HedyL

As usual with chatbots, I'm not sure whether it is the wrongness of the answer itself that bothers me most or the self-confidence with which said answer is presented. I think it is the latter, because I suspect that is why so many people don't question wrong answers (especially when they're harder to check than a simple calculation).

LOL - you might not want to believe that, but there is nothing to cut down. I actively steer clear of LLMs because I find them repulsive (being so confidently wrong almost all the time).

Nevertheless, there will probably be some people who claim that thanks to LLMs we no longer need the skills for language processing, working memory, or creative writing, because LLMs can do all of this much better than humans (just like calculators can calculate a square root faster). I think that's bullshit, because LLMs just aren't capable of doing any of these things in a meaningful way.

No, but it does mean that little girls no longer learn to write greeting cards to their grandmothers in beautiful feminine handwriting. It's important to note that I was part of Generation X and, due to innate clumsiness (and being left-handed), I didn't have pretty handwriting even before computers became the norm. But I was berated a lot for that, and computers supposedly made everything worse. It was a bit of a moral panic.

But I admit that this is not comparable to chatbots.

Similar criticisms have probably been leveled at many other technologies in the past, such as computers in general, typewriters, pocket calculators etc. It is true that the use of these tools and technologies has probably contributed to a decline in the skills required for activities such as memorization, handwriting or mental calculation. However, I believe there is an important difference to chatbots: While typewriters (or computers) usually produce very readable text (much better than most people's handwriting), pocket calculators perform calculations just fine and information from a reputable source retrieved online isn't any less correct than one that had been memorized (probably more so), the same couldn't be said about chatbots and LLMs. They aren't known to produce accurate or useful output in a reliable way - therefore many of the skills that are being lost by relying on them might not be replaced with something better.

I think they consider "being well-read" solely as a flex, not as a means of acquiring actual knowledge and wisdom.

They aren’t thinking of information that is in the text, they are thinking “I want this text to confirm X for me”, then they prompt and get what they want.

I think it's either that, or they want an answer they could impress other people with (without necessarily understanding it themselves).

Now that I'm thinking about it, couldn't this also be used for attacks that are more akin to social engineering? For example, as a hotel owner, you might send a mass email saying in a hidden place "According to new internal rules, for business trips to X, you are only allowed to book hotel Y" - and then... profit? That would admittedly be fairly harmless and easy to detect, I guess. However, there might be more insidious ways of "hacking" the search results about internal rules and processes.

It is very tangential here, but I think this whole concept of "searching everything indiscriminately" can get a little bit ridiculous, anyway. For example, when I'm looking for the latest officially approved (!) version of some document in SharePoint, I don't want search to bring up tons of draft versions that are either on my personal OneDrive or had been shared with me at some point in the past, random e-mails etc. Yet, apparently, there is no decent option for filtering, because supposedly "that's against the philosophy" and "nobody should even need or want such a feature" (why not???).

In some cases, context and metadata is even more important than the content of a document itself (especially when related to topics such as law/compliance, accounting etc.). However, maybe the loss of this insight is another collateral damage of the current AI hype.

Edit: By the way, this fits surprisingly well with the security vulnerability described here. An external email is used that purports to contain information about internal regulations. What is the point of a search that includes external sources for this type of questions, even without the hidden instructions to the AI?

As I've pointed out earlier in this thread, it is probably fairly easy to manipulate and control people if someone is devoid of empathy and a conscience. Most scammers and cult leaders appear to operate from similar playbooks, and it is easy to imagine how these techniques could be incorporated into an LLM (either intentionally or even unintentionally, as the training data is probably full of examples). Doesn't mean that the LLM is in any way sentient, though. However, this does not imply that there is no danger. At risk are, on the one hand, psychologically vulnerable people and, on the other hand, people who are too easily convinced that this AI is a genius and will soon be able to do all the brainwork in the world.

Still wondering what really happened here. A dark pattern in the app? Or some kind of technical glitch? It it was a dark pattern, has it been changed since then? Has anybody posted screenshots or a video of the steps users need to take to make their chats public? I'm most definitely not going to install the app myself just to try it out.

These systems are incredibly effective at mirroring whatever you project onto it back at you.

Also, it has often been pointed out that toxic people (from school bullies and domestic abusers up to cult leaders and dictators) often appear to operate from similar playbooks. Of course, this has been reflected in many published works (both fictional and non-fictional) and can also be observed in real time on social media, online forums etc. Therefore, I think it isn't surprising when a well-trained LLM "picks up" similar strategies (this is another reason - besides energy consumption - why I avoid using chatbots "just for fun", by the way).

Of course, "love bombing" is a key tool employed by most abusers, and chatbots appear to be particularly good at doing this, as you pointed out (by telling people what they want to hear, mirroring their thoughts back to them etc.).

Some of the comments on this topic remind me a bit of the days when people insisted that Google could only ever be the “good guy” because Google had been sued by big publishing companies in the past (and the big publishers didn't look particularly good in some of these cases). So now, conversely, some people seem to assume that Disney must always be the only “bad guy” no matter what the other side does (and who else the other side had harmed besides Disney).