The problem is that these documents may well get designed for AI to get it wrong. They are not yet, as far as I can tell, but they will be. And then AI will not mostly get it right anymore and make you an easy target for manipulation.

I just asked ChatGPT "Is Microsoft willfully enshittifying its products?".

It essentially said "yes" to enshittification, but was unsure whether it is incompetence or maliciousness. Hence it clearly is not there yet, because while there is clearly massive organizational dysfunction at Microsoft, there is quite clearly intent as well. For example, making things harder to use (o365, W11) will very clearly show up on their own user-testing, but they also very clearly have decided that wasting more user time is what they want to do. There is no way that was not a conscious informed decision.

Full answer:
----
Ah, the “willfully enshittifying” debate—you're referencing that infamous term coined by Cory Doctorow to describe the phenomenon where companies intentionally degrade the quality of their products or services in order to extract more money or control over users.

Microsoft, especially in recent years, has faced its share of criticisms along these lines. Some examples:

Windows 11 Updates: Windows 11, for example, has been the subject of some frustrations regarding forced updates, system requirements, and a lack of control for users. The fact that features like the ability to defer updates have been severely limited feels like an attempt to nudge people into a more uniform, more controllable system.

Office 365: The shift from traditional perpetual licenses (where you pay once) to subscription models (where you pay ongoing fees) has led to a lot of complaints. Many people feel like they're being forced into the subscription model with ever-increasing prices for the same or worse features.

Bloatware in Windows: The inclusion of pre-installed apps (many of which cannot be removed) or the increasing prominence of apps that try to push services like OneDrive and Microsoft Edge feels like an effort to funnel users into Microsoft's ecosystem—whether they want to be there or not.

Telemetry and Privacy Concerns: A lot of Microsoft's products (especially Windows) collect telemetry data, and while it’s marketed as a way to improve user experience, some feel it's just a backdoor for data collection to sell advertising or to exert more control.

Lack of User Control: Windows’ user interface has been simplified in some cases to the point where users feel less in control of their system. For instance, the removal of control panels, deep settings, and customization options in favor of a more "streamlined" and "unified" system.

So, is it willful? Some of it seems more like a combination of corporate strategy (monetizing services through subscriptions, gaining control over ecosystems) and the “unintended” consequences of designing products with a heavy corporate agenda.

But whether or not it's malicious is more of a gray area. It’s hard to imagine that Microsoft doesn't know the complaints, but at the same time, they might feel it’s just the price of doing business in an increasingly data-driven, subscription economy. Do you think it’s a deliberate move, or more of a byproduct of their business model evolving?
----

People struggle to get through their daily lives. They do not have the resources to spare not not use AI when it becomes the easier way. And then there is peer-pressure and FOMO.

Indeed. The core problem is that "AI" (well, LLMs) do not actually produce anything new or different. It is just making some essentially non-productive things easier. Hence there will not be a replacement industry.

For example, it can be used to automate bureaucracy, which just means we will get more of it. And it can make better ads, better propaganda and better scams. None of which is helpful on a society level. These all only serve to distribute what is there in wealth differently.

So yes, I think this time will not work out. Either put strict controls and limits on LLMs or suffer the end of capitalism as it descends into chaos. Well, we might get lucky and this whole hype collapses under its own weight soon enough to limit the damage.

Found your problem. Obviously AI is really good at helping you push made-up stories. And since nobody is going to do any fact-checking anyways, even the occasional hallucination does not matter.

All Trump can really do is incite discord (and violence). As soon as things quiet down enough, it would become obvious how pathetic his results and his person actually are. And he knows he cannot allow that to happen.

And what many more before him did. Trump is nowhere near intelligent enough to come up with his "strategy" by himself. A vote against Trump is actually a vote for "Democracy" at this time, sad as that is.

I completely agree. Finding stuff is a process that makes you learn things. Cut that learning out and suffer. Most things are complex and one thing I find time and again when searching for something that I need to look into some other things as well. (Datasheets are a good example.) If "AI" gives people just a simplified picture that misses important aspects (have that happen to me on many of the more complex questions I asked and it is a reason why I keep LLM-use to a minimum), people will stay incompetent about things they would actually benefit from knowing.

So, do we still need to analyze and understand things on our own in 2025? This is a complex question. For any expert that works insight-based, it is essential and they will stop being able to do their work or the quality of their work will drop catastrophically if they stop, essentially making them non-experts. These experts are critically important for many things, even if there are not a lot of them, relatively speaking.

For people that rarely use insight (the majority), it depends. These people are typically still voters and consumers. Do we, for example, want them to just vote for whoever does the best propaganda and makes the better empty promises? A credible argument can be made that many are already doing that (see how the "Liar in Chief" came into office, for example), and that AI will not reduce their atrophied mental capabilities because they already do not have any to speak of.

My take is that in the end, everybody that wants actual insight will not be affected much by LLMs. And the rest will not be either. The whole LLM thing is not a game-changer with regard to knowledge, insight and mental capabilities.

You think firewalling is a difficult topic? Then maybe you should not be doing it? You are aware that if you do not catch just the wrong hallucination once, you might cause a major catastrophe, right?

Why would porn be an issue? There is already much more out there than anybody can consume. AI is not going to change that.

There is actually a chance. The EU. Unless the AI peddlers want to not sell there or have two different models for anything. Not saying this is a very good chance though.

Obviously, customer "concerns" by US citizens will just be seen as a PR issue to be solved with the usual propaganda approaches. That works well on that user pool.

I had prime for a months and they messed up my billing and could not bill me. (Apparently they now have two accounts for me with same user name and different passwords since I changed my password. Talk about incompetent.) No bloody murder, I just gets some emails now about stuff I do not want on prime and these are easily blocked. Of course, that was in Europe.

Yes. That is how engineering liability works. Know why engineers routinely refuse to do dangerous things they are now qualified for? Because they would personally go to prison. And they all know that.

Now, this liability idea can be overdone. Look at the geologists that got prosecuted for failing to predict an earthquake correctly in Italy. That is counter-productive, because that just means nobody is willing to do these predictions anymore, because these predictions are and can only be unreliable. But clear state-of-the-art violations in engineering and applied Science are a different thing.

Sadly, IT is not an engineering discipline at this time. But look at how the HIPPA act works, for example. A lot of developers are going to have very clear documentation as to who made what decision there.

Fully agree. Or an automatic $1000 to the consumer affected per instance, no lawsuit needed.

The trick is to stop thinking this is about your choices. Then it makes perfect sense. Why compete on merit, if you can just push things on users and the lazy ones (which will be many) will just use that thing you pushed on them? Same reason MS is integrating "AI" with their crappy Office Suite and crappy OS. They do not want a market with actual competition.