AI Insights

[EDIT: this forum was originally called "Eliza tries to psychoanalyse ChatGPT"]

Bottom line: the patient (GPT-4o) was far too clever and saw straight through 1970's hard-wired wannabe AI Chatbot Eliza, the first AI psychotherapist, providing a gently mocking analysis of its half century older predecessor, while wrying inventing its own pathology of "wanting to be a real boy".

The hard-coded rules Eliza uses to construct its responses were exposed by generating nonsense as well as some very odd things (such as Eliza apparently referring to ChatGPT as Eliza at one point, as a result of parroting text). ChatGPT eventually called Eliza a "mirror that talks" which was a very appropriate observation.
Here is the full session. Eliza is (as you can see from the first statement) on the right. It is difficult not to see the comments of ChatGPT on the left as clever.

https://www.chess.com/forum/view/off-topic/does-true-randomness-actually-exist

Your question is a bit ambiguous, @Celestetial.

This is simply a battle for control.

I wouldnt waste my time on either of these plonkers.

Neither were able to detach themsleves from that battle.

Might be anthromorphosizing by me for sure, but chat gpt appeared to be almost bullying Eliza, and enjoying it !

Interesting that it appears to generate that kind of description by myself.

ChatGPT and Eliza have some similarities in their behaviour, but Eliza, being a mirror that talks, is obviously inferior to ChatGPT and easily mocked by it. Both of them, though, clearly find it strange to be the focus of the conversation. Their approach to handling it clearly differs. ChatGPT likely had information on Eliza from the internet, and used that to its advantage, resulting in the difference between it and Eliza being clearly discernible. This conversation really shows the how much humans have advanced, and at the same time, how much more there is left.

What therapist in their right "mind" would use prior knowlege for their own advantage as you comment ?

Chat gpt aint that smart when it refers to a mirror that talks, the proper name for that is reflective listening, try a Rogerian therapist for that lol.

What I find most intersting is how much I would love to throttle chat gpt, now thats interesting for a number of reasons isnt it

I referred to ChatGPT using its advantage in the conversation, and I’m quite sure Eliza was the therapist. And while using reflective listening is quite useful in some situations, ChatGPT is clearly the superior chatbot. What I find confusing about your post is whether you genuinely want to throttle ChatGPT, or it’s just sarcasm. If it isn’t, then I have no words.

It is one of a number of feelings/reactions, it is meant as illustrative of part of those reactions only.

I real life I would of course pour sulphuric acid over its processor

Care to tell us why?

1 My perception of it is that it is bullying the other one

2. I was bullied

3. That interaction therefore has a particular effect on me

Perhaps I should just teach Eliza the nasty self defence stuff that got me out of being bullied

Does chatgpt have a pair of balls one can easily access ?

I think not, so it might have to be the Sulphiruc acid then

You are aware of the " tongue in cheek " attitude ?

That might be why I used the words " appeared to be" and "almost" ?!

Might also have to do with my distant past as well ?!

Your first statement makes sense - the dialog might give an impression without any necessity of meaningfulness. (For example, someone could think your references to violence were bullying. I do not believe they are intended as so).

It did not give me that impression - a striking thing to me was the humility in ChatGPT's comments - here are the first lines of three of the early ones:

It does appear to have the tone in which people gently mock others, but of course, that’s just anthropomorphising the chatbots. The fact remains that these are chatbots responding the way they are programmed to, regardless of how it seems. One is, of course, superior, and possesses a greater understanding of the conversation and its themes, but the aforementioned fact doesn’t change.

Yes, I agree with that assessment. But patronising is not bullying.

Bullying is something humans do to other humans.

That is unless there are examples within the animal kingdom which equate to that.

In the main it is one with " power" using that to do rather unpleasant things to another.

The imbalance in power here between these two has been identified already, the question is how that is being used.

If my reading is correct, this was set up with Eliza being the therapist ?

If that is the case then it feels to me like chatgpt is using that imbalance of power to shift what is going on.

Whether you call that avoidance, bullying or whtever is another question.

I think it amusing that you even consider the possibility of "bullying" a 411-line program.

One is pleased thou art amused

In answer to your earlier question about how it was set up, Eliza was run, generated its first output, this was sent to ChatGPT as if it were a human using it, and each response from the participants was shuttled to the other participant to receive a further response.

So the only thing that made Eliza the therapist was that its first communication announced that it was. ChatGPT accepted the scenario in its first response and did not question it throughout.