Another interesting tidbit: PN: I think it might be fruitful, for the time being, to set aside the task of teaching you how to construct interesting p...
GPT4 as an Albert Einstein suffering an acute case of dyslexia This is how I perceived GPT4 to be in my latest interaction with it. Witness and behold...
Quite! It seems like dozens of the challenges that AI researchers had endeavored to overcome by breaking down complex cognitive tasks into component s...
Another interesting tidbit: PN: After I provided you with GPT4b's summary of the initial part of my conversation with you, is there something about me...
GPT4 may still be a dumb machine and a far cry away from having achieved AGI (artificial general intelligence). Nevertheless it is in many respects mu...
More on GPT4's proclivity to hallucinate and the limits of its (huge) working memory The following conversation (which I only provide excepts from) be...
GPT4's memory recall process and proclivity to hallucinate GPT4 has a huge knowledge base but the way it accesses it is rather similar to the way huma...
What it has done successfully is to satisfy a design requirement that met my expectations while being provided with instructions that were very abstra...
I understand. I didn't point out the error on that specific occasion because pointing that out in the past had resulted in GPT4 attempting to create a...
Georges Perec's palindrome and GPT4's anterograde anmesia Part 4 - A case of mistaken identity (continued) In the following conversation, I supplied a...
One of the most salient weakness that I saw being reported is that it limits the number of follow-up questions to three. This would make the experimen...
It is a failed attempt at producing a palindrome. The beginning (Evil rats) matches the ending (star, live) in reverse order. So, GPT4 clearly underst...
Georges Perec's palindrome and GPT4's anterograde anmesia Part 3 - A case of mistaken identity PN: I just went over another iteration in a separate di...
It already has had access to the internet for a while as its implementation in Microsoft's Bing Chat. Bing Chat is an alternative way to access GPT4 f...
Georges Perec's palindrome and GPT4's anterograde anmesia Part 2 PN: This looks good! Your ability to synthesize complex ideas amazes me. Let us there...
Georges Perec's palindrome and GPT4's anterograde anmesia Part 1 GPT4 again blew my mind so hard that my skull burst open an my brains ended up splatt...
Regarding LLMs proclivity to hallucinate or generate bullshit (in the Frankfurtian sense), I had indeed witnessed this often while interacting with Ch...
Damn! I think that the next experiment that I attempted with GPT4 was a little bit of a failure. It does again illustrate how much of a smart puppy th...
In the following exchange, I provided GPT4 with two chapters from a paper of mine on the topic of free-will, determinism and responsibility. For the s...
Part 2 PN: Your description is again correct but still doesn't match the representation that you produced. Let us try something else. Are you able to ...
GPT4 once again blew my mind. I again used the Socratic maieutic method to steer it away from dead ends, but the problems I was proposing to it were q...
Follow-up to my previous post regarding the robot reply to Searle's Chinese Room thought experiment: PN: Interestingly enough, you are replying from a...
Thanks for that! Just for fun, I asked Big Sister for her take as well (and then pressed her further on some issues). PN: Do you think that Searle's C...
I can see two reasons why that might not necessarily be the case. First, even if the designers drew explicitly on their own knowledge of the Chinese l...
So, here is my post again, processed by GPT4. I was a bit disappointed by its commentary since it's basically another instance of its oft-repeated boi...
Thanks, Banno, for providing this exchange, and interesting line of inquiry. Searle's Chinese Room thought experiment aims primarily at establishing t...
An issue that arose in my past exchanges with GPT4 concerns its ability to intuit the global structure of a problem before proceeding to breaking it u...
Yes, I think GPT3.5 had provided a better account of the source of the human error while GPT4 had provided a better account of the language model's ab...
Here is a short exchange, comparing GPT-3.5 (ChatGPT) and GPT4. This experiment is inspired by a recent Reddit thread. PN: If I am a runner in a footr...
So, here is how GPT4 reacted to my own original presentation of the parallel that I had noticed between van Inwagen's Consequence Argument and the arg...
And this one is for @"Banno" PN: Do you know the rules of the game of naughts and crosses? GPT4: Yes, I'm familiar with the game of noughts and crosse...
Certainly, in the light of those results, GPT4 could help me in such a way. I had initially though of using this formal similarity in a paper titled A...
So, in this most recent exchange, GPT4 had my mind completely blown. Being provided only with very vague clues, it was able to capture my intended mea...
Thanks for providing great questions! I need to give due thought to them before replying, and I'm a bit sick (with some sort of flu) right now. And I ...
Discussing with GPT4 a similarity between LLMs and my former philosophy teacher Richard Bodéüs. (Or, how to discursively unpack complex insights and i...
Here is yet another follow-up to my discussion with GPT4 regarding my discussion with @"sime" about GPT4 PN: "There is another issue from the exchange...
Here is a follow-up to the discussion in the preceding post. PN: 'While reacting to my intervention in the supplied exchange, you offered the followin...
I submitted part of my exchange with @"sime" to GPT4 and asked it for comments. I would have liked it to be more critical and less conciliatory (even ...
This is accurate as a low-level description of the generative model's design. But I think its ability to reason is a high-level emergent feature stemm...
Yes, this rather harks back to my discussion with it regarding the contrast between its superior cognitive skills and its comparatively poorer abiliti...
It seems to really struggle with most tasks that involve processing spatial representations. I discussed with GPT4 the mechanical structure of the con...
Follow-up on my previous post... The first part of my probing of GPT4's scholarly knowledge had yielded results somewhat below my expectations, but I ...
Probing the extent of GPT4's scholarly knowledge and discussing the neuroscience of emotions. Hi GPT4, In the present conversation I would like to ass...
I wasn't familiar with this particular metaphor from Dennett's. I'll look it up. It also evokes Leibnitz's windmill analogy and Dennett's own intentio...
This worry of yours seem to mirror a similar worry that had been voiced by @"magritte" in their post that appeared a few hours before yours. I also sh...
I agree with you, to some degree, and I surmise that GPT4 itself would likely agree even more! I am quite concerned in my investigations, and my probi...
GPT-3.5 and GPT-4 both assessing GPT-3.5's earlier performance: PN: "During a previous conversation I had given you the following instruction: "Can yo...
Now looking back at my recent exchanges with the bot regarding mathematical proofs, it looks like I had accidentally made use of the default language ...
Now, here is an interesting example of GPT4 not only failing to make stuff up but stubbornly refusing to hallucinate. PN: Could you now attempt to pro...
Comments