The KIT -- Knowledge and Information Technology -- No. 288

The KIT

Knowledge & Information Technology

No. 288 - 17 May 2021

In this Issue

· AI/ML Trustworthiness Issues

· IT Dependency Management

· Analytics Setup Guidebook

· Microsoft and Data Residency

· Personal Knowledge

Management

Our Consulting Services

IT Strategy and Roadmap

Enterprise Architecture

Business Process Modeling

Enterprise Software Selection

IT Innovation Briefings

IT Due Diligence for M&A

Cloud Computing Adoption

Enterprise Security Maturity

Knowledge Management

Communities of Practice

Taxonomy Development

Enterprise Social Media Adoption

For more information

Visit us: www.cebe-itkm.com

E-mail us: info@cebe-itkm.com

Phone: +1 415 870 4856

+33 970 444 992

Twitter: @cbaudoin

See: Previous KIT issues

Forward this issue to colleagues and friends!

Key Questions in AI and ML Trustworthiness

During a call last week with people from a research center on AI and their defense sector client, the following four questions were submitted for discussion. We find this interesting as it applies to many other domains, including civilian autonomous vehicles.

How do you verify that unintended bias is not incorporated during development or learning?
What has been the most efficient V&V (validation and verification) approach during the development process?
Do industry standards exist that contribute or hinder certification of AI/ML applications?
What system redundancies are or should be in place to ensure that AI systems can work in environments where sensors are disabled or data links are denied?

In the last question, "denial" would refer to hostile jamming in a defense context. But even outside of military operations, loss of signal for any other reason, as well as malicious hacker activity, poses the same issues.

There were no definite answers on that call -- only a lot of good thinking. What do you think? Let us know and will publish the best comments in the next issue.

IT Dependency Management from a Risk Perspective

Back in March, we looked at 8folios, a young company (12 people) that offers a software-as-a-service application to manage software and system dependencies that pose risks, including with respect to data protection regulations.

The product is not a traditional portfolio management tool (it overlaps with such tools, but does not capture the same attributes); it should really be described as an "IT knowledge management tool." With it, IT users record dependency information that is often dispersed across documents, spreadsheets, architecture diagrams, notes, and people's heads. This information may include applications, data elements, integration points, technologies, domain names, security certificates, and more. The result is the ability to analyze the risk posed by those dependencies, the opportunities for consolidation, possible compliance issues, etc. Like all KM applications, 8folios' success is highly dependent on the quality and completeness of information entered by users.

Analytics Setup Guidebook

One reviewer said it best: "I'm shocked to be telling you this next sentence: I read a free ebook from a company and actually loved it." This praise goes to The Analytics Setup Guidebook from Holistics, which covers topics that include, listed in a more logical order than in the book itself:

modeling data for analytics
consolidating and transforming data, and the role of the data warehouse
setting up and deploying a simple and modern analytics stack
evaluating business intelligence (BI) applications

Microsoft and Data Residency

In a May 6 announcement, rather grandiosely entitled Answering Europe's Call: Storing Data in the EU, Microsoft's Brad Smith wrote that "we will go beyond our existing data storage commitments and enable you to process and store all your data in the EU." This applies to "all of Microsoft’s core cloud services – Azure, Microsoft 365, and Dynamics 365."

This is good news, but is overdue and not as impressive as it sounds, for several reasons. One is that Microsoft didn't really have a choice. They're not really "answering the call" from Europe, like someone who volunteers to join a noble mission: they're reacting to customers who, as data owners or data custodians, said "we can't use Microsoft's cloud offerings without violating the GDPR unless you can guarantee that the data does not leave the EU." Moreover, Microsoft offered such a guarantee to the U.S. government several years ago with their "Microsoft Office 365 Government Cloud." They can't even claim to be just now focusing on European residency needs, since they offered in January 2017 a "differentiated option" of Office 365 (now Microsoft 365) hosted at a data center in Germany to comply with German and (pre-GDPR) EU laws. But let's not facts get in the way of marketing...

Personal Knowledge Management

The somewhat unexplored world of PKM is starting to get richer, and The Sweet Setup has just issued a comparative review of two offerings, Obsidian and Roam. It is increasingly clear that PKM is essentially a combination of three things:

note-taking -- a space originally occupied by OneNote (from Microsoft) and Evernote
wikis -- good personal wikis were notoriously hard to find and deploy
knowledge graphs -- including the ability to visualize and traverse the connections between ideas

These two products (as well as others such as TheBrain or DevonThink) are starting to effectively package those capabilities together. Will the older products keep up or become obsolete? And will there be standards to ensure that those new tools, once you store your knowledge in them, do not lock you in or put you at risk if the data is in the cloud and the provider disappears? It's fine to say "you can export all your data in JSON format," but what if no one else can read it? We're waiting for answers.

Thanks to Claude Bardy, a Paris-based consultant, for helping us track this space. He and Claude Baudoin are planning a couple of papers on this topic: one about the concepts, strategy and use cases of PKM, the other about the requirements and selection of a PKM tool.

Seen Recently...

"When you tweet something, your friends read it in your voice and with all the knowledge of you. Other people read it like a document is read in court. Devoid of character and context."

-- Emily Freeman, @editingemily