Daily Tech Digest - December 20, 2024

The Top 25 Security Predictions for 2025

“Malicious actors will go full throttle in mining the potential of AI in making cyber crime easier, faster and deadlier. But this emerging and ever-evolving technology can also be made to work for enterprise security and protection by harnessing it for threat intelligence, asset profile management, attack path prediction and remediation guidance. As SOCs catch up to secure innovations still and yet unraveling, protecting enterprises from tried and tested modes of attack remains essential. While innovation makes for novel ways to strike, criminals will still utilize what is easy and what has worked for them for years.” ... Organizations are urged to embrace scalable, cloud-native security information and event management (SIEM) solutions. These tools improve threat detection and response by integrating logs from cloud and endpoint systems and automating incident management with security orchestration, automation, and response (SOAR) features. ... While targets like edge devices will continue to capture the attention of threat actors, there’s another part of the attack surface that defenders must pay close attention to over the next few years: their cloud environments. Although cloud isn’t new, it’s increasingly piquing the interest of cyber criminals. 


Why AI language models choke on too much text

Although RNNs have fallen out of favor since the invention of the transformer, people have continued trying to develop RNNs suitable for training on modern GPUs. In April, Google announced a new model called Infini-attention. It’s kind of a hybrid between a transformer and an RNN. Infini-attention handles recent tokens like a normal transformer, remembering them and recalling them using an attention mechanism. However, Infini-attention doesn’t try to remember every token in a model’s context. Instead, it stores older tokens in a “compressive memory” that works something like the hidden state of an RNN. This data structure can perfectly store and recall a few tokens, but as the number of tokens grows, its recall becomes lossier. ... Transformers are good at information recall because they “remember” every token of their context—this is also why they become less efficient as the context grows. In contrast, Mamba tries to compress the context into a fixed-size state, which necessarily means discarding some information from long contexts. The Nvidia team found they got the best performance from a hybrid architecture that interleaved 24 Mamba layers with four attention layers. This worked better than either a pure transformer model or a pure Mamba model.


The End of ‘Apps,’ Brought to You by AI?

Achieving the dream of a unified customer experience is possible, not by building a bigger app but by AI super agents. Much of the groundwork has already been done: AI language models like Claude and GPT-4 are already designed to support many use cases, and Agentic AI takes that concept further. OpenAI, Google, Amazon, and Meta are all making general-purpose agents that can be used by anyone for any purpose. In theory, we might eventually see a vast network of specialized AI agents running in integration with each other. These could even serve customers’ needs within the familiar interfaces they already use. Crucially, personalization is the big selling point. It’s the reason AI super agents may succeed where super apps failed in the West. A super agent wouldn’t just aggregate services or fetch a gadget’s price when prompted. It would compare prices across frequented platforms, apply discounts, or suggest competing gadgets based on reviews you’ve left for previous models. ... This new ‘super agents’ reality would yield significant benefits for developers, too, possibly even redefining what it means to be a developer. While lots of startups invent good ideas daily, the reality of the software business is that you’re always limited by the number of developers available. 


A Starter’s Framework for an Automation Center of Excellence

An automation CoE is focused on breaking down enterprise silos and promoting automation as a strategic investment imperative for achieving long-term value. It helps to ensure that when teams want to create new initiatives, they don’t duplicate previous efforts. There are various cost, efficiency and agility benefits to setting up such an entity in the enterprise. ... Focus on projects that deliver maximum impact with minimal effort. Use a clear, repeatable process to assess ROI — think about time saved, revenue gained and risks reduced versus the effort and complexity required. A simple question to ask is, “Is this process ready for automation, and do we have the right tools to make it work?” ... Your CoE needs a solid foundation. Select tools and systems that integrate seamlessly with your organization’s architecture. It might seem challenging at first, but the long-term cultural and technical benefits are worth it. Ensure your technology supports scalability as automation efforts grow. ... Standardize automation without stifling team autonomy. Striking this balance is key. Consider appointing both a business leader and a technical evangelist to champion the initiative and drive adoption across the organization. Clear ownership and guidelines will keep teams aligned while fostering innovation.


What is data architecture? A framework to manage data

The goal of data architecture is to translate business needs into data and system requirements, and to manage data and its flow through the enterprise. Many organizations today are looking to modernize their data architecture as a foundation to fully leverage AI and enable digital transformation. Consulting firm McKinsey Digital notes that many organizations fall short of their digital and AI transformation goals due to process complexity rather than technical complexity. ... While both data architecture and data modeling seek to bridge the gap between business goals and technology, data architecture is about the macro view that seeks to understand and support the relationships between an organization’s functions, technology, and data types. Data modeling takes a more focused view of specific systems or business cases. ... Modern data architectures must be scalable to handle growing data volumes without compromising performance. A scalable data architecture should be able to scale up and to scale out. ... ... Modern data architectures must ensure data remains accurate, consistent, and unaltered through its lifecycle to preserve its reliability for analysis and decision-making. They must prevent issues like data corruption, duplication, or loss.


Cybersecurity At the Crossroads: The Role Of Private Companies In Safeguarding U.S. Critical Infrastructure

Regulation alone is not a solution, but it does establish baseline security standards and provide much-needed funding to support defenses. Standards have come a long way and are relatively mature. Though there is still a tremendous amount of gray area, and a lack of relevance or attainability for certain industries and smaller organizations. The federal government must prioritize injecting funds into cybersecurity initiatives, ensuring that even the smallest entities managing critical infrastructure can implement strong security measures. With this funding, we must build a strong defense posture and cyber resiliency within these private sector organizations. This involves more than deploying advanced tools; it requires developing skilled personnel capable of responding to incidents and defending against attacks. Upskilling programs should focus on blue teaming and incident response, ensuring that organizations have the expertise to manage their security proactively.A critical component of effective cybersecurity is understanding and applying the standard risk formula: Risk = Threat x Vulnerability x Consequence. This formula emphasizes that risk is determined by evaluating the likelihood of an attack (Threat), the weaknesses in defenses (Vulnerability), and the potential impact of a breach (Consequence). 


Achieving Network TCO

TCO discussion should shift from a unilateral cost justification (and payback) of technology that is being proposed to a discussion of what the opportunity costs for the business will be if a network infrastructure investment is canceled or delayed. If a company determines strategically to decentralize manufacturing and distribution but is also wary of adding headcount, it's going to seek out edge computing and network automation. It’s also likely to want robust security at its remote sites, which means investments in zero-trust networks and observability software that can assure that the same level of enterprise security is being applied at remote sites as it is at central headquarters. In cases like this, it shouldn’t be the network manager or even the CIO who is solely responsible for making the budget case for network investments. Instead, the network technology investments should be packaged together in the total remote business recommendation and investment that other C-level executives argue for with the CIO and/or network manager, HR, and others. In this scenario, the TCO of a network technology investment is weighed against the cost of not doing it at all and missing a corporate opportunity to decentralize operations, which can’t be accomplished without the technology that is needed to run it.


The coming hardware revolution: How to address AI’s insatiable demands

The US forecast for energy consumption on AI is alarming. Today’s AI queries require roughly 10x the electricity of traditional Google queries - a ChatGPT request runs 10x watt-hours versus a Google request. A typical CPU in a data center uses approximately 300 watts per hour (Electric Power Research Institute), while a Nvidia H100 GPU uses up to 700 watts per hour, a similar usage of an average household in the US per month. Advancements in AI model capabilities, and greater use of parameters, continue to drive energy consumption higher. Much of this demand is centralized in data centers as companies like Amazon, Microsoft, Google, and Meta build more and more massive hyperscale facilities all over the country. US data center electricity consumption is projected to grow 125 percent by 2030, using nine percent of all national electricity. ... While big tech companies certainly have the benefit of incumbency and funding advantage, the startup ecosystem will play an absolutely crucial role in driving the innovation necessary to enable the future of AI. Large public tech companies often have difficulty innovating at the same speed as smaller, more nimble startups.


Agents are the 'third wave' of the AI revolution

"Agentic AI will be the next wave of unlocked value at scale," Sesh Iyer, managing director and senior partner with BCG X, Boston Consulting Group's tech build and design unit, told ZDNET. ... As with both analytical and gen AI, AI agents need to be built with and run along clear ethical and operational guidelines. This includes testing to minimize errors and a governance structure. As is the case with all AI instances, due diligence to ensure compliance and fairness is also a necessity for agents, Iyer said. As is also the case with broader AI, the right skills are needed to design, build and manage AI agents, he continued. Such talent is likely already available within many organizations, with the domain knowledge needed, he added. "Upskill your workforce to manage and use agentic AI effectively. Developing internal expertise will be key to capturing long-term value from these systems." ... To prepare for the shift from gen AI to agentic AI, "start small and scale strategically," he advises. "Identify a few high-impact use cases -- such as customer service -- and run pilot programs to test and refine agent capabilities. Alongside these use cases, understand the emerging platforms and software components that offer support for agentic AI."


Having it both ways – bringing the cloud to on-premises data storage

“StaaS is an increasingly popular choice for organisations, with demand only likely to grow soon. The simple reason for this is two-fold: it provides both convenience and simplicity,” said Anthony Cusimano, Director of Technical Marketing at Object First, a supplier of immutable backup storage appliances. There is more than one flavour of on-premises StaaS, as was pointed out by A3 Communications panel member Camberley Bates, Chief Technology Advisor at IT research and advisory firm The Futurum Group. Bates pointed out that the two general categories of on-premises StaaS service are Managed and Non-Managed StaaS. Managed StaaS sees vendors handling the whole storage stack, by both implementing and then fully managing storage systems on customers’ premises. However, Bates said enterprises are more attracted to Non-Managed StaaS. ... “Non-managed StaaS has become surprisingly of interest in the market. This is because enterprises buy it ‘once’ and do not have to go back for a capex request over and over again. Rather, it becomes a monthly bill that they can true-up over time. We have found the fully managed offering of less interest, with enterprises opting to use their own resources to handle the storage management,” continued Bates.



Quote for the day:

“If you don’t try at anything, you can’t fail… it takes back bone to lead the life you want” -- Richard Yates

No comments:

Post a Comment