August 05, 2024

Faceoff: Auditable AI Versus the AI Blackbox Problem

“The notion of auditable AI extends beyond the principles of responsible AI, which focuses on making AI systems robust, explainable, ethical, and efficient. While these principles are essential, auditable AI goes a step further by providing the necessary documentation and records to facilitate regulatory reviews and build confidence among stakeholders, including customers, partners, and the general public,” says Adnan Masood ... “There are two sides of auditing: the training data side, and the output side. The training data side includes where the data came from, the rights to use it, the outcomes, and whether the results can be traced back to show reasoning and correctness,” says Kevin Marcus. “The output side is trickier. Some algorithms, such as neural networks, are not explainable, and it is difficult to determine why a result is being produced. Other algorithms such as tree structures enable very clear traceability to show how a result is being produced,” Marcus adds. ... Developing explainable AI remains the holy grail and many an AI team is on a quest to find it. Until then, several efforts are underway to develop various ways to audit AI in order to have a stronger grip over its behavior and performance.?

A developer’s guide to the headless data architecture

We call it a “headless” data architecture because of its similarity to a “headless server,” where you have to use your own monitor and keyboard to log in. If you want to process or query your data in a headless data architecture, you will have to bring your own processing or querying “head” and plug it into the data — for example, Trino, Presto, Apache Flink, or Apache Spark. A headless data architecture can encompass multiple data formats, with data streams and tables as the two most common. Streams provide low-latency access to incremental data, while tables provide efficient bulk-query capabilities. Together, they give you the flexibility to choose the format that is most suitable for your use cases, whether it’s operational, analytical, or somewhere in between. ... Many businesses today are building their own headless data architectures, even if they’re not quite calling it that yet, though using cloud services tends to be the easiest and most popular way to get started. If you’re building your own headless data architecture, it’s important to first create well-organized and schematized data streams, before populating them into Apache Iceberg tables.

The Hidden Costs of the Cloud Skills Gap

Properly managing and scaling cloud resources requires expertise in load balancing, auto-scaling, and cost optimization. Without these skills, companies may face inefficiencies, either by over-provisioning or under-utilizing resources. Inexperienced or overstretched staff might struggle with performance optimization, resulting in slower applications and services, which can negatively impact user satisfaction and harm the company's reputation. ... Employees lacking the necessary skills to fully leverage cloud technologies may be less likely to propose innovative solutions or improvements, potentially leading to a lack of new product development and stagnation in business growth. The cloud presents abundant opportunities for innovation, including AI, machine learning, and advanced data analytics. Companies without the expertise to implement these technologies risk missing out on significant competitive advantages and exciting new discoveries. The bottom line is that skilled professionals often drive the adoption of new technologies because they have the knowledge to experiment in the field.

Architectural Retrospectives: The Key to Getting Better at Architecting

The traditional architectural review, especially if conducted by outside parties, often turns into a blame-assignment exercise. The whole point of regular architectural reviews in the MVA approach is to learn from experience so that catastrophic failures never occur. ... The mechanics of running an architectural retrospective session are identical to those of running a Sprint Retrospective in Scrum. In fact, an architectural focus can be added to a more general-purpose retrospective to avoid creating yet another meeting, so long as all the participants are involved in making architectural decisions. This can also be an opportunity to demonstrate that anyone can make an architectural decision, not only the "architects." ... Many teams skip retrospectives because they don’t like to confront their shortcomings, Architectural retrospectives are even more challenging because they examine not just the way the team works, but the way the team makes decisions. But architectural retros have great pay-offs: they can uncover unspoken assumptions and hidden biases that prevent the team from making better decisions. If you retrospect on the way that you create your architecture, you will get better at architecting.

Design flaw has Microsoft Authenticator overwriting MFA accounts, locking users out

Microsoft confirmed the issue but said it was a feature not a bug, and that it was the fault of users or companies that use the app for authentication. Microsoft issued two written statements to CSO Online but declined an interview. Its first statement read: “We can confirm that our authenticator app is functioning as intended. When users scan a QR code, they will receive a message prompt that asks for confirmation before proceeding with any action that might overwrite their account settings. This ensures that users are fully aware of the changes they are making.” One problem with that first statement is that it does not correctly reflect what the message says. The message says: “This action will overwrite existing security information for your account. To prevent being locked out of your account, continue only if you initiated this action from a trusted source.” The first sentence of the warning window is correct, in that the action will indeed overwrite the account. But the second sentence incorrectly tells the user to proceed as long as two conditions are met: that the user initiated the action; and that it is a trusted source.

Automation Resilience: The Hidden Lesson of the CrowdStrike Debacle

Automated updates are nothing new, of course. Antivirus software has included such automation since the early days of the Web, and our computers are all safer for it. Today, such updates are commonplace – on computers, handheld devices, and in the cloud. Such automations, however, aren’t intelligent. They generally perform basic checks to ensure that they apply the update correctly. But they don’t check to see if the update performs properly after deployment, and they certainly have no way of rolling back a problematic update. If the CrowdStrike automated update process had checked to see if the update worked properly and rolled it back once it had discovered the problem, then we wouldn’t be where we are today. ... The good news: there is a technology that has been getting a lot of press recently that just might fit the bill: intelligent agents. Intelligent agents are AI-driven programs that work and learn autonomously, doing their good deeds independently of other software in their environment. As with other AI applications, intelligent agents learn as they go. Humans establish success and failure conditions for the agents and then feed back their results into their models so that they learn how to achieve successes and avoid failures.

Read more here ...