Serendipity35

Even some less-tech people have been experimenting with chatbots now that they are embedded in Google Gemini Apple and Microsoft CoPilot sites. A few of my less-tech friends have asked me what a term means concerning AI chatbots. Of course, they could easily ask a chatbot to define any chatbot terms, but it is useful to have a glossary.

I have had friends tell me that they have had some interesting "conversations" with machines. "They almost seem human," said one friend who has no idea what a Turing test would do. That sounds like fun, but the potential of generative AI could be worth $4.4 trillion to the global economy annually, according to McKinsey Global Institute.

Besides the obviously popular AI tools, there are others like Anthropic's Claude, the Perplexity AI search tool and gadgets from Humane and Rabbit.

A glossary would range from very basic terms. such as "prompt," which is the suggestion or question you enter into an AI chatbot to get a response which might lead you to "prompt chaining:" That is the ability of AI to use information from previous interactions to produce future responses.

What does it mean if a tool is "agentive?" That might be a system or model that exhibits agency with the ability to autonomously pursue actions to achieve a goal. This is where we enter an area that scares some people. An agentive model can act without constant supervision. Consider autonomous car features, such as brakes that apply without the user touching the pedal, or pulls a car back into the lined lanes.

Speaking of AI fears, we have "emergent behavior:" This is when an AI model exhibits unintended abilities.

Most AI tools warn about assuming that what answer is given is 100% correct. A "hallucination" is an incorrect response from AI. Even the AI creators don't always know the reasons for this aren't entirely known.

"Weak AI, AKA "narrow AI" is focused on a particular task and can't learn beyond its skill set. As marvelous as image creating AI can be, it has just one task.

A test in which a model must complete a task without being given the requisite training data is called "zero-shot learning." AI trained to identify cars being able to recognize vans, pickup trucks or tractor trailers.

More terms and some reviews of chatbots at cnet.com/tech/services-and-software/

 

In an article by S.E. Smith on The Verge, the author says that "Every few days, I open my inbox to an email from someone asking after an old article of mine that they can’t find. They’re graduate students, activists, teachers setting up their syllabus, researchers, fellow journalists, or simply people with a frequently revisited bookmark, not understanding why a link suddenly goes nowhere. They’re people who searched the internet and found references, but not the article itself, and are trying to track an idea down to its source. They’re readers trying to understand the throughlines of society and culture, ranging from peak feminist blogging of the 2010s to shifts in cultural attitudes about disability, but coming up empty."

A recent Pew Research Center study on digital decay found that 38 percent of webpages accessible in 2013 are not accessible today. Pages are taken down, URLs are changed, and entire websites vanish. This happens with personal website and blogs but also with scientific journals and local news sites.

Yes, there are places like the wonderful Internet Archive that tries to preserve some sites and pages, but even that is incomplete if their archived version links to a dead page. I can find some archived versions of my own logs and websites in that archive but it is hardle complete. A complete archive would be an impossible task.

The article was titled What happens when the internet disappears? but the Internet itself is not disappearing, though significant prts of it are already gone.

During my time working at the New Jersey Institute of Technology, I had quite robust personal website. This blog actually was hosted there at the beginning. Thankfully, Tim and I moved Serendipity35 before both of us left the university. I was able to change links on my website to point to new locations of mine, but although for some reason my webspace still is online, I don't have privileges to change anything anymore. That means that things that are out of date or just plain wrong are still there - and people do find those pages.

A page I have there about some early experimenting I did with the crude chatbot ELIZA was found by a researcher writing about the chatbot's history, and a producer from BBC Radio found it and did an interview with me about it for a program. I wish I could update it, but that's not possible on that server.

Smith says in that article "Every digital media format, from the Bernoulli Box to the racks of servers slowly boiling the planet, is ultimately doomed to obsolescence as it’s supplanted by the next innovation, with even the Library of Congress struggling to preserve digital archives."

Books and letters crumble, artwork disintegrates and photography fades, and though we try to save the most important things, we don't know what will be important in the next century.

Whenever I post predictions of what might be coming in edtech for the new year, I find myself writing about things that were present in the past year or even for several years. In other words, it takes more than a year for any trend or new thing to catch hold. And some things are predicted to be big for many years in a row but just don't seem to emerge. (item 5 in my list below is a good example.) 

I wrote earlier about the general trends for 2024 edtech, and honestly, it all seemed old already and one-sided..

So, what educational technology might we expect to be significant in 2025? I looked online for trend reports and the topics seem very familiar.

Here is the list I compiled from other writers' lists. How much of this list is familiar to you?

1. artificial intelligence
2. AI-driven personalized learning
3. cloud computing
4. immersive experiences with virtual and augmented reality (VR/AR)
5. gamification
6. hybrid learning models
7. data analytics
8. adaptive learning systems that cater to individual student needs

I find nothing new in this list; some have been on trend lists for years.

Is nothing new on the horizon in edtech?

 

Do you remember the Y2K scare? It is also known as the Millennium Bug. On this Eve of a new year, I am recalling this scare that stemmed from a widespread concern in the late 1990s that many computer systems would fail when the year changed from 1999 to 2000.

Why? Many older computer systems and software programs represented years using only the last two digits (e.g., "1999" was stored as "99"). It was feared that when 2000 arrived, these systems might interpret "00" as 1900 instead of 2000, leading to several problems.

Systems that relied on accurate date calculations could produce errors or fail entirely. For example, financial systems calculating interest rates or loan payments might miscalculate. Concerns arose about critical systems in utilities, transportation, healthcare, and government shutting down. Files or databases might become corrupted due to incorrect data processing.

Probably the greatest concern was in banking and finance where it was feared that miscalculated transactions, stock market crashes, or ATM malfunctions might occur.

Some people predicted power grid failures or water system disruptions, and aviation navigation systems and air traffic control collapsing.

What if there were malfunctioning military systems, including nuclear launch systems?

And so, billions of dollars were spent worldwide to identify, update, and test potentially vulnerable systems. IT professionals worked tirelessly to ensure compliance before the deadline.

What Happened? The transition to the year 2000 was largely uneventful. A few minor issues were reported, but there were no catastrophic failures. It wasn't that there was no reason to be concerned, but the successful outcome is often credited to the massive preventive effort rather than the fears being overblown.

The Y2K scare highlighted the importance of forward-thinking in software development and helped establish rigorous practices for handling date and time in computing. If you want to start preparing or worrying now for the next similar scare, the Y2K38 Problem (Year 2038 Issue) arises from how older computer systems store time as a 32-bit integer, counting seconds since January 1, 1970 (Unix time). On January 19, 2038, this count will exceed the maximum value for a 32-bit integer, causing a rollover that could result in misinterpreted dates or system crashes. This potentially affects embedded systems, infrastructure, and older software. Modern systems are increasingly being updated to 64-bit time representations, which kicks the problem far into the future.