Daily Tech Digest - January 31, 2024

Rethinking Testing in Production

With products becoming more interconnected, trying to accurately replicate third-party APIs and integrations outside of production is close to impossible. Trunk-based development, with its focus on continuous integration and delivery, acknowledges the need for a paradigm shift. Feature flags emerge as the proverbial Archimedes lever in this transformation, offering a flexible and controlled approach to testing in production. Developers can now gradually roll out features without disrupting the entire user base, mitigating the risks associated with traditional testing methodologies. Feature flags empower developers to enable a feature in production for themselves during the development phase, allowing them to refine and perfect it before exposing it to broader testing audiences. This progressive approach ensures that potential issues are identified and addressed early in the development process. As the feature matures, it can be selectively enabled for testing teams, engineering groups or specific user segments, facilitating thorough validation at each step. The logistic nightmare of maintaining identical environments is alleviated, as testing in production becomes an integral part of the development workflow.

Enterprise Architecture in the Financial Realm

Enterprise architecture emerges as the North Star guiding banks through these changes. Its role transcends being a mere operational construct; it becomes a strategic enabler that harmonizes business and technology components. A well-crafted enterprise architecture lays the foundation for adaptability and resilience in the face of digital transformation. Enterprise architecture manifests two key characteristics: unity and agility. The unity aspect inherently provides an enterprise-level perspective, where business and IT methodologies seamlessly intertwine, creating a cohesive flow of processes and data. Conversely, agility in enterprise architecture construction involves deconstruction and subsequent reconstruction, refining shared and reusable business components, akin to assembling Lego bricks. ... Quantifying the success of digital adaptation is crucial. Metrics should not solely focus on financial outcomes but also on key performance indicators reflecting the effectiveness of digital initiatives, customer satisfaction, and the agility of operational models.

Cloud Security: Stay One Step Ahead of the Attackers

The relatively easy availability of cloud-based storage can lead to a data sprawl that is uncontrolled and unmanageable. In many cases, data which must be deleted or secured is left ungoverned, as organizations are not aware of their existence. In April 2022, cloud data security firm, Cyera, found unmanaged data store copies, and snapshots or log data. The researchers from this firm found out that 60% of the data security issues present in cloud data stores were due to unsecured sensitive data. The researchers further observed that over 30% of scanned cloud data stores were ghost data, and more than 58% of these ghost data stores contained sensitive or very sensitive data. ... Despite best practices advised by cloud service providers, data breaches that originate in the cloud have only increased. IBM’s annual Cost of a Data Breach report for example, highlights that 45% of studied breaches have occurred in the cloud. What is also noteworthy is that a significant 43% of reporting organizations which have stated they are just in the early stages or have not started implementing security practices to protect their cloud environments, have observed higher breach costs.

Five Questions That Determine Where AI Fits In Your Digital Transformation Strategy

Once you understand the why and the what, only then can you consider how your organization can use insights from AI to better accomplish its goals. How will your people respond, and how will they benefit? Today’s organizations have multiple technology partners, and they may have many that are all saying they can do AI. But how will your organization work with all those partners to make an AI solution come together? Many organizations are developing AI policies to define how it can be used. Having these guardrails ensures that your organization is operating ethically, morally and legally when it comes to the use of AI. ... It’s important to consider whether your organization is truly ready for AI at an enterprise or divisional level before deciding to implement AI at scale. Pilot projects can help you determine whether the implementation is generating the intended results and better understand how end users will interact with the processes. If you can't achieve customization and personalization across the organization, AI initiatives will be much tougher to implement.

A Dive into the Detail of the Financial Data Transparency Act’s Data Standards Requirements

The act is a major undertaking for regulators and regulated firms. It is also an opportunity for the LEI, if selected, to move to another level in the US, which has been slow to adopt the identifier, and significantly increase numbers that will strengthen the Global LEI System. While industry experts suggest regulators in scope of FDTA, collectively called Financial Stability Oversight Council (FSOC) agencies, initially considered data standards including the LEI and Financial Instrument Global Identifier published by Bloomberg, they suggest the LEI is the best match for the regulation’s requirements for ‘Covered agencies to establish “common identifiers” for information reported to covered regulatory agencies, which could include transactions and financial products/instruments.” ... The selection and implementation of a reporting taxonomy is more challenging as it will require many of the regulators to abandon existing reporting practices often based on PDFs, text and CSV files, and replace these with electronic reporting and machine-readable tagging. XBRL fits the bill, say industry experts, although there has been pushback from some agencies that see the unfunded requirement for change as too great a burden.

Data Center Approach to Legacy Modernization: When is the Right Time?

Legacy systems can lead to inefficiencies in your business. If we take one of the parameters mentioned above, such as cooling, one example of inefficiency could lie within an old server that’s no longer of use but still turned on. This could be placing unneccesary strain on your cooling, thus impacting your environmental footprint. Legacy systems may no longer be the most appropriate for your business, as newer technologies emerge that offer a more efficient method of producing the same, or better, results. If you neglect this technology, you might be giving your competitors an advantage which could be costly for your business. ... A cyber-attack takes place every 39 seconds, according to one report. This puts businesses at risk of losing or compromising not only their intellectual property and assets but also their customer’s data. This could put you at risk of damaging your reputation and even facing regulation fines. One of the best reasons to invest in digital transformation is for the security of your business. Systems that no longer receive updates can become a target of cyber-attacks and act as a vulnerability within your technology infrastructure. 

4 paths to sustainable AI

Hosting AI operations at a data center that uses renewable power is a straightforward path to reduce carbon emissions, but it’s not without tradeoffs. Online translation service Deepl runs its AI functions from four co-location facilities: two in Iceland, one in Sweden, and one in Finland. The Icelandic data center uses 100% renewably generated geothermal and hydroelectric power. The cold climate also eliminates 40% or more of the total data center power needed to cool the servers because they open the windows rather than use air conditioners, says Deepl’s director of engineering Guido Simon. Cost is another major benefit, he says, with prices of five cents per KW/hour compared to about 30 cents or more in Germany. The network latency between the user and a sustainable data center can be an issue for time-sensitive applications, says Stent, but only in the inference stage, where the application provides answers to the user, rather than the preliminary training phase. Deepl, with headquarters in Cologne, Germany, found it could run both training and inference from its remote co-location facilities. “We’re looking at roughly 20 milliseconds more latency compared to a data center closer to us,” says Simon.

Can ChatGPT drive my car? The case for LLMs in autonomy

Autonomous driving is an especially challenging problem because certain edge cases require complex, human-like reasoning that goes far beyond legacy algorithms and models. LLMs have shown promise in going beyond pure correlations to demonstrating a real “understanding of the world.” This new level of understanding extends to the driving task, enabling planners to navigate complex scenarios with safe and natural maneuvers without requiring explicit training. ... Safety-critical driving decisions must be made in less than one second. The latest LLMs running in data centers can take 10 seconds or more. One solution to this problem is hybrid-cloud architectures that supplement in-car compute with data center processing. Another is purpose-built LLMs that compress large models into form factors small enough and fast enough to fit in the car. Already we are seeing dramatic improvements in optimizing large models. Mistral 7B and Llama 2 7B have demonstrated performance rivaling GPT-3.5 with an order of magnitude fewer parameters (7 billion vs. 175 billion). Moore’s Law and continued optimizations should rapidly shift more of these models to the edge.

The Race to AI Implementation: 2024 and Beyond

The biggest problem is that the competitive and product landscape will be undergoing massive flux, so picking a strategic solution will be increasingly difficult. Younger companies that are less likely to be able to handle the speed of these advancements should focus on openness so that if they fail, someone else can pick up support, interoperability, and compatibility. If you aren’t locked into a single vendor’s solution and can mix and match as needed, you can move on or off a platform based on your needs. Like any new technology, take advice about hardware selection from the platform supplier. This means that if you are using ChatGPT, you want to ask OpenAI for advice about new hardware. If you are working with Microsoft or Google or any other AI developer, ask them what hardware they would recommend. ... You need a vendor that embraces all the client platforms for hybrid AI and one with a diverse, targeted solution set that individually focuses on the markets your firm is in. Right now, only Lenovo seems to have all the parts necessary thanks to its acquisition of Motorola.

Quote for the day:

"It's fine to celebrate success but it is more important to heed the lessons of failure." -- Bill Gates

No comments:

Post a Comment