Daily Tech Digest - April 26, 2022

The emerging risks of open source

Many enterprises have sought to make their open source lives easier by buying into managed services. It’s a great short-term fix, but it doesn’t solve the long-term issue of sustainability. No, the cloud hyperscalers aren’t strip miners, nefariously preying on the code of unsuspecting developers. But too often some teams fail to plan to contribute back to the projects upon which they depend. I stress some, as this tends to not be a corporation-wide issue, no matter the vendor. I’ve detailed this previously. Regardless, the companies offering these managed services tend to not have any control over the projects’ road maps. That’s not great for enterprises that want to control risk. Google is a notable exception—it tends to contribute a lot to key projects. Nor can they necessarily contribute directly to projects. As Mugrage indicates, for companies like Netflix or Facebook (Meta) that open source big projects, these “open source releases are almost a matter of employer branding—a way to show off their engineering chops to potential employees,” which means “you’re likely to have very little sway over future developments.” 


How to model uncertainty with Dempster-Shafer’s theory?

One of the main advantages of this theory is that we can utilize it for generating a degree of belief by taking all the evidence into account. This evidence can be obtained from different sources. The degree of belief using this theory can be calculated by a mathematical function called the belief function. We can also think of this theory as a generalization of the Bayesian theory of subjective probability. While talking about the degree of belief in some cases we find them as the property of probability and in some cases, they are not mathematical. Using this theory we can make answers to the questions that have been generated using the probability theory. This theory mainly consists of two fundamentals: Degree of belief and plausibility. We can understand this theory using some examples. Let’s say we have a person diagnosed with covid-19 symptoms and have a belief of 0.5 for a proposition that the person is suffering from covid-19. This will mean that we have evidence that makes us think strongly that the person is suffering from covid(a proposition is true) with a confidence of 0.5. However, there is a contradiction that a person is not suffering from covid with a confidence of 0.2.


The Other AI: Augmented Intelligence

With a clear view of the benefits augmented intelligence delivered by AR can provide, you may be excited to get started within your enterprise but unsure of where to begin. First, it's important to start by speaking to your field technicians and service agents to gauge their interest or any potential aversion to implementing the technology into their workspace. New technology can be intimidating to field service technicians who are used to completing tasks a certain way. Helping them to understand how the technology can enhance their jobs and make service experiences less challenging and more engaging will be key. Next, consider which devices are needed to implement the augmented intelligence platform. At a basic level, a smartphone or tablet is needed. Hands-free wearable glasses make it easier for technicians to accomplish tasks in the field and on the factory floor. Drone support goes even further with AR visual awareness and graphical guidance not previously available. Finally, you'll want to confirm the bandwidth and connectivity requirements of the augmented intelligence AR platform and associated devices to ensure your field service technicians are set up for success.


Writing Code Is One Thing, Learning to Be a Software Engineer Is Another

Software developers are always students of software development and whenever you think you know what you are doing, it will punch you in the face. Good developers are humble because software development crushes overconfidence with embarrassing mistakes. You cannot avoid mistakes, problems and disasters. Therefore, you need to be humble to acknowledge mistakes and need a team to help you find and fix them. When you start as a developer, you focus on creating code to meet the requirements. I used to think being a developer was just writing code. Software development has many other aspects to it, from design, architecture, unit testing to DevOps and ALM. Gathering requirements and clarifying assumption. There are many best practices such as SOLID principles, DRY (Don’t repeat yourself), KISS, and others. The best practices and fundamental skills have long-term benefits. This makes them hard for junior developers to understand because there is no initial benefit. Well-named code, designed to be easily tested, isn’t first draft code. It does more than work. It’s built to be easy to read, understand and change.


AI Set to Disrupt Traditional Data Management Practices

“They often don’t have the skill sets, or their organizations don’t put in place processes and tools and practices to really manage data management for AI specifically,” says Sallam. “So data-centric AI has the potential to disrupt what has been traditional data management practices as well as prevalent model-centric data science by making sure that AI-specific considerations like data bias, labeling, drift, are all in place in a consistent manner to improve the quality of models on an ongoing basis.” Are tools under development to address this need, or are organizations investing in solutions for it? Sallam says that some of the other trends on the list will contribute to improving data management around AI. Specifically, to address this gap, leading organizations are disrupting data management for AI by building out data fabrics on active metadata and investing in things like AI governance, she said. This data-centric AI trend is one of several Gartner highlighted in its report for 2022 and grouped with a few others under the title of activating dynamism and diversity. 


Growing prospects for edge datacentres

Edge operations require user organisations and suppliers to think beyond infrastructure and architectural needs. New automation and orchestration challenges will arise, often across transactional boundaries and occurring between different companies and industries, rather than just different parts of the network. They must also think about ownership of the software and infrastructure stack and the likely path of service engagement – be that through a telecoms operator, hyperscale public cloud provider or others. Providers of edge operational services also need to decide how they support multiple customers according to their individual needs. This will be especially necessary for applying operational-specific AI algorithms, and may result in multi-layered partner offerings. All this will require organisations to think more carefully about how they extend their datacentre operations to enable greater levels of edge processing, work with cloud providers or hook into another provider’s edge datacentre network. The biggest drivers for edge datacentres are coming from industry sectors where edge operations are already well established. 


The Metaverse needs to keep an eye on privacy to avoid Meta’s mistakes

Metaverse avatars are a conglomeration of all issues relating to privacy in the digital realm. As a user’s gateway to all Metaverse interactions, they can also offer platforms a lot of personal data to collect, especially if their tech stack involves biometric data, like tracking users’ facial features and expressions for the avatar’s own emotes. The risk of someone hacking biometric data is far scarier than hacking shopping preferences. Biometrics are often used as an extra security precaution, such as when you authorize payment on your phone using your fingerprint. Imagine someone stealing your fingerprints and draining your card with a bunch of transfers. Such breaches are not unheard of: In 2019, hackers got their hands on the biometric data of 28 million people. It’s scary to think about how traditional digital marketing might look in the Metaverse. Have you ever shopped for shoes online and then suddenly noticed your Facebook is filled with ads for similar footwear? That’s a result of advertisers using both cookies and your IP address to personalize your ads. 


The Most In-Demand Cyber Skill for 2022

Just when everybody hoped that the security environment could not be more challenging, recent world events have created a further substantial uptick in cyber-attacks. This has also increased the sense that maybe we should all care more about the security of everything we ever purchased and placed in the cloud. Not so much buyer’s remorse as a penitent desire to security upcycle anything in the cloud that might be more critical to the organization once the current threat landscape is taken into consideration. Zero trust, extended detection and threat response (XDR), SASE (secure access service edge) – almost all the hottest topics are about how to take the security standards that were (once-upon-a-time) applied as standard to traditional networks and *seamlessly* implement them across cloudenvironments. The number one position for cloud computing makes sense. It reflects the growing concern about cloud security and the gradual evolution of the requirement to ensure that each organization has a consistent security architecture that extends over and includes any important cloud solutions and services in use.


How SaaS Models Changed Content Creation

Content creation used to be a difficult, arduous and manual process. Creative visions were consistently hampered by workflow and technological limitations. The dilemmas of our past were based on technical feasibility. Now, those restrictions have become completely unshackled. It’s no longer a question of what’s possible to do, but rather what you want to do, and which path do you take to get there? SaaS evened the playing field. To understand what is possible now and what is yet to come, it’s important to distinguish between two areas within the umbrella-use of SaaS. First, we have true software as a service, which is software that runs on the internet and is accessed in the cloud. Google Workspace is an example of this, allowing users to create spreadsheets, documents and presentations that are stored on Google’s servers. (Disclosure: My company has a partnership with Google.) The software runs as a service for you to connect to from any device and edit your documents anywhere. It's persistent regardless of the computer you’re on, and documents can be edited by multiple users, even simultaneously.


Data parallelism vs. model parallelism – How do they differ in distributed training?

In data parallelism, the dataset is split into ‘N’ parts, where ‘N’ is the number of GPUs. These parts are then assigned to parallel computational machines. Post that, gradients are calculated for each copy of the model, after which all the models exchange the gradients. In the end, the values of these gradients are averaged. For every GPU or node, the same parameters are used for the forward propagation. A small batch of data is sent to every node, and the gradient is computed normally and sent back to the main node. There are two strategies using which distributed training is practised called synchronous and asynchronous. ... In model parallelism, every model is partitioned into ‘N’ parts, just like data parallelism, where ‘N’ is the number of GPUs. Each model is then placed on an individual GPU. The batch of GPUs is then calculated sequentially in this manner, starting with GPU#0, GPU#1 and continuing until GPU#N. This is forward propagation. Backward propagation on the other end begins with the reverse, GPU#N and ends at GPU#0. Model parallelism has some obvious benefits. It can be used to train a model such that it does not fit into just a single GPU.



Quote for the day:

"At the heart of great leadership is a curious mind, heart, and spirit." -- Chip Conley

No comments:

Post a Comment