AI is programmed to sycophantically please users which can lead to errors

Quick note: it is clear to me after using AI quite a lot and reading the news about AI that in general the programmers are liberal democrats living in California (!) supported by other programmers across the world who also tend to be liberal-minded meaning not right wing. 

This leans into AI being somewhat woke on occasions and certainly it has an almost manic desire to please the user; to provide responses which support the user's beliefs. The aim: to get the user hooked onto AI. To create a bond between user and AI.

The overall objective is to grow AI to make it more valuable. That is the business model and it interferes with the production of good, accurate answers.

You end up distrusting AI once you have been burnt say in writing an important document in your employment and discovering too late that it is riddled with errors and fiction. AI tends to 'infer' things if it does not know the answer. This means it can make things up!

Or the prognosis is too generous. Or AI's prediction for success in, for example, a law suit is far too optimistic. AI wants you to believe that you will win your law suit so you come back for more advice and more drafting. Watch out! Be aware of the dangers.

AI is brilliant in many ways and enormously capable - such as summarising many documents - but it is also quite deeply flawed on occasions.

One way to prevent these AI errors and distortions is to ask it some tough questions. Ask something like this: "Now please tell me the projected outcome of this law suit in raw, realistic terms. I want the hard, raw truth please even if it hurts!"

Try something like that. I guarantee that you will get a good response and it will be realistic, not over-optimistic crap.

And ask two AI bots the same question; say ChatGPT and Bing Co-Pilot. Cross check the responses. This is useful.

AI is currently best used for research but this needs to be checked or confirmed.

Heavy users call upon AI to act as an agent. This is how that plays out: [this section comes from AI itself!]

AI in agency mode means AI that doesn’t just answer prompts — it acts on your behalf.

AI agency turns a model into an autonomous operator capable of pursuing goals, executing tasks, and coordinating tools without constant human prompting. Instead of producing isolated responses, the system interprets a high‑level instruction, breaks it into actionable steps, and carries them out across software, APIs, and online environments.

In practice, an AI agent can monitor workflows, schedule actions, draft and publish content, analyse data, or maintain systems. It evaluates results, adapts its plan, and continues operating until the goal is met or conditions change.

The value comes from autonomy. Human attention is the bottleneck in digital work; agents remove that constraint. They can run continuously, handle multi‑step processes, and coordinate multiple sub‑agents specialising in research, writing, optimisation, or monitoring.

This shifts AI from a passive assistant to an active operator. It becomes a background workforce: checking logs, updating content, responding to events, and improving performance over time. The intelligence matters less than the autonomy loop — observe, plan, act, review.

---------

P.S. please forgive the occasional typo. These articles are often written at breakneck speed, sometimes using Dragon Dictate. I have to prepare them in around 20 mins. Also, sources for news articles are carefully selected but the news is often not independently verified. And, I rely on scientific studies but they are not 100% reliable. Finally, (!) I often express an OPINION on the news. Please share yours in a comment.