March 15, 2024

AI hallucination mitigation: two brains are better than one

LLMs have been characterized as stochastic parrots — as they get larger, they become more random in their conjectural or random answers. These “next-word prediction engines” continue parroting what they’ve been taught, but without a logic framework. One method of reducing hallucinations and other genAI-related errors is Retrieval Augmented Generation or “RAG” — a method of creating a more customized genAI model that enables more accurate and specific responses to queries. But RAG doesn’t clean up the genAI mess because there are still no logical rules for its reasoning. In other words, genAI’s natural language processing has no transparent rules of inference for reliable conclusions (outputs). What’s needed, some argue, is a “formal language” or a sequence of statements — rules or guardrails — to ensure reliable conclusions at each step of the way toward the final answer genAI provides. Natural language processing, absent a formal system for precise semantics, produces meanings that are subjective and lack a solid foundation. But with monitoring and evaluation, genAI can produce vastly more accurate responses.

The Courtroom Factor in GenAI’s Future

There are a lot of moving parts. You kind of hit that on the head. Certainly, every day there’s something new, some development, but let me focus on my area of expertise, which is litigation and where I see some of the domestic generative AI litigation perhaps trending or where I think we’re going to see an increase in litigation going forward. I think that’s going to be twofold. I think you’re going to continue to see the intellectual property issues attended to generative AI litigated. I think that’s one area that’s inevitable. I think the other area that we’re really going to start to see, and we already are seeing an uptick in litigation, is in the use and deployment of generative AI by companies. Let me frame it this way. As companies attempt to take advantage of the promise of generative AI, they’re going to, they already have, and they will continue to deploy generative AI tools, and generative AI system, more advanced systems in terms of machine learning, and generative aspects of AI in their businesses. I think we’ll see a steady increase in use -- and some folks would say misuse -- of AI. It’s trickling out where plaintiffs allege that the business or the entity has done something wrong using AI.?

Next-Gen DevOps: Integrate AI for Enhanced Workflow Automation

In DevOps, the ability to anticipate and prevent outages can mean the difference between success and catastrophic failure. In such situations, AI-powered predictive analytics can empower teams to stay one step ahead of potential disruptions. Predictive analytics uses advanced algorithms and machine learning models to analyze vast amounts of data from various sources, such as application logs, system metrics, and historical incident reports. It then identifies patterns, correlations, and detects anomalies within this data to provide early warnings of impending system failures or performance degradation. This enables teams to take proactive measures before issues escalate into full-blown outages. ... Doing things by hand introduces the possibility of human error and is way too time-intensive — so it comes as no surprise that the industry is turning toward automation. Tools that utilize artificial intelligence can identify potential issues by analyzing code repositories at speeds that cannot be replicated by humans. On the ground level, this means that various potential issues — bottlenecks in terms of performance, code that doesn’t meet best practices or internal standards, security liabilities and code smells — can be identified quickly and at scale.

Key MITRE ATT&CK techniques used by cyber attackers

Half of the top threats are ransomware precursors that could lead to a ransomware infection if left unchecked, with ransomware continuing to have a major impact on businesses. Despite a wave of new software vulnerabilities, humans remained the primary vulnerability that adversaries took advantage of in 2023, comprising identities to access cloud service APIs, execute payroll fraud with email forwarding rules, launch ransomware attacks, and more. As organizations migrate to the cloud and rely on a growing array of SaaS applications to manage and access sensitive information, identities are the ties that bind all these systems together. Adversaries have quickly learned that these systems house the information they want and that valid and authorized identities are the most expedient and reliable way into those systems. Researchers noted several broader trends impacting the threat landscape, such as the emergence of generative AI, the continued prominence of remote monitoring and management (RMM) tool abuse, the prevalence of web-based payload delivery like SEO poisoning and malvertising, the increasing necessity of MFA evasion techniques, and the dominance of brazen but highly effective social engineering schemes such as help desk phishing.

Data management trends: GenAI, governance and lakehouse

Nearly every major database and data platform vendor had some form of generative AI news in 2023. Some vendors included generative AI as a tool to act as an assistant, helping users to conduct different tasks. Managing data platforms and writing different types of data queries has long been a complicated exercise and generative AI simplifies it. Among the many vendors that integrated some form of AI assistant, Dremio launched its Text-to-SQL AI-powered tool in June, which enables users to generate SQL queries more easily. In August, Couchbase announced Capella iQ, a generative AI tool that helps developers write database application code. Also in August, SnapLogic rolled out its SnapGPT AI tool to help users build data pipelines using natural language. ... Whether it's for AI, data operations or analytics, the topic of data governance is increasingly important. Being able to understand where data comes from, how to make it available and use it is important for security, privacy, accuracy and reliability. Over the course of 2023, multiple vendors expanded and enhanced data governance capabilities to help manage data.

The importance of "always-ready" data

Imagine living in a world where data is prepared on an ongoing basis – that is, data prepared so quickly, regardless of the amount, that it is always ready. Such a reality would enable enterprises to respond promptly to evolving business needs and unexpected challenges. Moreover, it would minimize backlogs of tickets and requests, granting data engineers time to be more proactive and productive. One way to facilitate this is through the use of a cloud data lakehouse. With it, data can be prepared directly on cloud storage, without the long load times that ETL- or ELT-based (extract, load, and transform) data processing typically takes. For enterprises that manage complicated and data-heavy workloads, the result is game-changing on multiple fronts. Agile data infrastructure underscored by superior cost performance will give enterprises an efficient means of adapting to changing market dynamics, new projects, and fluctuating customer demands. Beyond the flexibility it grants data engineers, always-ready data also empowers them to conduct ad-hoc queries and analytics as a way to derive actionable insights and predictions on the fly.?

Read more here ...