AI Words of Wisdom

AI is the tech news of the decade. It's here to stay, moving ahead at a frightening speed, and it's going to impact everything we do.

In fact, even the vocabulary around it changes so rapidly it's hard to understand. It's somewhat akin to morphing a language 100 years in 6 months time. Though it might be wise to check those estimates with AI just to be sure!

Artificial Intelligence is the overall technology, and that can be defined (by Oxford Languages) as: "the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, decision-making, and translation between languages."

Not that long ago, books were written, films and TV series were made, about the dangers of trusting AI, or, as in the case of a movie like Blade Runner, the dangers of being AI which had become self-aware, and therefore struggled to maintain itself against a promised "retirement." Today, AI is trusted to manage all sorts of tasks, including chatting with you, answering questions, taking notes at a meeting, assessing customer reviews, and generating imagery on command.

One of the limitations of AI, we were assured, was that while it could do what computer code once did - basic, boring, repetitive chores, though on a more sophisticated and casual basis than strings of complex code that had to be written by a programmer. AI would only replace transcription, measurement, grammar correction, and generating imagery for film backgrounds or ads.

The above more or less describes the concept of generative AI. The AI learns the rules and patterns of a particular field, and can then offer solutions such as a paragraph of text, a summary of a list or set of inputs, or an image depicting a concept or, for example, historical character.

Generative AI shows itself off in such apps as "ChatGPT, Copilot, Gemini, Grok, and DeepSeek; text-to-image artificial intelligence image generation systems such as Stable Diffusion, Midjourney, and DALL-E; and text-to-video AI generators such as Sora." (Wikipedia)

For a time, at least at a surface level, AI was a bit like a toy. It could merge your face with a movie clip, or create a funny story, ask for a summary of the top 25 search results on a given topic as a basis for research. The important point was though it would generate material based on its programming, it relied on human intelligence to create the request and parameters.

Enter agentic AI. According to one source, Moveworks.com, a company engaged in consultation and implementation of agentic solutions, "Just a few years ago, only 48% of organizations were experimenting with AI. Today, that number has surged to 72%—and it’s still growing.

But we’re entering a new phase of AI adoption—one powered by agentic AI. Unlike traditional AI, which relies on predefined rules, agentic AI can reason, learn, and make decisions on its own with minimal human intervention.

"... by 2028, 33% of enterprise software applications will incorporate Agentic AI to manage complex tasks and workflows. Organizations that embrace this shift today will gain a massive competitive advantage—automating decision-making, streamlining operations, and unlocking new levels of productivity."

In other words, to keep up, you and/or your organization will have to "employ" agentic AI in order to keep pace with the implementation speed of the competition. Just as computer CPUs outpaced human brains in repetitive tasks like adding numbers or changing a date (except for Y2K?), agentic AI will be able to scan text, review an image, or create an generated army for a film background without resorting to tedious human reading and fixing, or even compositing images.

Then there is the observability factor. With things like ChatGPT and CoPilot (Microsoft), prompts (requests from the person seeking the AI help) are in "natural language." We don't need to write a string of code, but can instead in simple language describe what we want, ask a question, or engage in a conversation. If you're playing with it, you can simply enjoy what you get. If you need an analysis or something as a basis for business, the results become more sensitive. And a number of issues can arise, which one organization, https://www.datadoghq.com/, lists as: hallucinations, performance and cost issues, prompt hacking, security and data privacy issues, and model prompt and response variance. Hallucinations can include made-up answers or answers provided when one is lacking. Performance and cost issues can crop up as algorithms change or inputs degrade. Prompt hacking involves users manipulating the AI, or slyly urging it to respond with harmful or dishonest results. Security and privacy issues more or less speak for themselves, and model prompt and response variance can be either the AI generating nonsense, the user inputting nonsense, or both. Observability makes it possible for the AI provider to find, trace, and fix these types of problems.

Keeping track of where this technology is moving, and how it might impact your own work and day-to-day life, is a challenge. One quick resource: https://www.thoughtworks.com/radar