About CatTruxillo

Salrios0

Some banks in Mexico offer finance educations to their clients. They provide insights and best practices to use credit and savings services. This approach helps their clients to trust the process to request a credit card or a loan from their banker.

CatTruxillo

Two distinct design philosophies currently define how autonomous intelligent systems engage with the world: AI agents and agentic AI. The terms are similar, but they represent technologies with different attributes. The purpose of this introductory post is to distinguish between AI agents and agentic AI, introduce use cases from business and academia, and explore just a few of the ethical and practical consequences of implementing agentic systems. At the end I'll leave you with lots of resources that I am finding helpful in navigating this topic. This post assumes that you have at least some background in statistics or data, and have wandered into the realm of AI either out of curiosity or out of business need. But this post makes no reference to math or statistics. Anyone should be able to join this conversation, and I hope you will, in the comments below. An AI Agent executes specific tasks in response to inputs or instructions. AI agents tend to be reactive, predictable, and task-bound. Agentic AI operates with a degree of autonomy and self-direction in ways that are proactive, adaptive, and goal-forming. Agentic AI can operate without human supervision to accomplish multi-step tasks, adapting its approach based on changes in conditions. Let’s look at three examples: Julia is a marketing coordinator who uses a voice assistant to take notes and book meetings. The assistant responds to her input, but it does not initiate tasks or adjust behavior beyond the predefined logic. Julia’s voice assistant is typical of an AI agent. It is useful, but its scope is limited. If her system were to begin suggesting rest breaks based on her calendar and the tone of her emails, it would be entering agentic territory. Agentic AI systems may proactively rearrange tasks, recognize behavioral patterns, and provide context-sensitive nudges. This increased autonomy would require clearer disclosures, especially when systems derive insights from personal signals. It's important to consider any privacy issues when systems infer meaning from sensitive data. Providers of agentic systems have an obligation to develop and evolve safeguards in parallel. Raj is a director of logistics, supervising a fleet of delivery drivers. He relies on an AI system that monitors fuel prices, weather patterns, and forecast delivery volume over time. The system can autonomously reroute shipments, reallocate warehouse inventory, and update stakeholders before Raj even flags an issue. Raj’s AI system does not wait for instructions. It acts on its own insights, reshaping operational strategy in real time. This agentic AI system illustrates a shift from automation tools to intelligent partners. Agentic AI can benefit optimizing operations by continuously reassessing business conditions and refining processes without explicit prompting. These systems might synthesize multi-source data to recommend interventions at scale. Note the difference in customer engagement between an AI agent (Julia) and agentic AI (Raj). AI agents offer support, while agentic AI adjusts tone, prioritizes escalations, and adapts workflows dynamically based on sentiment data that the system can access. Blanca is a professor who uses a suite of AI tools to identify contradictions across research studies and to propose new experiments for testing theory. She also uses it to update her classroom lectures based on student performance in prior semesters. Blanca's AI assistant navigates through databases and devises research ideas. It flags contradictions in literature and suggests new hypotheses. The system functions as an intellectual collaborator by identifying unexplored variables and recommending sources. This is an example of agentic AI. By adapting instructional content to cohort performance trends, Blanca’s system also enables better instructional support by diagnosing learning gaps and tailoring the content to be delivered to students. This application of agentic AI could be useful, for example, to counter the educational losses that a cohort of students experienced during the COVID-19 pandemic, tailoring educational assets to diagnose and fill in the gaps. Unique ethical issues arise from the use of agentic AI in different settings. For example, in Blanca's academic use case, what are the implications of agentic AI on authorship transparency and academic integrity? A few years ago, universities scrambled to publish generative AI policies and those policies continue to evolve in response to how students and faculty use AI, and as the tools available to them become more sophisticated. The time for agentic AI ethical guardrails is here, but there is still a great deal to learn about the downstream impact. Here are some questions I’ve been thinking about regarding ethics and governance for agentic AI. Agentic systems challenge traditional supervision of models. What level of oversight is appropriate? What thresholds should trigger human intervention? Continuous learning of AI systems carries a risk of carrying forward historical biases into future decisions. Ongoing auditing and explainability protocols are needed to assess potential bias. Who is accountable for implementing these protocols? Agentic AI can make impactful decisions. How do we attribute responsibility, for better or for worse? Public policy worldwide is evolving. How do we determine liability, especially when outcomes involve public trust or personal data? As AI becomes more a part of everyday life, the consequences extend beyond technical issues and into the realm of the personal and societal. We want to design AI systems that reflect our values and goals, and that serve to improve the world. I'd like to ponder these questions over the next few months and share those thoughts with you here-- let me know what you think about these issues, too. I would like to hear from you. What kinds of AI applications have you been dreaming of, or even designing? What ethics-related topic in analytics should I take on next? Please leave suggestions in the comments! There's a strong appetite for information about these topics right now. If you’re interested in reading more about agentic AI, please check out these posts: The hard part of agentic AI: Designing for trust - SAS Voices Building trust: A principled approach to ethical AI agent development - SAS Voices From Chat to Decision: A Blueprint for Intelligent AI Assistants SAS Agentic AI Accelerator – Register and Publish Models From Chat to Decision: A Blueprint for Intelligent AI Assistants Agentic AI: Powering the Next Evolution in Data and AI Lifecycle Want to learn to learn to do this in SAS? Check out this course: Agentic AI - How to with SAS® Viya®, which is part of the SAS Decisioning Learning Subscription. Want to chew on ethical dilemmas and learn about responsible and trustworthy AI? Sign up for this free e-learning course: Responsible Innovation and Trustworthy AI, which is part of the AI Literacy for Everyone learning path. Additional references that I found valuable in writing this post: Russell, S., & Norvig, P. (2020). Artificial Intelligence: A Modern Approach (4th ed.). Pearson. Floridi, L., & Cowls, J. (2019). A Unified Framework of Five Principles for AI in Society. Harvard Data Science Review, 1(1). https://hdsr.mitpress.mit.edu/pub/l0jsh9d1/release/8 OpenAI. (2023). GPT-4 Technical Report. https://openai.com/research/gpt-4 Bubeck, S. et al. (2023). Sparks of Artificial General Intelligence: Early Experiments with GPT-4. arXiv preprint. https://arxiv.org/abs/2303.12712 Whittlestone, J., et al. (2019). The Role and Limits of Principles in AI Ethics: Towards a Focus on Tensions. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society. Find more articles from SAS Global Enablement and Learning here.

mareezeshak · ‎08-23-2025

I completely agree. Mobile security encryption is essential to protect personal data, and at the same time, ensuring that it can help detect and prevent illegal or harmful activities is very important.

jomana-khatib · ‎08-01-2025

An example of peace of mind from AI transparency could be Google Maps or Waze showing real-time traffic data, estimated arrival times, and alternative routes. The transparency in how delays are calculated and why a route is suggested gives users confidence in the system’s advice and allows them to make informed choices—much like how EV drivers feel reassured by clear, real-time battery and range predictions.

jomana-khatib · ‎08-01-2025

Law schools have raised concerns about the lack of transparency in popular rankings, especially the heavy focus on graduate starting salaries, which may discourage public sector careers. To help prospective students make informed decisions, ranking systems should disclose key metrics such as: Graduation and bar passage rates Employment outcomes in both private and public sectors Student debt levels and repayment success Faculty qualifications and student-faculty ratios Curriculum quality and experiential learning opportunities Clear disclosure of these metrics would create a more balanced and meaningful evaluation of law programs.

jomana-khatib · ‎08-01-2025

1. One way that a changing environment can impact the accuracy of machine learning predictions: A changing environment can cause data drift, where the patterns in the data used to train the model no longer match real-world data. This leads to reduced prediction accuracy because the model’s assumptions become outdated. 2. Two ways to improve robustness in a dynamic environment: Continuous model monitoring and retraining: Regularly update the model with new data to adapt to changes and maintain accuracy. Implement fallback systems or human-in-the-loop approaches: These allow for oversight or manual intervention when the model encounters unfamiliar or high-risk situations.

jomana-khatib · ‎08-01-2025

Can a risk reduction strategy, like insurance contracts, actually increase risk? Yes. While insurance is designed to reduce risk, it can increase risk if used irresponsibly. In the 2008 financial crisis, firms relied on insurance to cover high-risk mortgage loans, assuming they were protected. This encouraged risky behavior, such as issuing subprime loans without proper checks. When claims surged, insurers couldn’t pay, leading to widespread collapse. So, a risk reduction strategy can backfire if it encourages complacency or risky actions. How does independent oversight improve robustness? Independent oversight ensures that financial practices are transparent, accountable, and follow regulations. It helps detect and correct risky behavior before it escalates. Oversight introduces checks and balances, such as audits, regulations, and reporting standards, that reduce the chance of systemic failure. Without oversight, organizations may act in self-interest, overlook critical risks, or exploit loopholes, weakening system robustness To sum up, while risk reduction tools like insurance can help manage uncertainty, they can also create more risk if misused or unchecked. Independent oversight plays a vital role in maintaining system robustness by enforcing accountability and preventing excessive risk-taking.

jomana-khatib · ‎08-01-2025

Documentation and oversight are critical for ensuring system robustness. Clear procedures, proper financial controls, and accountability mechanisms help prevent misuse, errors, and unethical behaviour. Without them like in the case of the failed cryptocurrency exchange informal practices, lack of transparency, and inadequate recordkeeping can lead to collapse and significant financial loss. Robust systems rely on formal governance, professional auditing, and detailed documentation to function reliably and securely.

jomana-khatib · ‎08-01-2025

Informed Consent Limits: Users in crisis may not be in a mental state to give truly informed consent, even if they click "agree." Data Sensitivity: Mental health data is highly sensitive; anonymization doesn't fully eliminate risks of misuse or re-identification. Trust & Transparency: Organizations must be transparent about data use and build trust by clearly explaining both current and future applications. Ethical Governance: Independent ethical review boards can oversee decisions about secondary data use. Opt-In Over Opt-Out: Whenever possible, prioritize explicit, opt-in consent for secondary uses—especially commercial ones. Benefit Sharing: If data is used for profit, users or their communities should directly benefit, not just the organization.

jomana-khatib · ‎08-01-2025

Accountability for inclusive safety testing lies with regulators, automakers, and researchers. Organizations like NHTSA must set inclusive standards, while manufacturers and scientists are responsible for developing and using diverse test models that reflect all body types. Collaboration among these groups ensures vehicle safety meets the needs of all drivers and passengers.

jomana-khatib · ‎08-01-2025

To make ASR (Automated Speech Recognition) systems more inclusive, we must go beyond diverse training data. It's essential to include dialects like AAVE, involve affected communities in development, and conduct regular bias audits. Systems should offer user feedback options, support contextual understanding, and be built by culturally aware teams.

jomana-khatib · ‎08-01-2025

Ignoring inclusivity in healthcare technology and research can lead to misdiagnosis, mistreatment, and unequal access to care, which directly impacts the health and well-being of underrepresented populations. In the EEG scenario, traditional electrode designs that don’t account for coarse or curly hair types result in inaccurate brain readings for individuals with such hair textures. This not only skews neuroimaging research findings but also delays or misguides clinical decision-making, potentially putting patients at greater health risk. Furthermore, these biases reinforce systemic inequities and reduce trust in healthcare systems among marginalized groups. If people believe that medical tools and procedures are not designed with them in mind, they may be less likely to seek care or participate in medical research, thereby widening health disparities.

jomana-khatib · ‎08-01-2025

Police can use AI-based tips as part of a broader investigative process, but they should not rely on them as the sole source of evidence. AI-generated clues, like phone pings, often come with uncertainty and lack context. Without human validation, there's a high risk of error and unjust outcomes. Safe use of such tips requires rigorous verification, cross-checking with other evidence, and a clear understanding of the technology’s limitations. To make AI-informed actions more human-centric, law enforcement must: Verify data before acting, especially when it could lead to forceful intervention. Implement safeguards to ensure that AI outputs are interpreted with caution. Train officers and analysts to question AI results and understand their limitations. Include empathy and ethical considerations, recognizing that behind every data point is a human life. Create channels for individuals to challenge AI-driven errors and be heard if wrongly impacted. AI and data tools can enhance policing, but should be used responsibly, prioritizing human lives, empathy, ethical thinking, and due process, ensuring AI serves humanity without harm.

jomana-khatib · ‎08-01-2025

This is a reality many drivers face every day. It’s why we need systems that are not only efficient but fair, flexible, and human-centric. A human-centric rating system should: Recognize factors beyond a person’s control, like severe weather or traffic delays. Include ways for individuals to provide context and appeal unfair ratings. Prioritize safety and effort, not just raw numbers. Be transparent about how scores are calculated. Treat people as people—not just data points. Technology should support us, not judge us unfairly. A human centric rating system recognizes that life is unpredictable. It builds in fairness, context, and the ability to respond to exceptions ensuring that drivers are treated with dignity and respect, even when the unexpected happens

jomana-khatib · ‎08-01-2025

In today’s data-driven world, it’s not enough for models to be accurate—they must also be aligned with human values. When building automated systems, especially in sensitive areas like healthcare, we must ensure that recommendations go beyond technical optimization. A truly human-centric approach considers not only cost or efficiency, but also a person’s wishes, beliefs, and life circumstances. For example, a model might suggest an in-home care plan because it’s cost-effective. But if the patient has no family support or the recommendation conflicts with their values, that outcome could do more harm than good. That’s why we must design systems to support, not replace, human decision-making. By integrating ethical principles, individual preferences, and transparent logic into our models, we can create solutions that are not only smart but also compassionate. Ultimately, the goal is to ensure that technology serves people not the other way around.

Online Status	Offline
Date Last Visited	a week ago