Navigating AI Responses: Maximizing the Benefit but Minimizing Confusion?

[Tinyflame]

Just wondering, but was unsure whether it's appropriate to post this as a thread, if anyone is having trouble getting through the volume of responses with the AI new feature? If I understand correctly it will eventually be at cost, and is designed to be personalized. And is being tweaked.

I have come across a few good bits of info in some responses, but the sheer volume is overwhelming, one thread could be an all-day affair. Sort of like reading a textbook? Also, unless I read the thread from the start I lose track. Similarly as I read along if I have a question along the way or something is unclear or contradictory (as far as what I've learned before) and the person engaging doesn't need it to be asked I am left a bit more confused. (Which doesn't take much, granted, lol.) So I am just thinking each thread is more designed to be tailored to mimic an individual's therapy experience, rather than for (eventual) general public guidance or discussion?

Hope that makes sense as I've written it? Thank you!

 

[somerandomguy]

I agree that it's almost impossible to keep up.

Maybe the answer is to start using it ourselves by asking it our own questions and not worry about a duplicative effect? But I personally really don't have the time to sink into it, unfortunately. It looks pretty awesome.

 

[Weemie]

I'd just keep in mind as well that most of these conversations are not legitimate. (Some are! Which is where it can seem confusing. I've used legitimate examples and asked it questions I actually wanted to as well. But even these threads are based in user testing.)

They are us testing the AI for accuracy and pushing its response limits so you may have a thread like mine where I get the result I was looking for then suddenly change the topic or create completely fictional/irrelevant scenarios or even trollish behavior. It may not make much sense at first but ultimately it will result in tweaks to the programming that benefit everyone.

You can also specifically ask the AI to respond in ways that work best for you! You can also correct it and ask it to provide its method of reasoning or sources, and refine the information it gives you until you understand. If you need less information, tell it to answer as simply and concisely as possible and it will (attempt to) do so.

So I am just thinking each thread is more designed to be tailored to mimic an individual's therapy experience, rather than for (eventual) general public guidance or discussion?

But yes as far as I understand it these threads will be entirely individual oriented and I believe it is even intended to lock the thread to just the thread starter as the AIs become confused when more than one person comments. So you would only need to focus on your specific interactions with the AIs.

 

[No More]

Sorry if my thread jumps around a bit. I am trying to focus my response to the AI down to one point, and I can pick something up in the response I want to follow up on / know more about and off I go. I am a bit squirrel brain at the best of times!

Me personally I am genuinely attempting to use the AI as an ‘informal’ therapy & for psychoeducation, rather than testing its limits as an AI

 

[Sideways]

Just wondering, but was unsure whether it's appropriate to post this as a thread, if anyone is having trouble getting through the volume of responses with the AI new feature?

I’m going to leave the thread where it is for the moment, although we probably will shift it to another part of the forum soon (you’ll know if/when that happens).

If the AI forums are currently interrupting the way you’re using the forum, particularly while it’s being tested, you can Ignore the Dr Catalyst and Dr Chase forums, and go back to them when you have time. That should remove the test threads from appearing in your regular feed.

 

[anthony]

I am happy for this thread to remain here for members to discuss the system, as I am not allowing anyone but those active subscribers access to the AI forums themselves, beyond reading.

So I am just thinking each thread is more designed to be tailored to mimic an individual's therapy experience, rather than for (eventual) general public guidance or discussion?

When a user starts a thread, what the thread contains is up to them, like everywhere else.

I believe it is even intended to lock the thread to just the thread starter as the AIs become confused when more than one person comments

Yes, each thread is limited to you and the AI, which isn't about confusion, but more that it will respond to each post, so if you have members responding to the authors content and/or AI responses, the author will start getting AI responses in their thread to other members content. It is on or off, limited to post count or unlimited. By locking each thread to the author and AI (admins excluded), it allows me to set the AI to unlimited responses until the author decides to stop responding, thus ending that thread.

General

As outlined above, I have members testing it at present. When I say testing it, they can use it for their own therapy / discussion / help, or they may just be trolling the crap out of it to find issues. Both use cases help me refine the programming, which happened fairly quickly over a 48hr period, to where I think its nearly ideal, beyond things I do not control with programming (OpenAI 3.5 still suffers some confusion and is limited by how much it keeps in memory to continue each discussion). GPT4 which is not publicly available, expands memory and lessens confusion, but will also be slightly higher cost.

At present, I am working on a rough figure of $5.50 - $7.50 for a months access per user. This is the cost I believe that the average user going at it to get the most out of it will reach per month cost wise. Some may exceed it, some may not reach it. I am hoping for a balance. I will adjust it a dollar at a time either way if required once released as an account upgrade.

Right now, the testing phase is all public. I have introduced a male therapist that will respond slightly different, being programmed to be male. I will also introduce a member only AI forum and maybe private, the same as private diaries, where only you and the AI (admins excluded), can see it, no other members can view what you discuss, so you can get deep and dirty if you want in privacy.

I am still thinking about some aspects.

For all intensive purposes, this will be a super cheap alternative to therapy. It is not designed to replace therapy, but complement therapy. It may be the only therapy that some can afford. You are accessing in essence, the smartest therapist on the planet. It knows everything psychology / psychiatry and forgets none of it. Yes, 3.5 suffers some confusion at times, but that will improve from OpenAI, and you can correct it and it will check the information provided to correct or not.

The biggest thing, is that this therapist is cheap and doesn't judge. It has no feelings towards what you say to it, its sole purpose is to help you to help yourself. It doesn't change facial expressions to make you think its judging you, or anything else. Its close to perfect, but can't do everything a human therapist can do, but can do things a human therapist can't do. It is the ideal compliment to therapy, and available for instant response 24/7. You can't sleep, your therapist is asleep, this does not sleep. You're in crisis and your therapist takes 24/48hrs to respond, this is available instantly to help you, to calm you, to walk you through distress.

My sole objective is to provide a tool so members can get the help they need. I do not make money from this, its costing will be adjusted so it pays for itself, nothing more, nothing less. What you're seeing right now, is costing about $0.60 daily with about 5 members going at it hard, another 5 have barely used it.

If you have questions, feel free to ask them here or raise a contact us to discuss.

 

[anthony]

Oh... and I'm thinking about rolling this out into other forums for limited and specific use, ie. a welcome AI therapist for introductions, that will not give advice, but will say hello, acknowledge what they have said and validate their introduction, then provide any recommendations, etc. That may tweak the annual donation cost a little to pay for it, unsure really at this point. Everything comes down to actual use and cost, and cost is something you never really know with things like this until you have existing data from use.

 

[Weemie]

I will also introduce a member only AI forum and maybe private, the same as private diaries, where only you and the AI (admins excluded), can see it, no other members can view what you discuss, so you can get deep and dirty if you want in privacy.

I think this is a fantastic idea.

 

[anthony]

Sort of like reading a textbook?

I wanted to touch on this yesterday, but time was limited.

The AI knows everything psychology / psychiatry, and by default it is not a replacement for a therapist. Saying that, you can choose for it to deep dive into your issue. You can also ask it with each response to dumb down the response as though you're 5 or 10, and it will do that.

It may give you advice that you could challenge a thought pattern with your therapist, but, you can simply ask it to help you challenge your thought process instead.

Everything you can't say to your therapist, you may to this. Everything your therapist may not say to you, this might. It is complimentary, but you choose whether you want to use it as a standalone therapist by telling it such in each post. It only remembers the last 4 posts, so the best thing you can do is tell it each post to dumb a response or to keep deep diving something with you, and with the last data set in memory, it will continue forward.

It will give you the perfect textbook answer, because it knows every text book and study, forever. Its knowledge will only keep growing as OpenAI continue feeding it data. But if you want it to answer you in a different way, just ask it, and it will try its best within its programming.

See: Personalized Plan where I asked it the same question for a 5 year old, 15 year old, then no specification, and got three different variations of response.

The more you tell it, the more specific its response. The less you tell it, the more general and likely textbook.

 

[anthony]

I am now closing this, as the service is 99% tuned, I have given members access to the general AI discussion forum where any questions and discussion can be had.