Daily Tech Digest - July 12, 2024

4 considerations to help organizations implement an AI code of conducts

Many organizations consider reinventing the wheel to accommodate AI tools, but this creates a significant amount of unnecessary work. Instead, they should subject any AI tool to the same rigorous procurement process that applies to any product that concerns data security. The procurement process must also take into consideration the organization’s privacy and ethical standards, to ensure these are never compromised in the name of new technology. ... It’s important to be conscious of the privacy policies of AI tools when using these in an enterprise environment — and be sure to only use these with a commercial license. To address this risk, an AI code of conduct should stipulate that free tools are categorically banned for use in any business context. Instead, employees should be required to use an approved, officially procured commercial license solution, with full privacy protections. ... Every organization needs to remain aware of how their technology vendors use AI in the products and services that they buy from them. To enable this, an AI code of conduct should also enforce policies to enable organizations to keep track of their vendor agreements.


From Microservices to Modular Monoliths

You know who really loves microservices? Cloud hosting companies like Microsoft, Amazon, and Google. They make a lot of money hosting microservices. They also make a lot of money selling you tools to manage your microservices. They make even more money when you have to scale up your microservices to handle the increased load on your system. ... So what do you do when you find yourself in microservice hell? How do you keep the gains you (hopefully) made in breaking up your legacy ball of mud, without having to constantly contend with a massively distributed system? It may be time to (re)consider the modular monolith. A modular monolith is a monolithic application that is broken up into modules. Each module is responsible for a specific part of the application. Modules can communicate with each other through well-defined interfaces. This allows you to keep the benefits of a monolithic architecture, while still being able to break up your application into smaller, more manageable pieces. Yes, you'll still need to deal with some complexity inherent to modularity, such as ensuring modules remain independent while still being able to communicate with one another efficiently. 


Deep Dive: Optimizing AI Data Storage Management

In an AI data pipeline, various stages align with specific storage needs to ensure efficient data processing and utilization. Here are the typical stages along with their associated storage requirements: Data collection and pre-processing: The storage where the raw and often unstructured data is gathered and centralized (increasingly into Data Lakes) and then cleaned and transformed into curated data sets ready for training processes. Model training and processing: The storage that feeds the curated data set into GPUs for processing. This stage of the pipeline also needs to store training artifacts such as the hyper parameters, run metrics, validation data, model parameters and the final production inferencing model. Inferencing and model deployment: The mission-critical storage where the training model is hosted for making predictions or decisions based on new data. The outputs of inferencing are utilized by applications to deliver the results, often embedded into information and automation processes. Storage for archiving: Once the training stage is complete, various artifacts such as different sets of training data and different versions of the model need to be stored alongside the raw data.


RAG (Retrieval Augmented Generation) Architecture for Data Quality Assessment

RAG is basically designed to leverage LLMs on your own content or data. It involves retrieving relevant content to augment the context or insights as part of the generation process. However, RAG is an evolving technology with both strengths and limitations. RAG integrates information retrieval from a dedicated, custom, and accurate knowledge base, reducing the risk of LLMs offering general or non-relevant responses. For example, when the knowledge base is tailored to a specific domain (e.g., legal documents for a law firm), RAG equips the LLM with relevant information and terminology, improving the context and accuracy of its responses. At the same time, there are limitations associated with RAG. RAG heavily relies on the quality, accuracy, and comprehensiveness of the information stored within the knowledge base. Incomplete, inaccurate or missing information or data can lead to misleading or irrelevant retrieved data. Overall, the success of RAG hinges on quality data. So, how are RAG models implemented? RAG has basically two key components: a retriever model and a generator model.
 

NoSQL Database Growth Has Slowed, but AI Is Driving Demand

As for MongoDB, it too is targeting generative AI use cases. In a recent post on The New Stack, developer relations team lead Rick Houlihan explicitly compared its solution to PostgreSQL, a popular open source relational database system. Houlihan contended that systems like PostgreSQL were not designed for the type of workloads demanded by AI: “Considering the well-known performance limitations of RDBMS when it comes to wide rows and large data attributes, it is no surprise that these tests indicate that a platform like PostgreSQL will struggle with the kind of rich, complex document data required by generative AI workloads.” Unsurprisingly, he concludes that using a document database (like MongoDB) “delivers better performance than using a tool that simply wasn’t designed for these workloads.” In defense of PostgreSQL, there is no shortage of managed service providers for Postgres that provide AI-focused functionality. Earlier this year I interviewed a “Postgres as a Platform” company called Tembo, which has seen a lot of demand for AI extensions. “Postgres has an extension called pgvector,” Tembo CTO Samay Sharma told me.


Let’s Finally Build Continuous Database Reliability! We Deserve It

While we worked hard to make sure our CI/CD pipelines are fast and learned how to deploy and test applications reliably, we didn’t advance our databases world. It’s time to get continuous reliability around databases as well. To do that, developers need to own their databases. Once developers take over the ownership, they will be ready to optimize the pipelines, thereby achieving continuous reliability for databases. This shift of ownership needs to be consciously driven by technical leaders. ... The primary advantage of implementing database guardrails and empowering developers to take ownership of their databases is scalability. This approach eliminates team constraints, unlocking their complete potential and enabling them to operate at their optimal speed. By removing the need to collaborate with other teams that lack comprehensive context, developers can work more swiftly, reducing communication overhead. Just as we recognized that streamlining communication between developers and system engineers was the initial step, leading to the evolution into DevOps engineers, the objective here is to eliminate dependence on other teams. 


Digital Transformation: Making Information Work for You

With information generated by digital transactions, the first goal is to ensure that the knowledge garnered does not get stuck between only those directly participating in the transaction. Lessons learned from the transaction should become part of the greater organizational memory. This does not mean that every single transaction needs to be reported to every person in the organization. It also doesn’t mean that the information needs to be elevated in the same form or at the same velocity to all recipients. Those participating in the transaction need an operational view of the transaction. This needs to happen in real time. The information is the enabler of the human-to-computer-to-human transaction and the speed of that information flow needs to be as quick as it was in the human-to-human transaction. Otherwise, it will be viewed as a roadblock instead of an enabler. As it escalates to the next level of management, the information needs to evolve to a managerial view. Managers are more interested in anomalies and outliers or data at a summary level. This level of information is no less impactful to the organizational memory but is associated with a different level of decision-making. 


Generative AI won’t fix cloud migration

The allure of generative AI lies in its promise of automation and efficiency. If cloud migration was a one-size-fits-all scenario, that would work. But each enterprise faces unique challenges based on its technological stack, business requirements, and regulatory environment. Expecting a generative AI model to handle all migration tasks seamlessly is unrealistic. ... Beyond the initial investment in AI tools, the hidden costs of generative AI for cloud migration add up quickly. For instance, running generative AI models often requires substantial computational resources, which can be expensive. Also, keeping generative AI models updated and secure demands robust API management and cybersecurity measures. Finally, AI models need continual refinement and retraining to stay relevant, incurring ongoing costs. ... Successful business strategy is about what works well and what needs to be improved. We all understand that AI is a powerful tool and has been for decades, but it needs to be considered carefully—once you’ve identified the specific problem you’re looking to solve. Cloud migration is a complex, multifaceted process that demands solutions tailored to unique enterprise needs. 


Navigating Regulatory and Technological Shifts in IIoT Security

Global regulations play a pivotal role in shaping the cybersecurity landscape for IIoT. The European Union’s Cyber Resilience Act (CRA) is a prime example, setting stringent requirements for manufacturers supplying products to Europe. By January 2027, companies must meet comprehensive standards addressing security features, vulnerability management, and supply chain security. ... The journey towards securing IIoT environments is multifaceted, requiring manufacturers to navigate regulatory requirements, technological advancements, and proactive risk management strategies. Global regulations like the EU’s Cyber Resilience Act set critical standards that drive industry-wide improvements. At the same time, technological solutions such as PKI and SBOMs play essential roles in maintaining the integrity and security of connected devices. By adopting a collaborative approach and leveraging robust security frameworks, manufacturers can create resilient IIoT ecosystems that withstand evolving cyber threats. The collective effort of all stakeholders is paramount to ensuring the secure and reliable operation of industrial environments in this new era of connectivity.


Green Software Foundation: On a mission to decarbonize software

One of the first orders of business in increasing awareness: getting developers and companies to understand what green software really is. Instead of reinventing the wheel, the foundation reviewed a course in the concepts of green software that Hussain had developed while at Microsoft. To provide an easy first step for organizations to take, the foundation borrowed from Hussain’s materials and created a new basic training course, “Principles of Green Software Engineering.” The training is only two or three hours long and level-sets students to the same playing field. ... When it comes to software development, computing inefficiencies (and carbon footprints) are more visible — bulky libraries for example — and engineers can improve it more easily. Everyday business operations, on the other hand, are a tad opaque but still contribute to the company’s overall sustainability score. Case in point: The carbon footprint of a Zoom call is harder to measure, Hussain points out. The foundation helped to define a Software Carbon Intensity (SCI) score, which applies to all business operations including software development and SaaS programs employees might use.



Quote for the day:

"Real leadership is being the person others will gladly and confidently follow." -- John C. Maxwell

No comments:

Post a Comment