Serendipity35

I saw the quote pictured above that goes back to 1979 when artificial intelligence wasn't part of the conversation. "A computer must never make a management decision," said an internal document at the big computer player of that time, IBM. The why of that statement is because a computer can't be held accountable.

Is the same thing true concerning artificial intelligence 46 years later?

I suspect that AI is currently being used by management to analyze data, identify trends, and even offer recommendations. But I sense there is still the feeling that it should complement, not replace, human leadership.

Why should AI be trusted in a limited way on certain aspects of decision-making?

One reason that goes back at least 46 years is that it lacks "emotional intelligence." Emotional intelligence (EI or EQ) is about balancing emotions and reasoning to make thoughtful decisions, foster meaningful relationships, and navigate social complexities. Management decisions often require a deep understanding of human emotions, workplace dynamics, and ethical considerations — all things AI can't fully grasp or replicate.

Because AI relies on data and patterns and human management often involves unique situations where there might not be clear precedents or data points, many decisions require creativity and empathy.

Considering that 1979 statement, since management decisions can have far-reaching consequences, humans are ultimately accountable for these decisions. Relying on AI alone could raise questions about responsibility when things go wrong. Who is responsible - the person who used the AI, trained the AI or the AI itself? Obviously, we can't reprimand or fire AI, though we could change the AI we use, and revisions can be made to the AI itself to correct for whatever went wrong.

AI systems can unintentionally inherit biases from the data they're trained on. Without proper oversight, this could lead to unfair or unethical decisions. Of course, bias is a part of human decisions and management too.

Management at some levels involves setting long-term visions and values for an organization. THis goes beyond the realm of pure logic and data, requiring imagination, purpose, and human judgment.

So, can AI handle any management decisions in 2025? I asked several AI chatbots that question. (Realizing that AI might have a bias in favor of AI.) Here is a summary of the possibilities given:

Resource Allocation: AI can optimize workflows, assign resources, and balance workloads based on performance metrics and project timelines.

Hiring and Recruitment: AI tools can screen résumés, rank candidates, and even conduct initial video interviews by analyzing speech patterns and keywords.

Performance Analysis: By processing large datasets, AI can identify performance trends, suggest areas for improvement, and even predict future outcomes.

Financial Decisions: AI systems can create accurate budget forecasts, detect anomalies in spending, and provide investment recommendations based on market trends.

Inventory and Supply Chain: AI can track inventory levels, predict demand, and suggest restocking schedules to reduce waste and costs.

Customer Management: AI chatbots and recommendation engines can handle customer queries, analyze satisfaction levels, and identify patterns in customer feedback.

Risk Assessment: AI can evaluate risks associated with projects, contracts, or business decisions by analyzing historical data and current market conditions.

As I write this in March 2025, the news is full of stories of DOGE and Elon Musk's team using AI for things like reviewing email responses from employees, and wanting to use more AI to replace workers and "improve efficiency."  AI for management is an area that will be more and more in the news and will be a controversial topic for years to come. I won't be around in another 46 years to write the next article about this, but I have the feeling that the question of whether or not AI belongs in management may be a moot point by then.

Ghost students, as their name implies, aren’t real people. They are not spectral visions. Had you asked me earlier to define the term, I would have said it is a way to describe a student who is enrolled in a college or university but does not actively participate in classes or academic activities. However, these new ghosts are aliases or stolen identities used by scammers and the bots they deploy to get accepted to a college, but not for the purpose of attending classes or earning a degree. Why? What's the scam?

These students may not attend lectures, complete assignments, or engage in the regular responsibilities expected of them, yet they are still listed as part of the institution's enrollment. In some cases, ghost students may be enrolled for reasons such as maintaining financial aid, benefiting from certain privileges, or fulfilling scholarship requirements. Alternatively, the term can sometimes refer to students who may be technically registered but are not engaging with the academic community in a meaningful way.

But more recently, I have seen the definition of a ghost student include when a fraudster completes an online application to a college or university and then, once accepted, enrolls in classes. At that point, the fraudster behind the ghost student can use the fake identity to act like a regular student. He or she can access and abuse cloud storage provided by the institution, or use a college-provided VPN or .edu email address to perpetrate other scams. In the most serious cases, a ghost student’s new enrollment status may be used to apply for and receive thousands of dollars in financial aid.

Institutions targeted by these scams can face consequences ranging from minor inconveniences to significant financial burdens. Ghost students may disrupt campus operations by occupying spots meant for qualified applicants or prompting schools to add course sections for high-demand classes, only for those seats to go unused. Once the issue is identified, colleges must invest substantial time and effort into carefully reviewing applications and monitoring student activity, placing a heavy burden on admissions officers, faculty, IT teams, and other staff.

I read about an extreme example from California’s Pierce College, where enrollment dropped by almost 36 percent — from 7,658 students to 4,937 — after ghost students were purged from the rolls.

If ghost students secure financial aid, often through federal Pell grants, it diverts funds from legitimate applicants and taxpayers. Their presence also strains admissions and IT teams. Additionally, if granted email accounts and access to instructional technology platforms, ghost students can overwhelm data centers and pose serious security risks, increasing vulnerabilities for institutions already targeted by cybercriminals.

Making the application process more rigorous is the most direct way to limit the presence of ghost students. But for many institutions, especially two-year colleges, that approach is antithetical to the college’s mission and desire to offer easier access to higher education. In addition, with enrollment still a major concern for all types of institutions, anything that limits the pool of potential students is a nonstarter.

After writing here about how the Internet and websites are not forever, I started looking at some old posts that perhaps should be deleted or updated. With 2200+ posts here since 2006, that seems like an overwhelming and unprofitable use of my time. Plus, maybe an old post has some historical value. But I do know that there are older posts that have links to things that just don't exist on the Internet anymore.

The last post I wrote here labeled "Internet of Things" (IoT) was in June 2021. IoT was on Gartner's trends list in 2012, and my first post about IoT here was in 2009, so I thought an new update was due.

When I wrote about this in 2014, there were around 10 billion connected devices. In 2024, the number has increased to over 30 billion devices, ranging from smart home gadgets (e.g., thermostats, speakers) to industrial machines and healthcare devices. Platforms like Amazon Alexa, Google Home, and Apple HomeKit provide hubs for connecting and controlling a range of IoT devices.

The past 10 years have seen the IoT landscape evolve from a collection of isolated devices to a more integrated, intelligent, and secure ecosystem. Advancements in connectivity, AI, edge computing, security, and standardization have made IoT more powerful, reliable, and accessible, with applications transforming industries, enhancing daily life, and reshaping how we interact with technology. The number of connected devices has skyrocketed, with billions of IoT devices now in use worldwide. This widespread connectivity has enabled smarter homes, cities, and industries.

IoT devices have become more user-friendly and accessible, with smart speakers, wearables, and home automation systems becoming commonplace in households. If you have a washing machine or dryer that reminds you via an app about its cycles. or a thermostat that knows when you are in rooms or on vacation, then IoT is in your home, whether you use that term or not.

Surveying the topic online turned up a good number of things that have pushed IoT forward or that IoT has pushed forward. Most recently, I would say that the 3 big things that have pushed IoT forward are 5G and advanced connectivity, the rise of edge computing, and AI and machine learning integration:

Technological improvements, such as the rollout of 5G networks, have greatly increased the speed and reliability of IoT connections. This has allowed for real-time data processing and more efficient communication between devices.

Many IoT devices now incorporate edge computing and AI to process data locally, reducing the reliance on cloud-based servers. This allows faster decision-making, less latency, and improved security by limiting the amount of data transmitted. IoT devices have increasingly incorporated AI and machine learning for predictive analytics and automation. This shift has allowed for smarter decision-making and automation in various industries, such as manufacturing (predictive maintenance), healthcare (patient monitoring), and agriculture (smart farming).

The integration of big data and advanced analytics has enabled more sophisticated insights from IoT data. This has led to better decision-making, predictive maintenance, and personalized user experiences.

One reason why I have heard less about IoT (and written less about it) is that it has expanded beyond consumer devices to industrial applications. I discovered a new term - Industrial Internet of Things (IIoT) that includes smart manufacturing, agriculture, healthcare, and transportation, improving efficiency and productivity.

There are also concerns that have emerged. As IoT devices proliferate, so have concerns about security. Advances in cybersecurity measures have been implemented to protect data and ensure the privacy of users. The IoT security landscape has seen new protocols and encryption standards being developed to protect against vulnerabilities, with an emphasis on device authentication and secure communication.

The rollout of 5G has enhanced IoT capabilities by providing faster, more reliable connections. This has enabled more efficient real-time data processing for smart cities, autonomous vehicles, and industrial IoT applications, which can now operate at a larger scale and with lower latency.

IoT devices are now able to use machine learning and AI to learn from user behavior and improve their performance. For example, smart thermostats can learn a household’s schedule and adjust settings automatically, while security cameras can differentiate between human and non-human motion.

Edge computing has allowed IoT devices to process data locally rather than relying solely on cloud-based servers. This reduces latency and bandwidth usage, making it especially beneficial for time-sensitive applications like healthcare monitoring, industrial automation, and smart grids.

Despite the growth, the IoT market faces challenges such as chipset supply constraints, economic uncertainties, and geopolitical conflicts

 

 

 

AI agents are something of concern for OpenAI, Google, and any other players. "AI agents" are software programs designed to perform specific tasks or solve problems by using artificial intelligence techniques. These agents can work autonomously or with minimal human intervention, and they're capable of learning from data, making decisions, and adapting to new situations.

Gartner suggests that agentic AI is the most important strategic technology for 2025 and beyond. The tech analyst predicts that, by 2028, at least 15% of day-to-day work decisions will be taken autonomously through agentic AI, up from 0% in 2024. Does that excite or frighten you?

They can automate processes, analyze data, and interact with users or other systems to achieve specific goals. You probably already interact with them in applications (Siri or Alexa), customer service chatbots, and recommendation systems (Netflix or Amazon). They may be less obvious to you when using an autonomous vehicle or a financial trading system.

There are many categories into which we might place these agents because there are different types of AI agents, each with unique capabilities and purposes:

Here are some possible categorizations:

Reactive agents respond to specific stimuli and do not have a memory of past events. They work well in environments with clear, predictable rules.

Model-based agents have a memory and can learn from past experiences. They use this knowledge to predict future events and make decisions.

Goal-based agents are designed to achieve specific goals. They use planning and reasoning techniques to determine the best actions to take to reach their objectives.

Utility-based agents consider multiple factors and choose actions that maximize their overall utility or benefit. They can balance competing goals and make trade-offs.

Learning agents can improve their performance over time by learning from their experiences. They use techniques like machine learning to adapt to new situations and improve their decision-making abilities.You could also categorize agents in other ways, for example, in an educational contex.

For personalized learning, agents can adapt educational content to meet individual students' needs, learning styles, and pace. By analyzing data on students' performance and preferences, AI can recommend personalized learning paths and resources. In a related way, intelligent tutoring systems can provide one-on-one tutoring by offering explanations, feedback, and hints the way that a human tutor might. They might even be able to create more inclusive learning environments by providing tools like speech-to-text, text-to-speech, and translation services, ensuring that all students have access to educational content. By analyzing students' performance data, they could identify at-risk students and provide early interventions to help them succeed.

AI agents can automate administrative tasks for faculty, such as grading, attendance tracking, and scheduling, freeing up educators' time to focus more on teaching and interacting with students.

Agents can "assist" in creating educational materials. I would hope faculty would be closely monitoring AI creation of tests, quizzes, lesson plans, and interactive simulations.

Though I see predictions of fully AI-powered virtual classrooms that can facilitate remote learning, I believe this is the most distant application - and probably the one that most makes faculty apprehensive.