Daily Tech Digest - March 19, 2024

Is The Public Losing Trust In AI?

Of course, the simplest way to look at this challenge is that in order for people to trust AI, it has to be trustworthy. This means it has to be implemented ethically, with consideration of how it will affect our lives and society. Just as important as being trustworthy is being seen to be trustworthy. This is why the principle of transparent AI is so important. Transparent AI means building tools, processes, and algorithms that are understandable to non-experts. If we are going to trust algorithms to make decisions that could affect our lives, we must, at the very least, be able to explain why they are making these decisions. What factors are being taken into account? And what are their priorities? If AI needs the public's trust, then the public needs to be involved in this aspect of AI governance. This means actively seeking their input and feedback on how AI is used. Ideally, this needs to happen at both a democratic level, via elected representatives, and at a grassroots level. Last but definitely not least, AI also has to be secure. This is why we have recently seen a drive towards private AI – AI that isn't hosted and processed on huge public data servers like those used by ChatGPT or Google Gemini.

Reliable Distributed Storage for Bare-Metal CAPI Cluster

By default, most CAPI solutions will use the “Expand First” (or “RollingUpdateScaleOut” in CAPI terms) repave logic. This logic will install an additional fresh new server and add it to the cluster first, before then removing an old server. While this is useful to ensure the cluster never has less total compute capacity than before you started the repave operation, it is problematic for distributed storage clusters because you are introducing a new node without any data to the cluster, while taking away a node that does contain data. So instead, we want to use the “Contract First” repave logic for the pool of storage nodes. That way, we can remove a storage node first, then reinstall it and add it back to the cluster, thereby immediately restoring data redundancy. ... So, if a different issue causes the distributed storage software to not install properly on the new node, you can still run into trouble. For example, Portworx supports specific kernel versions, and installing new nodes with a kernel version it doesn’t support can prevent the installation from succeeding. For that reason, it’s a good idea to lock the kernel version that MaaS deploys. Reach out to us if you want to learn how to achieve that.

Evaluating databases for sensor data

The primary determinant in choosing a database is understanding how an application’s data will be accessed and utilized. A good place to begin is by classifying workloads as online analytical processing (OLAP) or online transaction processing (OLTP). OLTP workloads, traditionally handled by relational databases, involve processing large numbers of transactions by large numbers of concurrent users. OLAP workloads are focused on analytics and have distinct access patterns compared to OLTP workloads. In addition, whereas OLTP databases work with rows, OLAP queries often involve selective column access for calculations. ... Another consideration when selecting a database is the internal team’s existing expertise. Evaluate whether the benefits of adopting a specialized database justify investing in educating and training the team and whether potential productivity losses will appear during the learning phase. If performance optimization isn’t critical, using the database your team is most familiar with may suffice. However, for performance-critical applications, embracing a new database may be worthwhile despite initial challenges and hiccups.

Surviving the “quantum apocalypse” with fully homomorphic encryption

There are currently two distinct approaches to face an impending “quantum apocalypse”. The first uses the physics of quantum mechanics itself and is called Quantum Key Distribution (QKD). However, QKD only really solves the problem of key distribution, and it requires dedicated quantum connections between the parties. As such, it is not scalable to solve the problems of internet security; instead, it is most suited to private connections between two fixed government buildings. It is impossible to build internet-scale, end-to-end encrypted systems using QKD. The second solution is to utilize classical cryptography but base it on mathematical problems for which we do not believe a quantum computer gives any advantage: this is the area of post-quantum cryptography (PQC). PQC algorithms are designed to be essentially drop-in replacements for existing algorithms, which would not require many changes in infrastructure or computing capabilities. NIST has recently announced standards for public key encryption and signatures which are post-quantum secure. These new standards are based on different mathematical problems

Teams, Slack, and GitHub, oh my! – How collaborative tools can create a security nightmare

Fast and efficient collaboration is essential to today’s business, but the platforms we use to communicate with colleagues, vendors, clients, and customers can also introduce serious risks. Looking at some of the most common collaboration tools — Microsoft Teams, GitHub, Slack, and OAuth — it’s clear there are dangers presented by information sharing, as valuable as that is to business strategy. Any of these, if not safeguarded or used inappropriately, can be a tool for attackers to gain access to your network. The best protection is to ensure you are aware of these risks and apply the appropriate modifications and policies to your organization to help prevent attackers from gaining a foothold in your organization — that also means acknowledging and understanding the threats of insider risk and data extraction. Attackers often know your network better than you do. Chances are, they also know your data-sharing platforms and are targeting those as well. Something as simple as improper password sharing can allow a bad actor to phish their way into a company’s network and collaboration tools can present a golden opportunity.

Improving computational performance of AI requires upskilling of professionals in Embedded/VLSI area

Implementing AI systems or applications requires intensive computational processors and low-cost power to deploy algorithms. Here, Very Large Scale Integration (VLSI) and embedded system design play a critical role. VLSI design involves the creation and miniaturisation of complex circuits, such as processors, memory circuits, and more recently, customized hardware for AI applications. On the other hand, embedded systems are computing systems for dedicated or specific functionalities, such as smart agriculture or industrial automation. The integration of VLSI with AI has the potential to revolutionise various sectors by enabling faster, more power-efficient, and customised hardware for AI applications. ... AI-based solutions are applied in designing and deploying communication systems to significantly enhance network performance and thereby the overall user experience. Dynamic allocation of resources, such as power and bandwidth, can be done efficiently by AI algorithms, which leads to improved spectral efficiency, reduced interference, and power consumption. Intelligent beam forming using AI algorithms enables wireless systems to focus their power and frequency band for specific users or devices.

Microsoft announces collaboration with NVIDIA to accelerate healthcare and life sciences innovation with advanced cloud, AI and accelerated computing capabilities

Microsoft, NVIDIA and SOPHiA GENETICS are collaborating to leverage combined expertise in technology and genomics to develop a streamlined, scalable and comprehensive whole-genome analytical solution. As part of this collaboration, the SOPHiA DDM Software-as-a-Service platform, hosted on Azure, will be powered by NVIDIA Parabricks for SOPHiA DDM’s whole genome application. Parabricks is a scalable genomics analysis software suite that leverages full-stack accelerated computing to process whole genomes in minutes. Compatible with all leading sequencing instruments, Parabricks supports diverse bioinformatics workflows and integrates AI for accuracy and customization. ... Microsoft aims to propel healthcare and life sciences into an exciting new era of medicine, helping unlock transformative possibilities for patients worldwide. The combination of the global scale, security and advanced computing capabilities of Microsoft Azure with NVIDIA DGX Cloud and the NVIDIA Clara suite is set to accelerate advances in clinical research, drug discovery and care delivery.

How Deloitte navigates ethics in the AI-driven workforce: Involve everyone

The approach to developing an ethical framework for AI development and application will be unique for each organization. They will need to determine their use cases for AI as well as the specific guardrails, policies, and practices needed to make sure that they achieve their desired outcome while also safeguarding trust and privacy. Establishing these ethical guidelines -- and understanding the risks of operating without them -- can be very complex. The process requires knowledge and expertise across a wide range of disciplines. ... On a broader level, publishing clear ethics policies and guidelines, and providing workshops and trainings on AI ethics, were ranked in our survey as some of the most effective ways to communicate AI ethics to the workforce, and thereby ensure that AI projects are conducted with ethics in mind. ... Leadership plays a crucial role in underscoring the importance of AI ethics, determining the resources and experience needed to establish the ethics policies for an organization, and ensuring that these principles are rolled out. This was one reason we explored the topic of AI ethics from the C-suite perspective. 

How to stop data from driving government mad

This would be a start, but everybody in large organisations knows that top-down initiatives from the centre rarely work well at the coalface. If the JAAC is to be effective at converting data into information, what insight could it glean from structures that have evolved to do this? And what could it learn from scientific fields that manage this successfully? First, deep neural networks learn by repeatedly passing information back and forth until every neurone is tuned to achieve the same objective. Information flow in both directions is the key. Neil Lawrence, DeepMind professor of machine learning at the University of Cambridge, notes that in government, "People at the coal face have a better understanding of the right interventions, although not what the central policy might be; a successful centre will have a co-ordinating function driven by an AI strategy, but will devolve power to the departments, professions, and regulators to implement it." Or, as Jess Montgomery, director of AI@Cam says: "Getting government data - and AI - ready will require foundational work, for example in data curation and pipeline building." 

Continuous Improvement as a Team

Conducting regular Retrospectives enables teams to pause and reflect on their past actions, practices, and workflows, pinpointing both strengths and areas for improvement. This continuous feedback loop is critical for adapting processes, enhancing team dynamics, and ensuring the team remains agile and responsive to change. Guarantee the consistency of your Retrospectives at every Sprint's conclusion. Before these sessions, collaboratively plan an agenda that promotes openness and inclusivity. Facilitators should incorporate practices such as anonymous feedback mechanisms and engaging games to ensure honest and constructive discussions, setting the stage for meaningful progress and team development. ... Effective stakeholder collaboration ensures the team’s efforts align with the broader business goals and customer needs. Engaging stakeholders throughout the development process invites diverse perspectives and feedback, which can highlight unforeseen areas for improvement and ensure that the product development is on the right track. Engage your stakeholders as a team, starting with the Sprint Reviews. 

Quote for the day:

“There's a lot of difference between listening and hearing.” -- G. K. Chesterton

No comments:

Post a Comment