • Issue #103

IBM takes a lot of criticism about its Watson business, including from me, but what if Watson ends up living up the the hype and its critics all end up eating crow? There’s a good article in the MIT Technology Review on Tuesday that looks at this question in the realm of health care—where IBM made some of its boldest claims and suffered some of its most notable setbacks—and raises some very good points about why there’s still a lot of promise in Watson.
Among them (paraphrasing):
  • Watson’s shortcomings appear worse because IBM overhyped its capabilities way too early on.
  • IBM still has the money and the partnership-making prowess to get access to health data, which is still a big challenge for anyone trying to operate in this space.
  • IBM Watson Health has some major executives who are also physicians, and have real-world insights into what the technology ultimately needs to deliver.
  • Watson actually does have some good technology behind it, and actually has delivered some meaningful results for customers and partners.
I think the article is probably a relatively accurate assessment of IBM’s current situation with regard to Watson in health care, and very probably applies to the Watson business overall. IBM promised the moon before the technology was really there to deliver and vowed to make Watson a $10 billion business (by 2023) before anybody really had an idea how to make money selling AI software. Big Blue tried to force its vision into reality by investing in Watson-based startups and applying Watson to seemingly every business under the sun, likely hoping something would catch on.
Now we’re at the point where deep learning has recast the conversation about AI, and where IBM has competition everywhere it looks. But it’s also still early enough that IBM—assuming it’s updating Watson’s underlying technologies to account for today’s state of the art—has just a good of chance of cracking major industries like health care as does anybody else. 
Its biggest hurdle might be the type of over-the-top marketing for which IBM notoriously has a penchant, and which its AI competitors largely very much do not. If IBM wants Watson to be taken seriously in the AI space, a little restraint will go a long way.
In semi-unrelated IBM news, here’s a story about how IBM is trying to talk some sense into Congress about the future of artificial intelligence. Bad regulations based on misunderstanding or outdated definitions can stifle technological advances, or at least lead to some really sticky legal situations. We’re seeing that right now with Microsoft and the Justice Department possibly headed to the Supreme Court to settle their case over data privacy.
The more power to any tech companies and any AI experts who can help ensure the United States ends up with laws and regulations that account for some of the riskier parts of AI while still encouraging a fast pace of innovation.

Sponsor: DigitalOcean
Sponsor: DigitalOcean
Highlights of the latest ARCHITECHT AI podcast
Melonee Wise of Fetch Robotics on why robots belong in warehouses
Fetch Robotics CEO Melonee Wise explains her company’s warehouse robots, as well as why today’s on-demand economy requires robots, and why people really like working with them.
Artificial intelligence
Its business model of selling to enterprises seems sound, especially if its technology allows for retro-fitting (which I believe it does). Also, Andrew Ng is now on the company’s board of directors.  •  Share
This link is to a partnership with Volvo around Nvidia’s in-car technology platform, while this one points to a story about how Volkswagen wants to use AI on transportation in general. VW is also working with D-Wave on quantum computing.
Like DeepMind, it’s coming straight outta Oxford. Unlike DeepMind, it has a clear and potentially lucrative product focus. It seems like this is becoming a more crowded space recently, though.
The unique thing about this is probably the crowdsourced nature of learning threat remediation, but everything has to have an AI model today, as well. Want proof? Check out this new analysis of AI in cybersecurity by CB Insights.
This survey, which is sponsored by chip-designer ARM, suggests people are less worried about AI than a lot of experts are. Many are even excited about it. The full report is here.
Bonsai has a novel concept for helping enterprises train reinforcement learning models, but this will help them rely less on the Bonsai platform if they already have some internal AI skills and models.  •  Share
This is actually really cool research, because many people won’t suffer any sort of assistant that forces an unnatural conversation or usage. Now to work this into existing NLP models …
This is directly related (in theory) to the research above, and also to my thoughts about IBM Watson. People want and expect a certain user experience from AI, and anything that doesn’t live up is a disappointment.
The fight is on among universities trying to attract AI talent, not only in the U.S. but also across Canada, Europe and China. Having a cross-discipline AI group could be a smart idea, rather than having a bunch of related but disconnected departments.
If that’s your thing, here’s a dataset including 6.7 million examples on robot-gripping.
Sponsor: CircleCI
Sponsor: CircleCI
Will this project to help the Snow Leopard Trust better identify snow leopards from video footage and cameras actually save the species? Probably not, but I’m all for trying and for applying deep learning in an attempt to save the planet. 
Cloud and infrastructure
The size of this fund aside, it’s remarkable how important Cloudflare has become not just security, but to web infrastructure and performance overall. If you’re looking for the big upcoming tech IPO, this could very well be it someday.  •  Share
This makes perfect sense given (a) the enterprise nature of many Box customers and (b) the amount of Microsoft they already use for documents, etc. Box co-founder Sam Ghods suggested this was coming on the ARCHITECHT Show podcast in February.
Disney is a remarkably progressive company when it comes to adopting new application architectures and technologies. This has some good info on how and why it adopted DevOps, and you can easily track down info on Disney and Docker, too.
Anisble, OpenShift and OpenStack accounted for 17 million-plus-dollar deals in the last quarter. Red Hat also predicts “a $200 million run rate in [cloud OS licenses] in the second quarter.”
I’m not sure how many people are following Microsoft’s stripped-down Nano Server, but it’s telling that they’ll now only be available as container images and not on bare metal. Why reinvent the wheel when everyone’s already looking at containers?
Sponsor: Bonsai
Sponsor: Bonsai
All things data
That is a lot of dough for the company, which analyzes log files. I don’t follow this space too closely, but I assume Splunk still dominates, and there are strong open source options, as well.
I’m surprised I didn’t read about this anywhere besides the press release wire, considering Khosla Ventures led this investment, and Andy Bechtolsheim and Diane Greene also invested. From what I gather, Xcalar is essentially a visual analytics tool that connects directly to Hadoop and other stores.
As I get older, I find data ruins the sports-fan experiences, but I’d probably want a lot of it if I spent hundreds of millions buying a team and wanted to recoup my investment. 
This is good advice from Silicon Valley Data Science, whose team knows a thing or two about data. What’s remarkable is that we’re still having this discussion, despite nearly a decade of talking about big data. I think we might have focused too much attention on the “big” part early on.  •  Share
Sponsor: Cloudera
Sponsor: Cloudera
Did you enjoy this issue?
The most interesting news, analysis, blog posts and research in cloud computing, artificial intelligence and software engineering. Delivered daily to your inbox. Curated by Derrick Harris. Check out the Architecht site at
Carefully curated by ARCHITECHT with Revue. If you were forwarded this newsletter and you like it, you can subscribe here. If you don't want these updates anymore, please unsubscribe here.