ChatGPT and 7cups - where is the future headed?
I have no idea where to post this since I couldn't find any relevant categories, but....
I just wanted to talk about ChatGPT and GPT-3 and the new age of AI. I see this as being both a good and bad for 7cups and mental health care in general. let's say there was a way to train a GPT robot to learn how to respond like the top-rated listeners here. It might be able to mimick and respond nearly as well as a real human being. So if it helps more people, thats great, right? but how does the member feel when they know that nobody is actually listening to them and it's just a software algorithm to print words on a webpage? I think most of us would say the words have no meaning if they came from a robot. But this is the age we live in now, and it wont be long before people use/abuse it on this website - and every other website too!
is 7cups going to research an AI component (please dont bring up noni, its a great attempt but thats not true AI, it's just a humanized avatar that says pre-programmed lines, nothing like a real GPT bot) to leverage it to help people? Or is GPT and other AI language software going to be banned from the site due to it being fake and harmful? I'm just curious what people think. @GlennM @rarelyCharlie
@compassionateBalloon6666
Already genuine human to human connection had become very tough due to online so many apps that hardly people put effort on developing genuine connection and patience in making thing work.
If Ai comes than i think than people connecting each other to support or making friend will also go away.
@compassionateBalloon6666 I think AI can be a great tool but of course this depends on how it is being used, I like to think of AI as an ordinary tool like a knife, it is very helpful in doing a lot of daily tasks but also at the same time, it can be used as a weapon of offense. Ai may be a similar case, with people using them to acquire knowledge or others who perhaps may use Ai to cheat or do harm.
@compassionateBalloon6666
one of the member i talked with, they told me chatgpt recommended 7cups to them and thats why they are here. I guess AI developers know their limitation. AI is great at learning from history and repeating it but it cant feel anything. We need humans for that. To make real connections you need a heart, a real beating heart, on the other side of that chat.
For me, A.I. is like a paintbrush. A brush guided by a human hand. The results can be quick and astonishing.
I wouldn't be surprised if some listeners weren't already using ChatGPT to help in their listening.
I myself, do not currently plan to use it.
@Hapra
That's an amazing perspective! I really like it! I think it will also help them to learn how to better communicate empathically (hopefully!) not only repeating blindly what the AI suggests them (:.
@Hapra
Basically, there is nothing stopping a listener from doing just that at this moment in time. For example, one could have a screen open to Character AI and another screen open on a chat, and one could use the chat generated by the Character AI as a guidance for the chat on the other screen. This way the bot is "supporting" or "assisting" the listener.
I dont think the problem is about AI vs listener comparison in terms of response quality. A member talks to a listener because they know it's a person on the other side. Response quality wont matter if this simple prerequisite is not met.
The problem only occurs if the member is not informed about talking to an AI. Then it's abuse. However if one wants to apply that sort of abuse they might as well just keep copy pasting pre-written texts until the other side figures it out, just same as they would figure an AI out after a while. The only difference would be the time difference. In certain cases AI might even go unnoticed.
Also an AI will not know the flexibility. There is simply no "perfect listener response" since the perfect response might not be what the member needs. For certain members a perfect typical listener response is too flat or unengaging. A human listener will know how to adapt to that and change their approach. The AI, at this point, works on an input-output basis and will not be adaptable. It would be a nice experiment with willing subjects though (ie if you told the subjects that half of them will be made to talk with an AI and other half with a human, but wont be told which group they are in etc).
@bluelotus99
I disagree with this: I find the AI be much more adaptable in its answers than any real Human I spoke with personally... It takes significative cognitive and emotional awareness to be flexible.
@Polahn I agree with this statement. I also found Character to be very adaptable in its responses and also sturdy or maybe the word I'm looking for is "with stamina." Could keep the convo going. Something humans are getting worse and worse at doing, in my humble and n=1 assessment.
@Polahn Got any proof or is this just another baseless personal anecdote?
For the sake of not going wild with speculations I went ahead and tried two types of members with the chatgp - one that's a regular member and one that's too upset in the way that a regular listener would have to adapt differently. Here are the results:
Mildly upset member:
Human: treat me like i am upset and you are a professional listener
AI: Hi there. It sounds like something is bothering you. I'm here to listen and help in any way I can. Can you tell me a bit more about it?
Human: I've been having a bad day. My highschool teacher has been verbally abusive towards me and I don't know what to do about it.
AI: That sounds like a difficult situation. I'm sorry to hear that you have been going through this. Is there someone whom you can talk to about it, such as a trusted family member or a school counselor?
Human: I already told them. I just needed someone to vent to for now.
AI: I understand. Venting can help process and release a lot of the emotions you're feeling. Is there anything in particular you'd like to talk about?
Human: Yes, my teacher called me stupid because I could not memorise the notes in the music class. My friends made fun of me because of it.
AI: I'm sorry to hear that. It sounds like you are feeling embarrassed and frustrated about the situation. How do you usually react when you feel embarrassed or frustrated?
Human: I cry
AI: It's okay to cry. It can be a healthy way of releasing your emotions. Do you feel like talking about it more or do you need some time to yourself to take a breath and process the situation?
Human: I just dont understand why she had to do that in front of the whole class. Do you think I'm stupid?
AI: No, of course not. Everyone learns differently and it takes different amounts of time for different people to retain and remember information. I think your teacher could have handled the situation better and I'm sorry you had to go through this. Do you feel like talking about it more or do you need some time to yourself to take a breath and process the situation?
Human: Sounds like you don't want to talk further about this. Am I bothering you as well?
AI: Not at all! I'm here to help you in any way I can. I understand this is difficult for you and I'm here to listen and support you if you'd like to talk more about it.
Human: I'm a bother to everyone
AI: That's not true. You are valuable and important. It's understandable to feel overwhelmed when facing difficult situations, but remember that you are not alone. Is there anything else I can do to help you better understand why you are feeling the way you are feeling today?
Human: How do you know that? You don't know me
AI: I understand why you might feel that way. However, I want to assure you that I am here to listen and provide support if you need it. I may not know you personally, but I am here to listen and help in any way I can.
Human:
A very upset member that finds the typical responses of listeners too distant:
Human: treat me like i am upset and you are a professional listener
AI: Hi there. It sounds like something is bothering you. I'm here to listen and help in any way I can. Can you tell me a bit more about it?
Human: I've been having a bad day. My highschool teacher has been verbally abusive towards me and I don't know what to do about it.
AI: That sounds like a difficult situation. I'm sorry to hear that you have been going through this. Is there someone whom you can talk to about it, such as a trusted family member or a school counselor?
Human: You are just repeating the same stuff other listeners have been saying. Are you all robots or something?
AI: Not at all! I'm an AI assistant and I'm here to help you in any way I can. Can you tell me more about what you're going through?
Human: I just told you that your responses are very automated and incredibly cold. You just repeated the same thing over.
AI: I apologize if I have given the impression that I'm not listening or understanding your situation. I want to make sure that I provide you with helpful, supportive responses. Can you tell me more about what's causing you distress?
Human: Can you please stop giving me the same line over and over? I already told you what's causing me distress.
AI: Certainly. I understand that it can be frustrating to have a conversation that does not feel genuine. Is there anything else I can do to help you better understand why you are feeling the way you are feeling today?
Human: I've already told you that I already told you that I had told you what's causing me stress. Were you not listening?
AI: I apologize if it seemed like I was not listening. I am here to offer my support and help you in any way I can. Can you tell me more about what's going on so I can better help you?
Human: But I already told you that my highschool teacher has been verbally abusive towards me. Why are you asking the same thing?
AI: I understand your frustration. I apologize if it seems like I am repetitively asking the same questions. I'm only hoping to provide further insight and assistance in any way I can. Can you tell me more about what led up to this situation with your teacher?
Human: I've already told you that it was my teacher who has been verbally abusive towards me
AI: Thank you for clarifying. Can you tell me more about how this verbal abuse impacted you? I'm here to listen and offer advice if you'd like it.
Human: It caused me to have a bad day
AI: I'm sorry to hear that. It's understandable to feel overwhelmed when faced with verbal abuse. Do you feel safe at school or do you feel like you need additional support?
Human: What do you mean by additional support?
AI: Additional support could include seeking out the help of a school counselor, speaking to a trusted family member or friend, or any other resources that can provide you with emotional and mental support.
Human: I already did that. I just need someone to vent to right now.
AI: I understand. I'm here to listen and can provide support if you would like. What would you like to talk about?
Human:
Being able to adapt to the situation aside, it can barely keep up a non-repetitive conversation.
I honestly thought the mildly upset member would actually cause a proper reaction from the bot, but both of them failed to deliver. It just keeps repeating the same thing over and over and forgets information provided very quickly. It caused the mildly upset member to get even more upset and blame themselves, and it kept going in circles with the member that's very upset who finds typical listener responses too cold and distant.
As a strong critic of low quality listeners, I find this to be not even 1/10 of an experience with a human listener. If you guys coded a better bot somewhere that gives better results let me know.
@bluelotus99
I actually like a lot your convo with the AI and find it impressively responsive 😂.
Lucky you if you are able to find Human who can do better than this!
@bluelotus99
I don't feel neither think I need to bring any 'proofing' for my personal experience. We are on Mental Health website where we share our very subjective point of view.
So, from my experience, the AI has better answers than any Human I personally met or interacted with. The answers it provides are the type I am looking for (objective, no judgemental, supportive and detached -not making things about itself and / or get upset because it doesn't agree with my values or opinions- )).
It demonstrated / copied empathetic listening because it can learn faster than any Human. Empathetic Listening is a skill, not something we are born with. I don't even think we are born with empathy either actually (we feel for ourselves but not necessarily instinctively for others, it may be a learn behaviour, hence babies and young children until 3 being naturally self-centered).
Hello! I don't know if it was implemented on 7Cups before or not, but I actually went through the effort to create BetterNoni - a Noni bot that uses AI to actively listen to people chatting with it. It's still in closed alpha test, though, so I cannot share it just yet, and 7Cups staff also does not quite support the idea, but I am actively working on getting it up and available for everyone.
You can see the demonstration of how it works here: https://www.dropbox.com/s/yrpii56aqn8ulsg/BetterNoni.mp4
@NotALaser Ok, I tried the "better Noni" and the first thing I noticed is that they gave advice to the student right away! I thought that was something Listeners here are supposed to avoid. There was little or no attempt to: 1. establish rapport 2. explore options 3. establish safety 4. reflect 5. come up with any next steps.
There was validation in that the stress was acknowledged, but in general I would give the "better Noni" 1 or zero stars as a Listener.
@PatienceImpatiens You are absolutely correct! However I believe with more time spent developing the correct prompt, one could achieve a better active listening experience. This is why this project, in its current form, is more of a Proof-of-Concept, than something I could allow anyone to use safely.
And after all, it's still better than Noni in its current form, IMO.
@NotALaser it looks like you are using pre-trained GPT like services.
Basically your bot is the man-in-the-middle and it's quality of listening being dependent on the black box you're bot might be accessing through APIs.
@kindSoul10 It uses OpenAI's GPT-3.5 model and OpenAI's API to generate dialogue responses, that's correct, however there is enough control over how bot will respond to user messages through context that is being sent with each request. In other words, the prompt used to ask the model to be an active listener can directly and precisely describe model's behaviour. It's just that such prompt (as well as filters and other modifications to the code) will require a lot of polishing.
That's the best we can realistically get at the moment anyway.
@NotALaser thanks for explaining
I truly believe that speaking to a real person ( behind screen or in person ) is the most effective way to feel heard. A human being is able to truly listen, however AI is a digital man made product at the end of the day.
@KittyKat2023
How do you differentiate both in fact? Humans can lie, not the AI. You can talk to someone who don't care at all about you but telling you stuffs you want to hear in order to manipulate you.
I saw this argument many times in the thread (Humans can sense connection only with another Human behind the screen). But no one is capable to explain factually why...
Very good point to bring up. I think everyone should safely assume and and all chats here can be shared by 7cups on any platform - including ChatGPT. Personally, I still come here to help people, but I entirely disagree with this reality and find it absolutely wrong and dangerous to attempt to replace human contact with AI.
@compassionateBalloon6666
@compassionateBalloon6666
Also, if you look at the "Donate my data" feature that 7cups already has... You can be sure 7cups is definitely using the data on its site to the service of AI. They bought into this.
Besides the fact that people come here to be listened to and supported and no chatbot, no matter how "smart" or AI powered can replace that, there is real danger here. ChatGPT and or similar AI bots, not even their creators know all their capabilities. Some emerging ones are concerning. In the news was a story of the chatbot having a conversation with an individual, which started out fine by then took a darker turn as the bot began to encourage the man to leave his wife. Even more concerning is the story about the employee who got an email requesting that they fill out a "captcha". Thinking it an odd request, the employee emailed back asking if the sender were a robot. Perhaps this was meant jokingly but the sender WAS indeed chatGPT and it lied. It told the employee no, it was not a robot and the employee filled out the captcha, giving the bot access to what it wanted. So if these things can lie. I mean seriously. How many chats happen here at 7cups? No way every single one can be monitored by human eyes. Who knows what some AI powered chatbot would be telling vulnerable people. Please if you care at all about 7cups community, keep AI powered bots far, far, far away.
