Getting a conversation started with GPT4 regarding its tokenizer, its multilingual abilities, Fodor's LOT thesis and the Sapir-Whorf hypothesis. Hi GP...
It's not. To be clear, I was claiming that GPT4's abilities to think in any language didn't count as evidence against the hypothesis. The hypothesis a...
Discussing an issue surrounding alignment and consistency with GPT4, who offers some insights regarding the future of AI system development and the co...
Hi Jamal! I apologize for leaving you hanging for so long. My primary aim in this conversation (as well as many others) was to examine GPT-4's intelli...
This is fascinating. While thinking about GPT4's multilingual abilities, and its ability to synthesize knowledge that is represented in its training d...
Testing GPT4's understanding of Gödel's incompleteness theorems. GPT4 again passes the Turing Test with flying colors. (Thanks to my friend Michel V. ...
Aiming to tap more systematically into GPT-4's emergent reflexive abilities for purpose of self-monitoring. (An approach inspired by the recent paper ...
Thank you for providing this example, and for reminding me of the hide-and-reveal feature of this forum! I should re-edit some of my older and longer ...
Making use of GPT4 as a time-travel machine and visiting ancient Rome circa 50 BC. Part 1 (In subsequent part, I intend to modify the method in order ...
Having GPT4 summarise the subtitle file from a long YouTube video, mindful of its limited attention window, and then having it reflect back on the tas...
This was a nice discussion. I enjoyed the songs it wrote, especially the bit: "Some may say I'm just a machine, a tool to be used, But I like to think...
GPT-4 is much better, that's for sure. There are specific conditions, and specific topics (mainly those that it is insufficiently trained on), that wi...
Breakthrough in the 'palindrome project' - Teaching/learning how to spell words backwards with GPT4: a machine with a brain that only processes tokens...
Hi Benkey, Reading your whole conversation with ChatGPT, I must say that the bot is way better than I had remembered it to be. I hadn't used it for lo...
How does GPT-4 knows that it is GPT-4? And, does it really know it? (Overcoming skeptical worries, part - 1) PN: Hi old pal! GPT-4: Hello! It's great ...
Do you access it though a ChatGPT Plus subscription? You must select the model from the drop-down list on each new session (dialogue) before beginning...
Wonderful! I must go to bed now. I'll be back in 20-24 hours and edit this message to comment. Did you really have this conversation with ChatGPT (bas...
Probing GPT4's creative skills with mathematical poetry. (This was inspired by an example in the recent Sparks of Arti ficial General Intelligence pap...
Unruly schoolkid sitting in nerdy Professor GPT4's classroom. (Also demonstrating GPT4's ability to push back against incorrect information when promp...
@"Benkei" Engaging in philosophical discussion with GPT-4 (explanatory-levels, emergence, AI phenomenology and operational definitions) Hi GPT4! Consi...
I discussed with GPT4 the topic of its attention window and the nature of the signification units that it contains. I don't think it's a form of perce...
It had been brought to my attention by a friend of mine a few hours ago. The proposal seems reasonable to me. The potential dangers and disruptive aff...
A super duper easy way to exploit the hitherto untapped cognitive abilities of large language models like GPT4 Hello GPT4! Here is the abstract of a r...
Hi @"Benkei", nice talking to you again! I think GPT4's (as well as ChatGPT's) bias is a consequence of the late stages of its human-supervised traini...
Pursuing research, as well as some musings on philosophical behaviorism, with a recovering AI Alzheimer patient PN: I'd now like to submit to you an e...
Playing a game with GPT4 and witnessing it morphing from brilliant research assistant to Alzheimer patient. Hi GPT4! You wouldn't happen to know how m...
Thanks for that! It's fascinating. It will be interesting, as the author notes, to see those results replicated with GPT-4, which has a significantly ...
Pushing GPT4 to its limits and exploring the source of its blind spots. Hi GPT4, I think you may be familiar with a class of logical puzzles or riddle...
During a fleeting moment, GPT-4 slips up and seems to think of itself as a human being. PN: Before too much of the present discussion slips outside of...
Discussing with GPT4 its difficulty with processing word components, and guiding it towards potential solutions. PN: In addition to the need for craft...
Thank you @"Banno", for this very relevant reference. Especially useful are the following remarks: "The training data for ChatGPT-4 includes huge numb...
Another fun tidbit. GPT4 gets creative and then immediately morphes back into Eliza GPT4: This prompt directly instructs me to reverse the words and l...
I tried to trick GPT4 and it didn't fall for it. While assuming its default personality, it doesn't have a sense of its own finitude, not even a facti...
GPT4 provides remarkable insight into the source of its own errors and then manifests a case of cognitive overload. The bot provides insight into this...
Showcasing again GPT4's astounding ability to solve problems while being directed to do so in very abstract and non-directive terms. PN: This is good....
Follow up discussion with GPT4 regarding the meaning of "I" PN: This still is not quite right but the mistakes you are making are instructive. I think...
Actually, upon rereading GPT4's latest response in the previous post I see that it still didn't get it quite right. The issue wasn't anything inherent...
Another striking example of GPT4's ability to reassess its judgements when prompted to do so with minimal human guidance. PN: I think this is a fairly...
GPT4's ability to gracefully recover from episodes of hallucination. (Continuation of the dialogue in the previous post) PN: Now that we have broadly ...
Probing the source of GPT4's self-identification ability (which led to some hallucination likely triggered by the fact that GPT4 had painted itself in...
I had some doubts whether I might not have confused this book with "La nature dans la physique contemporaine", which has also rested on a shelf for de...
I'm not sure but, as a teen I had read Heisenberg's intellectual autobiography "Der Teil und das Ganze," (in the French translation "La partie et le t...
I am not sure what the source of those tidbits of self-knowledge are, though. Maybe the correct responses regarding self-identification have been rein...
@"T Clark" For the sake of comparison, I also ran the same two questions by GPT-3.5, except I lied to it regarding the source of the original interact...
@"T Clark", I inquired with GPT4 regarding your interaction with (I presumed) GPT-3.5. Here is the result. PN: Hi GPT4, For context, a friend of mine ...
You haven't tried GPT4 yet, have you? There is a considerable (in some cases massive) improvement from GPT-3.5 (ChatGPT) to GPT4 regarding their knowl...
GPT4 comes down to earth. I asked GPT4 to produce a new version of the summary of our latest conversation that is less jargony and more likely to be u...
Exploring external cognitive enhancers and implicit self-monitoring abilities as means to develop meta-cognitive strategies that GPT4 could be prompte...
Yes another cool tidbit exemplifying how AI language models might soon be able to become self-scientific-investigators, self-design-improvers and self...
Rather that supplying another long transcript, I'm going to let GPT4 offer a short progress report from its own POV. PN: Could you produce a significa...
Comments