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?
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.
