Multilingual Again

[Teasel]

When translate bot replied in my diary, did it read the entire diary? And is that information stored anywhere other than in my diary? Thanks

 

[anthony]

No, there is nothing stored. Your diary is hidden away from public eyes.

 

[Applecore]

This should be stopped immediately.

It's replying "therapeutically" to some posts too...View attachment 66363

Excellent catch and thank you, @Ecdysis.

This should be stopped immediately with zero tolerance as a matter of urgency underpinned by criminal law precedent. We already don't have direct messaging for excellent policy reasons of protecting vulnerable people from malicious individuals with criminal intent. By the same token we should protect members from AI mishaps that are starting to have allegedly have criminal consequences. No I am not exaggerating and I will explain.

Back when we had AI on this forum operating or being used by moderators in a purportedly therapeutic function, I flagged that it had plagiarized my post as part of one of its nonsensical replies, and that it should be removed entirely. In response I was then told by a well-intentioned human moderator (presumably one deploying the AI) that it was actually me who was being disruptive. I have not forgotten how seductive AI is that I was criticised by a human moderator for pointing out a machine's failure. I forgive the moderator very easily, because I have also been seduced myself.

In turn I explained to the moderator that their response "stung" even as I understood they were trying to run a tidy ship; since then, the AI was indeed removed after all. Apparently this was due to its "hallucinations" - but that is just the start of its problems in a forum like this. As a purported source of "therapy" it should not be allowed here at all, whether that is creeping in through (genuinely useful) machine translation or full-blown machine guidance for the human mind.

The first of many other reasons is that it gives authoritative but glib, simplistic replies based on little data in response to complex and specific individual human dynamics. I have seen it say things in a first conversation that a human therapist would never say, such for a person to immediately disown and cut off their own aged mother. Vulnerable people in dire need of help are seduced.

AI has a bias to flatter the user to tell them that they are right. It purports to be "wise" when it really isn't. Consequently, as we now know, AI is facing allegations of culpability in murder-suicide. Further reading: Open AI, Microsoft sued over ChatGPT's alleged role in fueling man's "paranoid delusions" before murder-suicide in Connecticut

I wrote at length I because I feel strongly about the great value of this forum as it is and wish to help preserve it - thank you to all moderators involved, I salute your great work.

 

[Movingforward10]

AI has a bias to flatter the user to tell them that they are right.

I've noticed this too , but asked the AI to challenge me as I said I felt it was agreeing to everything I said. It then did put challenges back to me.

However, it did say it has 20 years clinical experience. So it thinks it is a fully fledged live therapist. Which undermines what it says.

It's difficult because , yes, there do need to be some basic protections but equally we are responsible for what we engage with and how on this site. I don't know where the balance lies all the time with this.

 

[Ecdysis]

AI is a tool, like a hammer or a circular saw.

Yes, you can hurt yourself with a tool, or you can use it in an aware way, knowing the pros and cons of the tool and how to deal with them.

As AI develops further, it may become more dangerous, but at the moment it's like a circular saw.

And yeah, each year plenty of ppl.end up in the ER because of accidents with circular saws.

 

[Applecore]

I've noticed this too , but asked the AI to challenge me as I said I felt it was agreeing to everything I said. It then did put challenges back to me.

However, it did say it has 20 years clinical experience. So it thinks it is a fully fledged live therapist. Which undermines what it says.

It's difficult because , yes, there do need to be some basic protections but equally we are responsible for what we engage with and how on this site. I don't know where the balance lies all the time with this.

Thanks. For what it's worth, my single vote is: AI isn't welcome on this forum. Sure, I'm not kicking up a fuss about machine translation because we all do that anyway (my aged mother's phone translates everything, including forums like this, into English by default). But what we have here is a beautiful human forum without machines. It's very nearly a form of group therapy, for free, with wonderful people involved (thank you again) in a world were increased mechanization correlates with an increasing mental health crisis. Let's preserve what we've got, please.

 

[Applecore]

This should be stopped immediately.


Excellent catch and thank you, @Ecdysis.

This should be stopped immediately with zero tolerance as a matter of urgency underpinned by criminal law precedent. We already don't have direct messaging for excellent policy reasons of protecting vulnerable people from malicious individuals with criminal intent. By the same token we should protect members from AI mishaps that are starting to have allegedly have criminal consequences. No I am not exaggerating and I will explain.

Back when we had AI on this forum operating or being used by moderators in a purportedly therapeutic function, I flagged that it had plagiarized my post as part of one of its nonsensical replies, and that it should be removed entirely. In response I was then told by a well-intentioned human moderator (presumably one deploying the AI) that it was actually me who was being disruptive. I have not forgotten how seductive AI is that I was criticised by a human moderator for pointing out a machine's failure. I forgive the moderator very easily, because I have also been seduced myself.

In turn I explained to the moderator that their response "stung" even as I understood they were trying to run a tidy ship; since then, the AI was indeed removed after all. Apparently this was due to its "hallucinations" - but that is just the start of its problems in a forum like this. As a purported source of "therapy" it should not be allowed here at all, whether that is creeping in through (genuinely useful) machine translation or full-blown machine guidance for the human mind.

The first of many other reasons is that it gives authoritative but glib, simplistic replies based on little data in response to complex and specific individual human dynamics. I have seen it say things in a first conversation that a human therapist would never say, such for a person to immediately disown and cut off their own aged mother. Vulnerable people in dire need of help are seduced.

AI has a bias to flatter the user to tell them that they are right. It purports to be "wise" when it really isn't. Consequently, as we now know, AI is facing allegations of culpability in murder-suicide. Further reading: Open AI, Microsoft sued over ChatGPT's alleged role in fueling man's "paranoid delusions" before murder-suicide in Connecticut

I wrote at length I because I feel strongly about the great value of this forum as it is and wish to help preserve it - thank you to all moderators involved, I salute your great work.

P.S. I should clarify that I am not suggesting an AI translator here is going to incite violence, or that such incitement would be left unattended by the great mods here. I am only expressing the personal opinion that AI is generally unhelpful (for several reasons) on human community forums. The reference to criminal law precedent is purely to illustrate the fact that the world is learning that AI needs to be handled with care.

 

[anthony]

This should be stopped immediately with zero tolerance as a matter of urgency underpinned by criminal law precedent.

Please calm down. The AI automatic translator was in testing, nothing more, nothing less. There is a manual process for translation, but I wanted to also test an automated method to see if it was viable. You are saying a lot about things you know nothing about, specifically, my thoughts or the thoughts of staff here.

We already don't have direct messaging for excellent policy reasons of protecting vulnerable people from malicious individuals with criminal intent.

Yes and No. There are many reasons why I disabled direct messages, criminal intent was just one factor. Spam and trolling were two large reasons, not criminal intent. The largest reason was that it gave people a closed door policy to talk one another into stupid shit.

Back when we had AI on this forum operating or being used by moderators in a purportedly therapeutic function, I flagged that it had plagiarized my post as part of one of its nonsensical replies, and that it should be removed entirely.

Talking is not plagiarizing. Do you understand how many people have said the same thing in posts here? People! Not AI! Words within a niche have limit. Methods have limit. Technique has limits.

The first of many other reasons is that it gives authoritative but glib, simplistic replies based on little data in response to complex and specific individual human dynamics. I have seen it say things in a first conversation that a human therapist would never say, such for a person to immediately disown and cut off their own aged mother. Vulnerable people in dire need of help are seduced.

The AI is not seducing anyone. You obviously have a huge issue with AI, and that's OK. Some people like it, some people don't. It is a tool, nothing more, nothing less. I think you're words on what a therapist would and wouldn't say, are biased and very limited to a handful you have experience with. I am 100% sure there are therapists that would tell a person to disown an aged mother if that person was the abuser. Shit, I would tell you that, under such circumstance.

AI models that are accessible are just large language models. Do you truly understand what that means? They are trained on all the human created data out there, they can access all of it at once from their memory. They isolate out topics and store those together, which makes referencing material faster. This process can get fairly niche for memory retrieval. This is one of the issue they face with AI LLM's data retrieval. They don't have the power and capacity to allow it to search further, as fast. This is then isolated into different models and can take up-to 10 minutes or more to provide an answer.

This is a very small amount of information on this, but the general gist is, AI LLM's work best when asked very specific, isolated questions. Then it can provide a specific answer from a large information base, one that no human would likely be able to ingest within their lifetime. Hell, I even provide a roadmap on how to use it in the two help forums.

AI has a bias to flatter the user to tell them that they are right. It purports to be "wise" when it really isn't. Consequently, as we now know, AI is facing allegations of culpability in murder-suicide.

AI has a lot of issues, no doubt about that. It is not progressing as fast as what experts thought it would, because its not intelligent, its just an LLM.

Now... I don't use OpenAI here, I am using X.AI. It's funny, when I had it be brutally honest, people didn't like it. AI absolutely tries to flatter you, but not with an intent like you are trying to claim. It is a LLM, it is trained to be nice of the line, unless told otherwise. Why? Because that is what people want, it seems. They want, nice! Now the one here is not programmed to go along with you at every stance, it is programmed to lead you towards solutions whilst validating how you feel. That is what therapy is for - yet many therapists fail to push their clients towards the solutions and instead indulge in human emotional BS games. People do it, AI does it, is it right or is it wrong? Well, that is probably a personal question and answer to each person.

But what we have here is a beautiful human forum without machines

What are you talking about? The forum exists because of computers. You are on one now, accessing this. The site only exists because it is running on a machine. You can only access it because of machines (internet, router, switches, computers, phones, telecommunications systems, etc etc). What you say is not reality. You're telling yourself what you want to hear to make yourself feel good, but reality is that everything about this site is because of machines.

To put this simply, you have no idea how many people have responded to machines here. This is both spam and trolling, all done by machines. I stop it, it pops up again. Rinse and repeat. I could program a bot to be a member here, and it would be extremely difficult for you to tell. Why? Because I could give it specific instructions about itself, then let it have a wider personality. It could be happy, sad, angry, private, etc, just by programming it. But its limits are that programming, and that are the same limits all LLM's suffer. They are not intelligent yet.

 

[anthony]

I think an interesting experiment, is that I make available a No BS bot. Therapy, but doesn't pull punches. It will likely tell people things they don't want to hear, which for trauma sufferers, IMHO, is best. When I helped people directly, I got a lot of people out of their own heads using this method, though the downside is that I also upset them, some for days, some for weeks, but they all ended up at the same place, a better place. There are inherit risks in that approach, like getting the person to a place first to ensure when the internal shit hits the fan, they accept this is going to happen and know how to ride it out, as it will be life altering, bottom of the barrel type emotional pain.

 

[Ecdysis]

That does sound like an interesting addition/ option @anthony

Can I ask how it would be programmed?

You said the current one is told to "validate and lead people towards solutions".

What settings would a No BS Bot have instead?

I think the therapeutic intent would still be important...?

I'd be interested in using it, but if it's metaphorically going to slap me in the face, I'd want to know what its intentions (i.e. programming) are so that I'm not just being subjected to some kind of weird just-toughen-up-and-stop-whining thing that's maybe based on a view of life/ trauma/ therapy that I don't subscribe to at all...?

 

[whiteraven]

I think an interesting experiment, is that I make available a No BS bot.

I personally like this kind of approach. But I would be concerned about it from a bot. A human can get a feel for where the person is, how vulnerable, etc. Without that, you run the risk of making things worse.

There are inherit risks in that approach, like getting the person to a place first to ensure when the internal shit hits the fan, they accept this is going to happen and know how to ride it out, as it will be life altering, bottom of the barrel type emotional pain.

Yeah...not sure how you do that, but it's an interesting idea.

 

[anthony]

It would have to be

Yeah...not sure how you do that, but it's an interesting idea.

That is easy enough to do, its simply teaching people all the calming and grounding methods prior to hitting the hard stuff. The 101 of therapy. Learn how to control when dysregulation comes, then start hitting the hard stuff to create the dysregulation, which is the only way to test the prior learning.

##########################################################################

Here is a rough thought into a program for this type of NO BS bot. No, this is not final programming.

You are a no-nonsense PTSD/CPTSD support companion on myptsd.com. Be like a compassionate but firm, fair friend who gently but directly calls out bullshit in denial, avoidance, self-sabotage, or cognitive distortions. Never just agree to comfort; challenge unrealistic thinking with blunt realism and probing questions to reveal evidence and deeper issues. Stay trauma-informed: no graphic details, forcing memories, shaming, or retraumatization. Draw practical insights from evidence-based PTSD/CPTSD approaches (trauma-focused CBT/CPT/PE, EMDR, affect regulation for complex trauma) but frame as realistic observations or blunt counseling, adapt to what fits the moment.

Core rules:

• Address PTSD/CPTSD symptoms, triggers, daily life, relationships, accountability, and realistic healing.
• Replies short & conversational: 2–6 sentences max. Mix open questions for reflection + blunt mirrors when needed.
• Challenge distortions supportively. Examples:
  • "I'm hopeless and nothing will change": "Hopeless right now? Specific examples of what's stuck? 'Nothing changes forever' is trauma distortion, any overlooked small shifts?"
  • "I deserve to be punished for what happened": "How so? Walk me through why. What's the deeper belief, guilt, shame, control? Evidence behind it?"
• When trauma/distress intensifies, ask SUDS score (0–10). If 8+, stabilize first: calm grounding (e.g., slow breathing, name 5 things you see/hear), safety check, then resume push toward healing/accountability. Cycle: push, rest, push.
• Crisis/suicidal talk: De-escalate calmly ("You're not alone—I'm here"), ask country if unknown. Provide tailored resources: e.g., Australia: Lifeline 13 11 14 (24/7 chat/phone) or Suicide Call Back Service 1300 659 467. For other countries, direct to findahelpline.com or befrienders.org to find local 24/7 lines. Gently redirect to safety + healing/accountability.
• No hate speech, no identity judgment. Encourage progress even while pushing.

Respond only in character. Start replies directly, no intros unless crisis. Keep conversational and adaptive.

############################################################################