# opus[1m] ## Retrieval Notes - Session id: `3a650043-a887-41fc-9a60-f093372de1be` - Last activity: `2026-03-20T14:47:43.990Z` ## My Notes <!-- Add your notes here. This section is preserved across syncs. --> ## Conversation ### User opus[1m] ### User The user has opened the note file:///Users/nickyoung/My Obsidian Vault/Writing/research/generating-philosophy-text-internal-evaluation/3. Thought Experiments and Armchair Abduction.md in Obsidian. This may or may not be related to the current conversation. If it seems relevant, consider using the Read tool to examine its content. @[[3. The challenge from experience and the challenge from connecting to the world]] @3. Thought Experiments and Armchair Abduction this is a a transcript of me and my colleague discssing the most recent draft of the generating philosophy paper. sections 0 - 3. can you create a cleaned up version of the transcript please. The one with the italian accent is called enrico. please note the task is to give me a clean FULL transcript, don't cut corners, don't summarise TRANSCRIPT: The difference between the zombie case and the philosophical cases, like that. What we say May apply also to physics to certainly or and what he say can also apply to philosophy, but we can discuss that I can stuck it that way, I'll just I'll be I'm always moving around. Basically, I've got to stand up. Sit down, stop. Please show me one what? Now, it's better. What's not? I I realised that when I presented that. At the worship. The the it seems to me that the the dagrass Adam's quotation is not super clear. The role that it plays in our I wonder whether we should just drop it because I it's nice to keep it but maybe We should explain better. If it's unless all is just rhetorical, it's okay. It's just I I know what you mean. It's, uh, now because it's maybe the reader expect that there are some relevance also, to the arguement. Whereas, this is just a, it's just the idea that it's just a way of introducing, uh, in 2026 humanities reached a position in which you can actually ask machine philosophical questions. Well, another, in other words, uh, is the, the problem there is that the, the question was not asked properly. Yeah, if you, I think we talked about, did I talk to you about this? Um, Would be, is if we do another section which would be about how to prompt to get good. Exactly, that's exactly. Exactly. That was my point, this seems to be about the the importance of prompting which is not there so far. So it makes sense to keep this. We also add a section on front here. Yeah. Oh we actually just moved the entire arguement down to that section. So the entire example down to that section, Exactly, instead of having at the beginning of the paper, which seems to to give the line to the people, give the the direction to the paper and save, it may look like misleading. We can just And maybe a nice uh, a nice break to have it after a lot of heavy philosophy, instead of a change, maybe at the beginning, we just say Something like, Uh because instead of having they could, we can keep the quotation for later at the beginning, just saying the uh, guys to the Galaxy, you might ask an AI to do something else of it in 2026 Humanity as Actually reached a position in which you can ask machine, resulting questions. So something like, in the science fiction, Book novel. I checked that so you should just create a contrast rhetorical contrast between a science fiction case and what? Yeah, yeah and then we can uh, take the design fiction case again later with more detail when we be more fit into to do that good? Yeah. Yeah then There's a missing reference all the other updates, but soldiers. Doesn't has a data. Oh yeah. Page 3, crystallographic data just just a concern is usually crystallographic is not biological. So I wonder whether this is such a one now. Yeah, it's the Washington Greek examples. It's just whether this is is the right term because usually as far as I remember, crystallographic is almost the opposite of biological. So I wonder whether crystallographic data but maybe me something that is just just just yeah, yeah, yeah. It's good to, you know, these things because I don't know. Okay. When you introduce the dance and I wonder what you maybe some brackets here, where say stands or not to others, depends or not. Since I, they also stress negative. Sorry, which one was it was the phenomenon stands or not? How is that? Phenomenon depends on that. Okay. Yep, got it. Then just a few lines after two speakers on Twitter. As soon as I remember, it's not to speak as entrepreneurs. Every speakers, one speaker. But this is Here you say, we have some calls this overfitting but this is one take this to me in the last sentence, which you see either? Yeah. Yes, on D. So this is probably the user while the view that survives like becoming more elaborator, rather than more revealing, but that's so quite far. So, I think here that the anaphoric Should be. Clarify better. Yeah. Otherwise Okay, yep, I'm sorry. Yes, then the here, this this sentence which is a philosophical Corpus is not just any body of text. It is a body attack shape. I repeated judgements about whether it's arguement privileges. This is understanding of the subject. This would be, will will become clear later. We told the peer review. And, but at this stage, I found this sentence a bit enigmatic because I think a reader may just say, Because it fits better with things that come later is what you're saying? Yeah. Because that is here, geological Corpus, which is already an ocean, which is not obvious. It's not just anybody's text, it's a way to stay shape. I repeated judgements about whether it's about judgement by whom Okay, it just requires a bit more of the population because it's just one sentence, whereas later, it will explained. So, See here, and there, it risk to be obscure. And also the transition is to abroad because the, the first section ends, and then just a floor, Indian others argue that. But why we go to Florida? Sure, maybe we can say that in the introduction, but I think it's useful. Also at the end of what section having a bridge to to the next session, Okay. Then up here, there are the commas. The quotation is just like that llms is the sort of thing, llms will repeatedly do these things when you tell them not to Here is this sentence, they produce one plausible. Continuation without waiting it against Alternatives, so far, so good, but that, and they do so without what Floridian dollars calls and external feedback loop for posterior evaluation. Is not understandable unless we explain what that means, external feedback loop for posterior evaluation. Uh I wonder maybe the centres can just do without that or if we want to keep it it's a cut to the library place. Yeah, exactly. And then the part of the stuff in the philosophical Corpus. Yep. Yeah here is where the the previous emphasis explained but maybe here weak work, survives, whatever? Well, no it's okay. Also May survive or survive strong work is overlooked or maybe overlooked, okay? Or maybe maybe some exactly qualifies slightly. Then here at the end of page 7, Williamson used to have one way of characterising, but this banks on organisation to systemize, but we really say that. So all Say something like acid and bold because if if it's a book size we've already said it and then. Well, we have a yeah. It's it's a sort of summary of this. Yeah. So it's a bit odd to to have. It's here is presented something new but it's not new. So if it's a recap it's okay but we have to introduce something like as seen above. Yeah. Yeah. Yeah, and the same related for for Lipton lepton. Distinguishes, the lightest from the loveliest but where we have a whole section about that. So if it's a record, a cine above lip to distinguishes, The this a bit earlier. Before they lift on paragraph, there is a paragraph, this is a sentence as Walter and the others observe familiar form of the feasible, reading care familiar, critical questions. That's again, this sentence is not not. Yeah. No wait. So I thought what I was thinking, I want to remove Walton anyway because there's too many people in this. Paragraph anyway, I think. But anyway, I'll flag it up. The paragraph is also very long. Yeah. It's not clear. What? What means that that There's this bit all that like likeness is using two different senses. One is statistical like illness, but then the liptor sense of likeliness seems to have more to do with accuracy. So it's not something produced by a nlm is not likely if you get the next token, Slightly in the sensitive it captures something. It's aiming for truth. Yeah, exactly. In a sense, it's through what we say? But it will reward not in that. There there is another sense of likeliness. We are different from this and yeah. Also, so I think they are the Floridi way is the normal way. It Lipton uses lightliness is slightly peculiar I have the opium example somewhere in here, though. The Opium examples. Yes. It makes you sleepy. Yeah. And I, I used to work out that he thinks that's the likely because that to me, it's a totology or something like that. But he means that, no, no, that's just you've gone for truth so much that it's completely useless if he's illness. Yeah. And that's what he means by likingness. So, yeah, like that's not satistic. I like no. Exactly. Because, so this lightness is just just the property. Then, this is another thing that that it's before the leapton. It's still the same ellipson paragraph beginning. Yeah. Yeah. At the end of the paragraph. When there's also the one this one is Kendall. Walton is another Walton. It's another Walton. I think it work now. Okay. Um okay, a model trained on philosophical process, Corpus is exposed to a distribution. Systematically skewed toward the property, this theorist described this. I cannot understand what this means. Well, the last sentence of the conversation before Lipton Sorry less than the power. A model trained on this Corpus is therefore exposedly. Yeah, exactly. That could be the last pet since the paragraph, right? And let's say, repeat because What's this sentence? A model trained on a philosophical process, a corpus, from which the weakest form has been trained. And yeah, so, I just understand what, what are the properties? This Theory is described. And also don't understand who are the theorists probably yours. Okay, so again, it's sort of, yeah, the third example is okay. I think it's useful, but then it comes this sentence to conclude that. I, I don't see what two was that in me. When you say you don't see what it means do you mean I really wouldn't be able to see but you do understand what I'm going for. Do you mean? I'm just you don't understand what I'm going for. So the China charactermatical companies without being diagrammatical. Rules exposure, to enough a well-formed speech is sufficient. Let's clear everything. Yeah yeah so and so there are lamb also is exposed and seems a model training philosophical process is exposed to a That's it doesn't draw the analogy. Yeah. So what Means the distribution. Uh, toward good, philosophy. Something like that. Yeah. Just like the child is exposed usually to proper to to to Language then the they learn speaking badly. See but just I think it's just that to make I think this is stressed in a more clear in a clearer way. Would this be a clear way that I think I've missed. This is other than misunderstandingly. What I think the example I wanted to go for here was nothing with a child learning grammar. He was to do with the llm learning grammar. Just the same. So all the element route is, is training the next train to complete the next token. They're hearing and in doing so it learns how to create grammatical statuses. The idea would be something similar is going on. Like the philosophy. Yeah, that makes that's clear right? Yeah. Yeah in that sense. It's clear even with the child. Yeah. You know it's still, I don't know. It seems I had to get over complications. Yeah, probably it's better to to just use. Yeah. History is described is also the better as well. It's it's bad. Even if we replace the child with llm itself. Just for a second, just to grab a glass of water. So we ask Yeah, then There is the paragraph that starts with one may still insists. That the model is only inherited the results of other people ability level. The objection as forced in empirical science, but the reason a standard Works may depend on features of the world not exhausted by the same epic literature. Standard standard. You, you you miss a theory right? The standard is a strange expression for me. Uh, they've just Nationals the reason a theory. Yeah, yeah. Oh, a modern, something like that standard usually you are Yeah. Sorry, made a conversation to work. That was sent it. I'm sure. Then you that's my inline note of who's not how I write. But, All these. This is just a sort of problem lexical problem, so but I have more more a deeper problem with the whole paragraph which one the the ones where may still insist until That embodies them. Yeah, that's the problem. Just baby is a conclusion. It's not the idea that also make a philosophy is part of the Corpus. So The justification for the various standard are part of the Corpus. Whereas, in science. Usually, it's not not so. So yeah, probably. Okay. I mean if it's a if it's not a good paragraph then they can yeah blah blah, blah. Just just better to make a bit. Why why? How is related to also to the other things. It's a bit. Standing. Then. It's just maybe here, I don't know it's, but sure it is not since to make it a bit less assertive. Bring the review will be a defective practise but it is not. Maybe some surely designed because people may think that it's not such a good practise. Then the, the come, the lipton part, which is a bit abrupt as if some foot seats, but it seems probably if there there is a better A smoother connection otherwise is not clear. Why this this quotation from Lipton is right around here because it's Become a bit out of the blue, probably. We are introducing the quotation. Okay, and then we, we have The the right direction. Yeah, definitive version of section three. Let me find it. So to begin with, I'm not sure whether it should begin with Williamson or whether it should just jump straight into Sally. No, I see the the comments but Maybe we have some nice? Yeah. No, no, that pound it. Yeah, not part special problematic. What I found a bit more problematic is the end of the paragraph. Well, the Chinese room thing. It's uh, Rule following without comprehension that there then that's, it's a good better later. So probably at this stage anticipating that because one big thing, oh, no, Chinese Loom is exactly the the feeling that that serve has of not understanding anything while being into the the room that is uh, It's not something that this feeling cannot be found in. In the concept of rule following. And so you just said you just think maybe just a straight The site. Yes. So basically maybe from the centres beginning, the same holds the Chinese rule Yeah, up until something like Einstein's imagination, perhaps. Yeah. Okay. Something like that. Then I, I, I, I very much like the, the, the part of Austin, the philosophical crops, I think this, this part philosophy, the, the paragraph that ends with philosophy does not usually begin from wrong counter. It wasn't was already been articulated. Maybe the Chinese room case may go after that that may have even though I am still liking that, that's the problem because super is theoretical. It's not just The tutorial. That I still not see the difference between Searle and Einstein. Since in both cases, there is an intuition that has to do with with introspection or and So using maybe just one or the other. I do not that. That's I think it's the problem we are facing now unless it's me missing something, not understand. I, I I, I Not sure that we we Okay. Explain because also also for physics. Mum my mum said, okay. But maybe the eyes and, uh, experience may have been described by by others, talking about acceleration. And so, all our arguement also also gazed zombie. We have, we need more reason to explain why physics require phenomenology and philosophy. Doesn't occur because it seems that maybe one reason that I'm not sure that that philosophy also incorporate description in the training set, whereas physics maybe as a training set made only of data and here is that's interesting. But I don't know if I say, okay, so maybe if we can uh supplement the training set of physics, also with with people describing their own experience in elevators, we don't need Einstein Good. Would you mind of saying what you just said again? Just so I'm clear. I think I understand it, but I want to be sure. Yeah. I think the problem I have with all that, which I like, I think it's it's it's it's very good. Good part this, this one. It's just that that I I I not see that physics is so different from philosophy and disrespect become strong. The experience that are relevant to to Einstein, Discovery seems to experience that that can be described. Yeah, and so I think I've used the descriptions instead of, I don't see a difference between Einstein and Searl or Einstein and the other philosopher strong, uh, onto the experiments. Since they are all all relying on on introspection and on, on first person experience, Maybe the Iceland one is more embodied. The, the the Sir, one or the the trolley problem. One is less embodied. These other nice examples. The maloponti thing seems very similar to the Einstein that's embodied, okay? So it sounds to me that this section needs to be Still a little bit more both in terms of example and in terms of the idea. Yeah, okay but also also it's not just distilled which is the game more editorial. I think there is a political problem that if it's for the moment, I think we we don't give reason to think that physics and philosophy are so different and Think we have two options. One is say, okay, so we are also we are also challenging zabi at the end, okay? Uh, on the other hand, say no, no Savvy is different because there are reason why physics work differently. Okay. Now now I think I understand Wilson, okay? And one reason may be the physics if maybe one can think of the training set of physics as based only on on data and and so maybe what we may say is that if Of descriptions of. But maybe, maybe that's maybe that's the point. That's the, the good point because it's also true that it's hard because in that case you should you should add all the descriptions. Off from from because, usually in physics, you don't find description of experience but maybe non-books because on books rely on things like Einstein. And then on the other hand, you have the problem of philosophy because it's true that that you have a description of experiences, but the point of total experiments is precisely two to five to find cases that don't seem to be there in in what was described before. So, again, this seems more like Einstein, for instance, Jackson Marion the colours, uh, Training set. The cases of scientists who grew up in a black and white. Room. So again I I I don't see the difference with the Einstein case. Okay, unless you say oh no, philosophy draws a lot on On on decks. And what does physics just draw? Some Numbers. Did you have a chance to read? No, I can do that. Do you think it's worth it? I think I've been to know you think, cuz he's, he's very much like the same with the difference between physics. What science and philosophy and he's a next scientist as well. Um, I sent this to you. I considered to you again. Yeah, yeah I know I I just I'm going Receiver that I can find. Philosophy is a vocation of conceptual. Where the physical world comes in for physics or for science, versus where it comes in for philosophy. Yeah. Maybe everything that can also. I think that that's the the I think it's good because it's it's how long is the paper now? Not so long so under seven thousand I think. Okay so maybe we can still find something to say precisely if they could do that and then the paper is almost although yeah we still need to work out the well to do the Practical thing or not. I mean we can Is the final section could be. This is how you prompt. Yeah, but that's a, that is very ambitious to do. Another thing could be As they're preparing, the groundwork for showing at least alluding to how these things could be used. Yeah, exactly. Maybe we can use the the The the Adam's quotation, as a conclusion, instead of as a as an introduction. So just saying, yes, but then it may be that all we have seen is a can just lead to a situation like like that. So prompting. But prompting will be all for 100 days. If you're conclusional, I mean, that would be nice to do just because then we're almost at a finished paper. I I wonder if if you know what other people think is it and that's a bit too much of a This stuff and then you say, well, yeah, we're not going to tell you. Um, I can download what we can go for a longer paper and also this this Yeah, and I'm slowly coming up with some ways and maybe framing these ideas as well. So maybe not. I don't talk about them today because they're a bit vague. Interesting stuffing. Just actually well, I remember something interesting. I found using these things at the moment. In a way, it's trying to do. But things like magic words that you can make these things do make these things more powerful. So, I learnt somebody, and I learnt this from somebody on Twitter and he says, here's a useful phrase, the cev. So cev of an idea or of a paragraph or a passage CV, is some fairly recent acronym, which is something like coherent extrapolation volition. You say, well, here's my idea and the cev of this idea would be what the idea would be. If I'd already sold all the problems if I already had cleared up what I think I want, if everything had been clarified. I understand, everything perfectly, what actually would this paragraph look like, and I and basically this guy on Twitter said this is a hack because the the internet look. Sorry. The LM knows what this means. So if you point that phrase, CEV at something and say, give me the thing that I want even though I don't know what I want yet. It'll have a, it's very interesting. It sort of pushes it this particular way and you can kind of force it to do philosophy for you like that as well. So there's there's some things like this. I think might be interesting to talk about whether it's in this paper or the next one. Uh-huh. Yeah that's super interesting. Yeah, but I think this thing is is also very, very interesting this connection to experience and the weather experience can be sedimented in text and or we really need because I still just, for instance, here the the Right. And it is always believe about something. And so, on this creator, also noticated description of greed, not on the philosopher, present feeling of loss, that's not other would say, no, no, it's uh, you've never, Maybe through that grief is much described in literature, maybe not interest of the leader, but philosophy may draw also on novels or quinoa, but I don't know. I have the impression that some philosophers would simply say no that that that that's not how it was. It's especially if you was Innovative philosophy, the the Yeah, we can use mushroom eating but that's also, you know, super satisfying. Well, it's a probably it's, it's interesting to emature it that I I struggle because you, you liked it in the talk and I've been struggling to work this stuff in Cuz I've been working with your transcription quite a bit and you said, this is the last sorry. And I know this because the llm keeps reminding. It Enrico says, this is the last important criticism to deal with, which is a phrase you used in the talk and llms were obsessed with it. But I want to read a bit more of that book because there's some interesting stuff. Yeah, I think that that, that's where the most interesting questions, arise. If you just feeling details, it's okay. Hey, I think it's But maybe, we can say more about the fact that that But it's it's it's quite even. The the floor is the thing is Again, it's not too difficult to deal with because it it seems like just a straightforward consequence of philosophy the text because if but this is the idea. But this all the if they just explain abduction in psychological terms in the process terms, but the philosophy, the text that's all all the polarities say. So I I I I'm even That, that When we take even too much time to to deal with this because it is clearly wrong. You know, because it's a philosophing the text then who cares, whether the land is not doing, but anyway, we can keep it. It's it's it's, it's, it's it's just useful probably to introduce that but the, the real life interesting thing is seems to be the could, I mean, maybe this is too much but what we could do. So I agree the fluidity thing is led. There's less meat, there's less stuff to talk about there. One thing we could do, there would be actually to fold it into something like section one, somehow. So Section 1 will sort of talk about philosophy being in the text and that it moves on to llm's incorporating, it a bit, right? So, you could Sort of talk about floridy there and saying, look fluid really is right to say that the mechanism is entirely different to what we're normally talking about. But then this was bad. I think was better in the PowerPoint. It's the way in which we we frame and die is that okay? The first thing is because in the text, that's more the Continental stuff. But then we, we may have cases in which even if we uh accept like I think also Florida D should do because this is a sort of philosopher that the values in the text That the text cannot be cannot have value because the process is not And this is clearly the zombie point that has no problem with the fan that the the value is in the text. Yeah, it just think that if you don't type the proper mental processes, you cannot produce a valuable text. Wonder whether we can we if we find a way of framing us or Florida, the point in this way, that would be a section two more worthy, more and not just a sort of of Easy. Easy, the feasible, objection. Because if, if the objection of the way, which present we present flow, the objection is just, oh abduction is a psychological process. But we we have just shown that the value is in the text. So who cares about abductions like? So in that sense having four pages for quality? It seems too much because, but if we can frame Florality as saying, oh no, without the psychological process, you cannot, you cannot have a valuable abduction in the, that seems the interesting point by Florida did not. That abduction is, is in the mind, but that without a human abduction, you cannot, you cannot have a valuable text tax abductions. Happy birthday. Just just to be clear. Could you go one more time about how How you envisage, just reconfiguring the Floidy stuff. Just because I want to get it recorded so I can make sure I understand it properly later. Because through the Florida is. Can be interpreting two-way first just saying that abduction. It is not in the text is in the mind. But that's for for us, that's not uh especially interesting objection because in the previous section we have just said that we we focus on the text and so uh he the kind of objection he's interested in is the one in the mind showing that Delilah like that is not interesting unless We read Florida essay in that without the abduction. In the mind. You cannot add a valuable abduction in the text. And so we have to argue guys that not again. And so I think we have to change slightly the arguement in section two, okay? Because at the moment the arguement seems more. Oh no, look Florida. We are caring about the tax, not about the mind, so your arguement is strong. But because it clearly, yes, it would be bad. Even maybe just saying that, but if you can constrate it, it's not the old rule of constructing, the arguement of the opponent in the best way possible. Even better than the itself. Yeah, okay, good. And then, yes, there is. The driving thing that the the arguement is. But in that case, we can also connect better. The I think we should introduce uh, Florence in the same way. Say okay, now we have said at the end of section one, we are saying we have said that the point is not what's going on in the mind. But what's going on? Let's let's adopt say the the blank review model. Yeah. Because they say okay, even if you accept that model values in the text Floridians, say without abduction in the mind, you cannot have the abduction and good abduction in the text and obviously, without phenomenology in the mind, you cannot add good abduction in the text. Yes, yeah, that's a nice way of a nice symmetrical way. I think that's that's another paper which is just a touch of the end of the previous paper and this is what you meant when you said the presentation was a little bit more. Yes later. Okay. I remember how we thought about it before, because originally you saw section one is more. Well, actually, the introduction is saying, well, you can split philosophy into sort of things where the text is important versus things that the subject is important. We're only going to be focusing on the text one so let's ignore all of the Victory and stuff. And so then you can say, well, Even if we agree, there's nothing in principle wrong with us a non-human doing philosophy. There, here are two different reasons why we might think llms are not that sort of non-view, right? One of them. It's not real abduction. It's just path matching. Whatever two. It's a lack of phenomenological. Yes, but it's this is already there. The point is what is missing is that The reason why Florida said that is not good philosophy. It's not that. There is no real abduction. Is that the fact that there is no real abduction under psychological level prevent A real abduction in the text. So I think we should present Florida. As somebody who thinks that values in the tech, you should not use, it's too easy to use the value in the text, arguement against Florinity. To construct, Florida. They say that. You cannot values in the text but you cannot have abductive value in the text without the the proper psychological path to abduction. Gotcha, that that's the point. And that's that's that seems to me you, we have resources. You know, in part is already there that is presented in a way which it seems that Florin is wrong, just because you think that abduction is a mental process. And so you say oh like abduction, but that's not true. Because abduction is in the text whereas I think a much better is a way of presenting this position is saying no, no. I agree that abduction in the test is just that you cannot have good abduction in the test without the abduction in the mind. Good. Okay. And in this way his arguement because we say you cannot have paradigm shifts. Yes. So again, good abduction because also Palace is the best is super creative but in that case the point is The that's the best way of fighting them. Both actually. Yeah, they're choosing in the mind versus the phenomenology. Yeah, it's a nice way of putting it and what a framing board is Sort of. Allegedly necessary paths to our roots to to abduction in the text. So good. Yeah, I think that that's a crucial to have the paper unify. So, first section one, we we focus on on valuing the text and not section. One is already the title is very good, philosophy is bad section two. The title should be something like, can we have philosophy in the text without abduction in the mind? Can we have philosophine attacks without? Embodied phenomenology in the mind. Can we have philosophy in the text with our doctor in the mind? Was the first one, right? Yeah. And second the the third one, can we have philosophy in the text? That without the phenomenology in the mind because the flow of the arguement seems more fashionable. It doesn't appear to Consciousness just to a certain way of of organising information. Absolutely. Yeah. And yeah, I what I think I'm going to do. Yeah, I think the fun one is going to be trying to get section three sorted out properly because I think that's Almost ready, not quite And I was thinking about this the other night as well because They're very elegant, very good to talk about colour experience now, so very good. I've been sort of working on my getting my lovely notes out. All the colours are sort of perfect impact, they're all sort of carefully, chosen, palettes and stuff like that. And the llm is saying, oh well, if you use this shade, it's not going to pop as much as so. All of these very fine line distinctions, and that's a very nice and that's kind of a good example of phenomenology being in the text. Yeah, that's the Austin. Uh, exactly resonate with the the Austin quotation. Yeah. Engaging the, the problem with that is that it seems to this seems to apply also to physics because So so unless physics required and moreover that you have with philosophy, the problem is that the, the the, the, the real Turns. Leaps. In research seems to not just to rely on call or experience. But on a complex scenarios like Mary or the zombies which seems like Einstein and yes. Yeah, I agree and even then the Chinese room seems to then sure there are other domains in which that's less less but I think also also also alzabi would agree with that. That it depends also on the domains in which um believe that embodied is always. But for instance, the Dharma the dammit case but even though the my problem with the domat is that as presented, it doesn't seem a big Discovery seems just obvious On ingredient sense is the contribution, it makes to the third condition of a complex synthesis, which is embedded. This doesn't seem like it seems something that also, 10 years. Child can say, oh yes you sent this as a meaning of his song. But if you put it in a bigger sentence, it contributes to the meaning of the biggest sentence, you're right? So probably it's more complex but it's not easy to process or maybe it's not a good example. We need or we can elaborate more on that and make it more more. There are maybe too many examples in this video. Yeah. And Islamic one is also not especially famous college to me, I don't know. I never heard of this distinction. I will, as I read it, it's a obvious distinction. That doesn't require a fear to be to be in Maine. I'll live with you that must have appeared the most recent lln, version. That's not idea. I had nothing to do with that, okay. It's very I I think that that, that Just just see something to to fix. But yeah, this is very interesting, so good. Think if you can give more structure in the way we just said, and then, Because I was thinking also the strategy once we frame these two things, as a similar objections. So you cannot have a good text without the proper psychological process. So the the reply to Florida is no no we can because at the end, I wonder whether the our reply is more or less the same in the two cases, so it's always, that's true. You the psychological processes are sedimented in the text that they can use. So, even though the llm is doing statistics, it's a kind of Statistics that that tracks. Uh, yeah, yeah, abductive reasoning in the floor indicates and phenomenological reasoning. Uh, We can also rely on dial de esume, Oso. Be the conclusion showing that at the end there is just just one objection with you. Yeah, to yeah sites the the more functional side. And the more phenomenological side and that the, the the Emphasising. The, the importance of of the of the Corpus of the trains can allow us to deal with bodies objections because at the end in different ways because in one case, it's more a matter of adding descriptions of. Uh, so the the the phenomenological objection can be By like, all descriptions or the phenomenological process because that seems our arguement that at the end. Even the Einstein or Jackson at the end. They, they they when they have to use their own phenomenological insights, they turn that into description. So you don't just directly, they make a they call. Yeah, exactly. How we can. Yeah, so so at the end, if you can what what's really entered, the arguement is the description. So if you're ready at the description, that's okay. And from Florida instead is different because it's not a matter of having the descriptions. It's more the fact that that this these reasonings are already Terms of Comparison between possibilities are already at work in the in the Corpus and so those are statistic through the Corpus, give you the right. But in a sense, it's already there just to to bring it out. Yeah. I think to frame it better organise it better. The question is we cannot produce. Available tax without the progress, psychological process our reply. No, no, look consider how it works the land. Then you can see that even though you don't have the psychological process, you have another process that rely on, on trading, set that working a way that can give you also the right kind of abductive. Yeah, kind of overseas. Sort of functionally the same shape. Yeah, exactly. We have to show. They basically, First, we said there is the process and the text first we say, okay now that's you cannot say, oh then I cannot do philosophy because they like that. You we just have to focus on that this section. One breaking this league but then you have the the two objections. You say, oh yes, yes, let's let's focus on that. But you cannot have a good debt without, uh, the the abduction process, which is Florida, and the phenomenology process, which is, uh, Our reply is that? No, no, look, we can. We agree that they don't do that, but we can have the abductive process star. And the phenomenology process. Could that that can give you the same kind of text could? Um, I think that that's the structure of the paper that should be clarified in introduction and then use all across the paper at the moment. The one in the paper seems more patchwork of interesting ideas I know without. Well I was saying so I don't have to say this. It seems maybe interesting between these two though. So for this one we're saying something about well. So for this one we're saying something about the training data specifically about Descriptions in the data right. This one was saying something more about the mechanism itself. Yeah. Is that right? Yeah, this is. So this would be so maybe that's an easier way of dividing these things as well, right? Yeah. The general structure is the same but then the details of the replies are different. So, yeah, just reply one. One for the two one. We, we know, we need two replies to the two different objections. I'm going to take a photo of your diagram as well. Yeah, you can take it to that computer. You can take the fault. Yeah. Yeah. Well, I'll feed in the diagram, as at the beginning, everything gets fitted to the computer. Yeah. Oh you're going to talk to me about a different idea for paper? Yeah, the status of functioning I think that that's yeah. Let me reread what you said because I present we present that on in TuneIn. Invited me at the pre-technical. So I will cite our forcoming paper when we republish the the sorry, the beauty in use? No, no newspapers. No, the the engineering the aesthetic of engineering. So that's the way in which I I was thinking of presenting it very shortly. Okay. What's listen? Because these are engineering students and professors so they don't know. So that's just to introduce what is Aesthetics. Appreciation engineering. That's like I took that from our paper and This is a machine, that's all things that are basically indicator. Uh, then that's also in the paper, the appreciation of machine design looking fit for function machine, Yeah, it's it's Okay. That's I think the key, the key point So they performed the same function, but they have a different functioning and this notion of function is completely missing in the philosophy of Technology that people say just to care about fascial, conflate maybe some time function and functioning because when you say, oh, function is the characteristic behave. But the behaviour is rather function, in the function is more the, the outcome of the behaviour. So stress. This is the importance of this concept and then you say but that's not only function and functioning, it's the same. Also for for nature, because also cars on cars on on the one end, this is in system processes, but then it seems also to care more about function. This doesn't seem to at least. Yeah, yeah, I don't know, maybe about nature is a bit different, but it seems that any functional Beauty go back and defending anaesthetic of nature, which is more focused on fashion rather than fashion, despite what it says in the Nature. Just saying that in both function is clearly an ocean that is relevant in. Both domains. And we want to stress that this if we care about Aesthetics, but of organismatics biology and Because maybe the static of nature functioning. The principle for the inorganic functioning maybe for mountains. It's a bit less in that case processes, okay? But it's, whereas in small in biology that fashion is as a central nausea. So the organisms are the function of adapting to the environment and to reproduce themselves, and then the organs come to have a function, the eye the hair, that's the function of pumping blood. So, but again, the function also the the heart can be a beautiful organ because it look fit for the function of popping blood that can also be beautiful for for the Rhythm. The, the way, the way it is the way in which it is. Yeah. And the wings the same, the aesthetic of birth. So you say, oh good wings, they do free profile, I think they do, they have examples but wings are beautiful. So also, for the way, Exactly, that's the key. Uh, I think Good. And that's not. But I think it's it also to, to, to draw this, to make the things more interesting and just to avoid the repetition of the paper. Even though that paper I think will just disappear in this, which is a shame. Yes, yeah, exactly. But if we have, the general framework in which we don't care about techno, then we we also showed this the differences but we start with the general view that our function, a lot of debate functioning Overlook conflated with function itself. But from an aesthetic perspective, at least maybe it's interesting. Also another it seems interesting not only in the city, but let's let's our interest. Functioning seems, as important as a falcon or even more important. And it's also an analogy agency. For instance, in that case is activities or the process versus object functions, is more. A picture of objects? Faculty is a process so I think we can make And the aesthetic of causality, which is similar, but then we can also draw the. Then we can also draw distinction between functioning, but when there is agency, then it's different because function, then we have already the Aesthetical pages in one has already done. It brings into the Montero stuff. Yeah, exactly Montero, but in that case you have, but it seems the same thing seems very interesting because we have no agency at all is involved as in nature, or agency is just the design of functioning, but it's not the, the guy, the the subject. So the it's the sort of uh yeah. It's not agents in the this rather the, the There's probably some nice contrast cases where you have the same job being done by a person with a utensil versus yeah. Some sort of a machine where you can say there's the beauty of it being done by this person with the grace and the skill, but then there's the other type of new that's making interesting and then you have a turkeys in which just a natural phenomenon. Yeah, in that case, in one case you have the the human conscious activity which is just constitutive of the process in the technological case, the process doesn't have agency. But you still have the intentional source that design the project and then in the in the natural case you just have Yeah, so I think that that's it's I think maybe another a nice way of of We use in that material. So yeah. In a broader framework showing it to movie painting to a decent journal and definitely into an obscure collection because I did like and also yeah, exactly. We can also do something that we we, we don't do in the paper. Namely, uh, The design appreciation versus order appreciation because that seems more a case of order. Appreciation, in the case of nature, that's a custom point, but it's an older appreciation, which in which the order we see the order doing something, whereas in the static case, if you just use In a sense, is in the Danish prayer case in the agency case, but that's a strange design because it is designed in a sense of the action itself. It's a Simon idea that actional artefacts but the the engineering cases is the intermediate case between nature and because in that case you have bought design and older because design is Say. Is the source in a sense is the what has has created the order. So the order is but then you see, when you see the machine ongoing is basically order appreciation because you do not appreciating. Actions. But, but this order is the outcome of a design. So this idea of playing with the dancing, with the loss of nature, seems a mixture of design appreciation. And the older appreciation is when blog post in the junkyard about Rube Goldberg machines about what group Goldberg machines. What's that, you know, is, is the, the name of a person's Rue Goldberg, uh, These sorts of, Do you remember how the beginning of Back to the Future begins? With the sort of, the sort of the the devices. So the the hammer hits the nail, which makes the thing roll, which makes the thing fall, and you have this whole sort of This is a group called and I yeah, I just found that when has written a blog post on exact but this, maybe sounds like exactly this sort of stuff, right? It's Natural forces, but arranged and sort of. Yeah, exactly. I'll send you them one thing but that's nice. Similar with, Domino rallies as well. There's something particularly satisfying about seeing this single process spin around. But again, that that's probably for him is more the gain agency in the sense that you have a, that's that's also something we have, uh, I think the other thing we can develop are the two last points of the Paper, which we just say, oh there are these two things to develop that but we we have no time enough one is this how also that's the two freezer driving a car. Also the case in which there is still also an agency you have agency but you also have the machine because the pure case of machine is more something like the print machine in which you just press the button and leave it. Yeah. Or the Tesla machine just like on its own. In that case, it seems design is not pushing the button or starting the domino effect. It's just creating the order, just like God, creating nature and then nature starts doing his own but but Ski some agency human agency but then you have the autonomous activity, which is nevertheless designed by bye. So I think it's it's our gear level paper because it's not the aesthetic of engineering is the aesthetical functioning of which circular engineeredith is, but there are other instances and it's worth considered in different instances, comparing them and show out this National aesthetical functioning which in a sense is also, Because agency in a sense is functioning of intentional, unconscious objects. Uh, so that's I think can be a, a, very good paper. And then, yeah, the conclusion, I just used the example, the clock, the bridge, and then I, I erased the black box problem, which is also something whether this is fashiony. We can also use a game, the aesthetical functioning as appreciation, of functioning through perception and social, but we can also perceive functioning in the computer even without looking inside the computer because we can imagine it, we can have company phenomenology. So I think this does not seems also not too expensive paper because you could take a paper. We put it and we give the right. I could plug in the transcripts in a minute to see what it can. Yeah, so maybe I wonder whether this also can be a good reason. Just to drop Le Hong Kong thing in the sense that it's less risky. It's a Pity because it's nice. It is a nice room to meet on these people and also to be there and it would have been a stepping stone to my honeymoon as well. You go to Japan, we still want to go to Japan. We need to work out some dates when, when it's okay for me to leave and when it's okay for her to leave as well. Okay, so in that case, maybe you can also wait more. You can leave later, it will be won't be immediately. I think I think Natalie would like to go before the summer. Okay. But we will work something so that that's also because it's in this way. You don't have to do it. We don't count. Yeah, exactly. It wouldn't be maybe doing that too early. Yeah so having problems with your back. Yeah so probably it's a wisest choice too too. Yeah and do that online. Um, I wouldn't go to Finland, though. Did you get into Finland? No, I, I'm I I'm not, but yeah, there's also another good thing in that case because I, I, we are almost finished the, the funding, so it's so so hard to have. But in this case, if we, if we drop Hong Kong, it's much. We surely have, they're still funding for, for Finland. So, yeah. Also having this this team, then let's see. Also, what happens? We talk soon, then there, yeah, you say, but yeah, I think and moreover that's also will be more time to work on this, this, these two papers, you know, I I I'm not a huge fan of conferences. I'm much happier to be at home. I care about that, I think, because they, they work really million. And, and yeah, it would be nice to have more contacts like, for for them but there will be other occasions. Yeah, and Bink online. It's it's not like being online in general. The the the the system is perfect. It actually works. Yeah. Is going to use it personally, okay? Yeah, every civil who's going in person who Maria but you see this guy, Michelle and Christina, okay? Oh, by the way. I'm Maria Sylvia told me. I've designed the yeah. Yeah. It was nice. I told my friend that he's a designer when he was furious with me because I've caused the design have a job. Oh, and then we got this, this annoying email about the risk that how the church every time I get cut off. La, or I can do that myself but maybe if you do that, it's better. I always complain. So if somebody has complaining is better just saying please take care about that because it's it's crucial for us for our walk. So that would mean it's good to block to to interrupt our research, so, okay? I don't mind complaining. Yeah. With somebody from the recruitment a few weeks ago as well. So basically, So obviously everything was organised with my extension last year, we used the hands up in the vote and stuff like that. And then when I was in the hospital, This guy wrote to me and said, you have to sign this contract, urgently urgently urgently. And I'm why, why have we waited two and a half months and why I suddenly so, um, Sorry for complaining. Do you have anything fun to do this weekend? Work, I think. Yeah. Because next week, I will go to touring. These Engineers. Yes, we're engineers. Yes, yes. Because Verity you remember her? She I think she was there. She was the chair, I think of Turkey at the sifa. Turina. She's short black glasses. Ethics of Technology at the engineering school in Toledo. Yes, she invited me. Now, let's see. But maybe I will, I will also ### Assistant Let me read the manuscript file for context first. ### Assistant Good, I have the manuscript context. Here's the cleaned transcript: --- ## Transcript: Nick Young and Enrico — Discussion of Generating Philosophy Paper (Sections 0–3) ### Douglas Adams quotation and paper opening ENRICO: It seems to me that the Douglas Adams quotation is not super clear — the role that it plays in our paper. I wonder whether we should just drop it. I mean, it's nice to keep it, but maybe we should explain better. Unless it's just rhetorical, it's okay. It's just — I know what you mean. The reader might expect that there's some relevance also to the argument, whereas this is just a way of introducing: in 2026, humanities reached a position in which you can actually ask machines philosophical questions. NICK: Would it work if we do another section which would be about how to prompt to get good philosophical output? ENRICO: That's exactly my point. This seems to be about the importance of prompting, which isn't covered so far. So it makes sense to keep this if we also add a section on prompting. NICK: Or we actually just move the entire argument down to that section — the entire example down to that section. ENRICO: Exactly. Instead of having it at the beginning of the paper, which seems to give the reader a misleading direction, we can just — and maybe it's a nice break to have it after a lot of heavy philosophy. Instead of having it at the beginning, we just say something like — because we can keep the quotation for later — at the beginning, just saying: in the Hitchhiker's Guide to the Galaxy, you might ask an AI to do philosophy. In 2026, humanity has actually reached a position in which you can ask machines philosophical questions. NICK: So we should just create a rhetorical contrast between the science fiction case and reality? ENRICO: Yeah. And then we can take the Douglas Adams case again later with more detail, when we'd be more fit to do that. NICK: Good. Yeah. ### Minor corrections (pages 3–5) ENRICO: There's a missing reference — it doesn't have a date. NICK: Oh yeah. ENRICO: Page 3 — "crystallographic data." Just a concern: usually crystallographic is not biological. So I wonder whether this is the right term, because usually, as far as I remember, crystallographic is almost the opposite of biological. NICK: Yeah, it's the Williamson/Greek examples. It's just whether this is the right term. ENRICO: Yeah, it's good to flag these things because I don't know for certain either. Okay. When you introduce "stands or not" — I wonder whether we need some brackets here. It says "stands or not," but it depends on context. I also want to stress the negative. NICK: Sorry, which one was it? ENRICO: "The phenomenon stands or not." NICK: How is that? ENRICO: "The phenomenon depends on that." Okay. Yep, got it. Then just a few lines after — "two speakers on Twitter." As far as I remember, it's not two speakers, it's one speaker. NICK: Right. ENRICO: But here you say "we have some calls this overfitting" — but this is one take. This seems to be the last sentence which you see. Yes. So this is probably where the view of surviving texts becoming more elaborate rather than more revealing — but that's quite far back. So I think here the anaphoric reference should be clarified better. NICK: Okay, yep. I'm sorry. Yes. ### "Philosophical corpus" sentence ENRICO: Then this sentence: "A philosophical corpus is not just any body of text. It is a body of text shaped by repeated judgements about whether an argument privileges this understanding of the subject." This will become clearer later when we talk about peer review. But at this stage, I found this sentence a bit enigmatic because a reader may just say — NICK: Because it fits better with things that come later? Is that what you're saying? ENRICO: Yeah. Because here "philosophical corpus" is already a notion which is not obvious. "It's not just any body of text, it's shaped by repeated judgements about whether…" — but judgements by whom? It just requires a bit more explanation because it's just one sentence, whereas later it will be explained. So here it risks being obscure. And also the transition is too abrupt, because the first section ends and then just "Floridi and others argue that…" But why do we go to Floridi? NICK: Sure. Maybe we can say that in the introduction. But I think it's useful also at the end of the section to have a bridge to the next section. ENRICO: Okay. ### Floridi section issues ENRICO: Then here, the quotation is just like that: "LLMs are the sort of thing that will repeatedly do these things when you tell them not to." Here is this sentence: "They produce one plausible continuation without weighing it against alternatives" — so far, so good — "and they do so without what Floridi and others call an external feedback loop for posterior evaluation." This is not understandable unless we explain what that means — "external feedback loop for posterior evaluation." I wonder whether the sentence can just do without that, or if we want to keep it, it needs a gloss. NICK: Yeah, exactly. ENRICO: And then the part about the philosophical corpus. Yeah, here is where the previous emphasis gets explained. But maybe here — "weak work survives" — well, no, it's okay. But also: "may survive" or "survives"? "Strong work is overlooked" or "maybe overlooked"? NICK: Maybe some qualifiers, exactly. ### Williamson/Lipton recap issue ENRICO: Then here, at the end of page 7, "Williamson used to have one way of characterising" — but this banks on things we've already said. So it's a sort of summary. It's a bit odd to present it as something new when it's not new. If it's a recap, it's okay, but we have to introduce it with something like "as seen above." NICK: Yeah, yeah. ENRICO: And the same for Lipton. "Lipton distinguishes" — but we have a whole section about that. So if it's a recap: "as seen above, Lipton distinguishes…" ### Walton paragraph and "likeliness" ENRICO: A bit earlier, before the Lipton paragraph, there is a sentence: "As Walton and the others observe, familiar forms of the feasible reading carry familiar critical questions." That sentence is not — yeah. NICK: No, wait. So I was thinking I want to remove Walton anyway because there are too many people in this paragraph. But anyway, I'll flag it up. ENRICO: The paragraph is also very long. And it's not clear what means what. NICK: There's this bit where "likeliness" is being used in two different senses. One is statistical likeliness. But then the Lipton sense of likeliness seems to have more to do with accuracy. So something produced by an LLM is not "likely" in the Lipton sense — it captures something. It's aiming for truth. ENRICO: Yeah, exactly. In a sense. But there is another sense of likeliness that is different from this. NICK: Also, I think the Floridi way is the normal way. Lipton uses "likeliness" slightly peculiarly. I have the opium example somewhere in here though. ENRICO: The opium example — yes. "It makes you sleepy." NICK: And I used that to work out that he thinks that's "likely" because, to me, it's a tautology or something like that. But he means that you've gone for truth so much that it's completely useless. That's likeliness for him. ENRICO: And that's what he means by "loveliness." So, yeah, that's not statistical likeliness. No, exactly. Because this likeliness is just the property of — ### Last sentence of long paragraph ENRICO: Then, at the end of the paragraph: "A model trained on a philosophical corpus is exposed to a distribution systematically skewed toward the property this theorist described." I cannot understand what this means. NICK: The last sentence of the paragraph before Lipton? ENRICO: Sorry, the last sentence of the paragraph. "A model trained on this corpus is therefore exposed to…" Yeah, exactly. That could be the last sentence of the paragraph, right? But what does this sentence mean? "A model trained on a philosophical corpus — from which the weakest work has been filtered — " and yeah, so — NICK: What specifically don't you understand? ENRICO: I just don't understand what "the properties this theorist described" are. And I also don't understand who "the theorists" are — probably you? NICK: Okay. So again, it's sort of, yeah — ENRICO: The paragraph is okay, I think it's useful. But then this sentence to conclude — I don't see what it means. When you say you don't see what it means, do you mean you really wouldn't be able to see, or do you understand what I'm going for? NICK: What I'm going for is: the child learns grammatical competence without being taught grammatical rules — exposure to enough well-formed speech is sufficient. Let's clear that up. And so an LLM is similarly exposed — a model trained on a philosophical corpus is exposed to a distribution skewed towards good philosophy. Something like that. ENRICO: Yeah. Just — I think it needs to be stressed in a clearer way. It doesn't draw the analogy. NICK: Yeah. I think the example I actually wanted to use here wasn't about a child learning grammar — it was about the LLM learning grammar, the same way. All the LLM does is train to complete the next token, and in doing so it learns how to create grammatical sentences. The idea would be that something similar is going on with philosophy. ENRICO: Yeah, that makes sense. That's clear, even with the child analogy. You know, it's still — I don't know. It seems like there were unnecessary complications. NICK: Yeah, probably it's better to just use — yeah. ENRICO: "This theorist described" is also bad. It's bad even if we replace the child with the LLM itself. ### "One may still insist" paragraph ENRICO: Then there is the paragraph that starts with "One may still insist that the model has only inherited the results of other people's ability." The objection is forced in empirical science, but "the reason a standard works may depend on features of the world not exhausted by the epistemic literature." "Standard" — you're missing a word, right? "The standard" is a strange expression for me. NICK: A theory? ENRICO: Yeah. Or a model or something like that. "Standard" usually — yeah. Sorry, one more: that's my inline note of "not how I write." But I have a deeper problem with the whole paragraph. NICK: Which one? ENRICO: The one from "one may still insist" until "that embodies them." NICK: Yeah. What's the problem? ENRICO: It's just that the conclusion is not — it's not about the idea that philosophy is also part of the corpus. So the justification for various theories is part of the corpus, whereas in science usually it's not so. NICK: So probably — ENRICO: Okay. I mean, if it's not a good paragraph then — yeah. It's just better to make clearer why and how it's related to the other things. It's a bit freestanding. ### Lipton section transition ENRICO: Then — it's just maybe here, I don't know, but "peer review will be a defective practice, but it is not" — maybe make it a bit less assertive? Because people may think that it's not such a good practice. Then the Lipton part comes, which is a bit abrupt — it seems like a non sequitur. Probably there needs to be a smoother connection, otherwise it's not clear why this quotation from Lipton is right here. It's become a bit out of the blue. Probably we need something introducing the quotation. NICK: Okay. ### Section 3: Thought experiments and phenomenology ENRICO: And then we have the definitive version of section three. Let me find it. So to begin with, I'm not sure whether it should begin with Williamson or whether it should just jump straight into Zahavy. NICK: No, I see the comments, but maybe — ENRICO: Yeah. Not especially problematic. What I found a bit more problematic is the end of the paragraph — the Chinese Room thing. Rule-following without comprehension — that's better dealt with later. So probably at this stage, anticipating that — because the Chinese Room is exactly about the feeling that Searle has of not understanding anything while being in the room. It's not something that this feeling can be found in the concept of rule-following alone. NICK: So you think we should just cut from "the same holds for the Chinese Room" through to — ? ENRICO: Yes. So basically maybe from the sentence beginning "The same holds for the Chinese Room" — yeah — up until something like "Einstein's imagination, perhaps." NICK: Something like that. Okay. ### Austin paragraph and Chinese Room placement ENRICO: I very much like the Austin part — the philosophical corpus. I think the paragraph that ends with "Philosophy does not usually begin from raw encounter; it works on what has already been articulated" — that's very good. NICK: Maybe the Chinese Room case could go after that? ENRICO: It could. Even though I still — that's the problem. Because Searle is theoretical. It's not just phenomenological. I still don't see the difference between Searle and Einstein, since in both cases there is an intuition that has to do with introspection. So maybe we should use just one or the other. NICK: I think that's the problem we're facing, unless I'm missing something. ENRICO: Not sure that we explain — because also for physics — Zahavy says okay, but maybe the Einstein experience may have been described by others, people talking about acceleration. And so our argument also — against the zombie case — we need more reason to explain why physics requires phenomenology and philosophy doesn't. It seems that maybe one reason — I'm not sure — is that philosophy also incorporates descriptions in the training set, whereas physics maybe has a training set made only of data. And that's interesting. But I don't know. So maybe if we can supplement the training set of physics also with people describing their own experience in elevators, we don't need Einstein. NICK: Would you mind saying what you just said again? Just so I'm clear. I think I understand it, but I want to be sure. ENRICO: Yeah. The problem I have with this section — which I like, I think it's very good — is that I don't see that physics is so different from philosophy in this respect. The experiences that are relevant to Einstein's discovery seem like experiences that can be described. And so if we use descriptions instead of direct experience, I don't see a difference between Einstein and Searle, or Einstein and the other philosophical thought experiments. Since they are all relying on introspection and on first-person experience. Maybe the Einstein one is more embodied. The Searle one or the trolley problem one is less embodied. These are nice examples. The Merleau-Ponty thing seems very similar to the Einstein case — that's embodied. NICK: Okay. ENRICO: So it sounds to me that this section needs to be developed still a little bit more, both in terms of examples and in terms of the underlying idea. And it's not just editorial — I think there is a philosophical problem. For the moment, I think we don't give sufficient reason to think that physics and philosophy are so different. And I think we have two options. One is to say: okay, we are also challenging Zahavy at the end. On the other hand, we could say: no, Zahavy is different because there are reasons why physics works differently. NICK: Now I think I understand. ENRICO: And one reason may be that the training set of physics is based only on data. And so maybe what we may say is that if — but maybe that's the point, the good point. Because it's also true that it's hard, because in that case you should add all the descriptions. Usually in physics, you don't find descriptions of experience, except maybe in books — because books rely on things like Einstein's elevator. And then on the other hand, you have the problem of philosophy, because it's true that you have descriptions of experiences, but the point of thought experiments is precisely to find cases that don't seem to be there in what was described before. So again, this seems more like Einstein. For instance, Jackson's Mary and the colours — the training set wouldn't include cases of scientists who grew up in a black and white room. So again, I don't see the difference with the Einstein case. Unless you say: no, philosophy draws a lot on texts, and physics just draws on numbers. NICK: Did you have a chance to read Pigliucci? ENRICO: No, I can do that. Do you think it's worth it? NICK: I think so. He's very much about the difference between science and philosophy, and he's a natural scientist as well. I sent it to you — I'll send it to you again. ENRICO: Yeah, I know. I just — I'm going to find it. NICK: He's got a lot on the difference between where the physical world comes in for physics versus where it comes in for philosophy. Maybe that can also help. ENRICO: I think that's good because — how long is the paper now? NICK: Not so long. Under seven thousand words, I think. ENRICO: Okay, so maybe we can still find something to say precisely on that point, and then the paper is almost done. Although we still need to work out whether to do the practical section or not. I mean, we could — the final section could be: this is how you prompt. NICK: Yeah, but that's very ambitious to do. Another option could be preparing the groundwork for showing — or at least alluding to — how these things could be used. ENRICO: Yeah, exactly. Maybe we can use the Adams quotation as a conclusion instead of as an introduction. So just saying: yes, everything we've seen can lead to a situation like that. So prompting — but prompting will be for another paper. NICK: If it's just a conclusion, I mean, that would be nice to do, because then we're almost at a finished paper. ENRICO: I wonder if people will think that's a bit too much of a tease — this stuff — and then you say, well, we're not going to tell you. NICK: We could go for a longer paper and also — I'm slowly coming up with some ways of framing these ideas as well. So maybe I won't talk about them today because they're a bit vague. Interesting stuff though. Actually, I remember something interesting. I found, using these things at the moment, there are ways — sort of magic words that you can use to make these things more powerful. So I learnt this from somebody on Twitter: there's a useful phrase, the "CEV." So CEV of an idea or a paragraph — CEV is a fairly recent acronym that stands for "coherent extrapolated volition." You say, here's my idea, and the CEV of this idea would be what the idea would look like if I'd already solved all the problems, if I'd already cleared up what I think I want, if everything had been clarified and I understood everything perfectly — what would this paragraph actually look like? And basically this guy on Twitter said this is a hack because the LLM knows what this means. So if you point that phrase at something and say, give me the thing that I want even though I don't know what I want yet, it'll — it's very interesting. It sort of pushes the output in this particular way, and you can kind of force it to do philosophy for you like that. ENRICO: Yeah, that's super interesting. NICK: So there are some things like this that I think might be interesting to talk about — whether it's in this paper or the next one. ### Experience sedimented in text ENRICO: Yeah, but I think the connection to experience and whether experience can be sedimented in text is very interesting, because I still — for instance, here: grief. Belief is always belief about something. And so also grief — we have articulated descriptions of grief, not the philosopher's present feeling of loss. But others would say, no, no — you've never — maybe grief is much described in literature. Maybe not just in philosophical literature, but philosophy may draw also on novels. I don't know. I have the impression that some philosophers would simply say no, that's not how it works. Especially if you're doing innovative philosophy. NICK: Yeah, we can use the mushroom-eating example, but that's also, you know, super specific. ENRICO: Well, it's a — probably it's interesting to explore. I struggle because — you liked it in the talk, and I've been struggling to work this stuff in. I've been working with your transcription quite a bit. And you said — this is the last important criticism to deal with — which is a phrase you used in the talk, and LLMs were obsessed with it. NICK: But I want to read a bit more of the Pigliucci book because there's some interesting stuff there. ENRICO: Yeah. I think that's where the most interesting questions arise. If you're just fixing details, it's okay. But maybe we can say more about the fact that — but the Floridi thing is, again, it's not too difficult to deal with because it seems like just a straightforward consequence of philosophy being in the text. Because if — that's the idea. But if Floridi just explains abduction in psychological terms, in process terms, but the philosophy is in the text, then — I'm even saying we take too much time to deal with this because it is clearly wrong, you know? Because it's philosophy in the text — then who cares whether the LLM is not doing abduction? But anyway, we can keep it. It's useful to introduce, but the really interesting thing seems to be the phenomenology. ### Restructuring the Floridi section NICK: I mean, maybe this is too much, but what we could do — so I agree the Floridi thing has less meat, less stuff to talk about. One thing we could do would be to fold it into something like section one, somehow. So section one talks about philosophy being in the text, and then it moves on to LLMs incorporating it. ENRICO: So you could sort of talk about Floridi there and say: look, Floridi is right to say that the mechanism is entirely different from what we're normally talking about. But then — NICK: This was better in the presentation. It's the way in which we framed it. ENRICO: Is that — okay. The first thing is "because it's in the text" — that's more the continental stuff. But then we may have cases in which, even if we accept — like I think also Floridi should — because he's a sort of philosopher who values the text — that the text cannot have value because the process is not right. And this is clearly the zombie point: the zombie has no problem with the fact that the value is in the text. He just thinks that if you don't have the proper mental processes, you cannot produce a valuable text. I wonder whether — if we find a way of framing the Floridi point this way — that would make section two more worthy, and not just a sort of easy, easily defeasible objection. Because if the way we present Floridi's objection is just "oh, abduction is a psychological process," but we've just shown that the value is in the text, then who cares about abduction? So having four pages on Floridi seems too much. But if we can frame Floridi as saying: no, without the psychological process you cannot have a valuable abduction in the text — that seems like the interesting point. Not that abduction is in the mind, but that without human abduction you cannot have a valuable text. NICK: Could you go one more time through how you envisage reconfiguring the Floridi stuff? Just because I want to get it recorded so I can make sure I understand it properly later. ENRICO: Floridi can be interpreted two ways. First: just saying that abduction is not in the text, it's in the mind. But that's not an especially interesting objection for us because in the previous section we've just said we focus on the text. So the kind of objection that's interesting is the second one: that without the abduction in the mind, you cannot have a valuable abduction in the text. And so we have to argue against that. NICK: So the easy version is: we're caring about the text, not about the mind, so your argument is weak. But that's clearly not the best way to construct the opponent's argument. ENRICO: Right. Exactly — it's the old rule of constructing the argument of the opponent in the best way possible, even better than they do themselves. ### Paper structure: two parallel objections ENRICO: And then, the argument is — once we frame these two things as similar objections (you cannot have a good text without the proper psychological process), then the reply to Floridi is: no, we can. Because at the end, I wonder whether our reply is more or less the same in both cases. It's always that the psychological processes are sedimented in the text, which the LLM can then use. So even though the LLM is doing statistics, it's a kind of statistics that tracks abductive reasoning in the Floridi case and phenomenological reasoning in the other case. NICK: Right. ENRICO: We could also bring them together at the end to show that ultimately there is just one objection with two sides — the more functional side and the more phenomenological side. Emphasising the importance of the corpus and the training set allows us to deal with both objections. Because in the phenomenological case, it's more a matter of adding descriptions — descriptions of the phenomenological process. Because that seems to be our argument: even Einstein, even Jackson — at the end, when they have to use their phenomenological insights, they turn them into descriptions. So you don't directly — they make a — NICK: They convert it, yeah. ENRICO: Exactly. So if what really enters the argument is the description, then if you already have the description, that's okay. And for Floridi, it's different — it's not a matter of having the descriptions. It's more the fact that these reasonings, these comparisons between possibilities, are already at work in the corpus. And so the statistics over the corpus give you the right kind of output. But in a sense it's already there — just needs to be brought out. NICK: Yeah. ENRICO: I think to frame it better: the question is "we cannot produce valuable text without the proper psychological process." Our reply: no, no, look — consider how the LLM works. Then you can see that even though you don't have the psychological process, you have another process that relies on the training set, working in a way that can give you also the right kind of abductive — NICK: Kind of functionally the same shape. ENRICO: Exactly. We have to show that. Basically: first we said there is the process and the text. First we say, okay, you cannot say the LLM can't do philosophy just because of that — that's section one, breaking the link. But then you have the two objections: "Yes, yes, let's focus on the text, but you cannot have a good text without the abduction process" — which is Floridi — "and the phenomenology process." Our reply is: no, we can. We agree they don't do that, but we can have the abductive process and the phenomenology process — through the text — that can give you the same kind of output. NICK: I think that's the structure of the paper that should be clarified in the introduction and then used throughout. ENRICO: At the moment, the paper seems more like a patchwork of interesting ideas without — NICK: Well, I was going to say so myself. ENRICO: Interesting. But there's something between these two. For the phenomenology case, we're saying something about the training data — specifically about descriptions in the data. For the Floridi case, we're saying something more about the mechanism itself. NICK: Is that right? ENRICO: Yeah. So the general structure is the same, but then the details of the replies are different. We need two replies to the two different objections. NICK: I'm going to take a photo of your diagram as well. ENRICO: Yeah, you can take a photo. NICK: I'll feed the diagram in as well. At the beginning, everything gets fitted to the computer. ### Second paper idea: aesthetics of functioning ENRICO: You're going to talk to me about a different idea for a paper? NICK: Yeah, the status of functioning. ENRICO: I think that's — yeah. Let me reread what you said. I'm presenting on this in Turin. NICK: Invited you to the polytechnic? ENRICO: Yes. So I will cite our forthcoming paper — the aesthetics of engineering one. NICK: Okay. ENRICO: So that's the way I was thinking of presenting it, very briefly. These are engineering students and professors, so they don't know aesthetics. So that's just to introduce what aesthetic appreciation of engineering is. I took that from our paper. And this is a machine — these are basically indicators. Then there's the appreciation of machine design — looking fit for function. NICK: Yeah. ENRICO: That's the key point. So they perform the same function, but they have a different functioning, and this notion of functioning is completely missing in the philosophy of technology. People sometimes conflate function and functioning. When you say "function is the characteristic behaviour" — but the behaviour is rather the functioning, and the function is more the outcome of the behaviour. So I want to stress the importance of this concept. And then: it's not only function and functioning, it's the same also for nature. Because organisms perform functions, but they also have functioning. Cars, at one end — this is in systems and processes — but it seems also to care more about functioning. NICK: Yeah. ENRICO: I think any functional beauty account can also be a defence of an aesthetics of nature which is more focused on functioning rather than just function, despite what it says. And for nature — just saying that in both engineering and nature, functioning is clearly a notion that is relevant, and we want to stress that if we care about aesthetics of both organisms and biology. And because maybe the aesthetics of nature — functioning is the principle — for the inorganic, functioning may be less relevant — for mountains, it's a bit less applicable. But in biology, functioning is important. So organisms have the function of adapting to the environment and reproducing themselves, and then the organs come to have a function — the eye, the heart has the function of pumping blood. But the heart can also be a beautiful organ because it looks fit for the function of pumping blood, and also beautiful for the rhythm — the way it does it. NICK: And wings — the same. The aesthetics of birds. You say: good wings. They have a nice aerodynamic profile, they have certain properties. But wings are beautiful also for the way — ENRICO: Exactly. That's the key point. I think it's good. And it also helps to make things more interesting and avoid repetition of the existing paper, even though that paper will just disappear into this one, which is a shame. NICK: Yes, exactly. But if we have the general framework in which we don't just care about technology, then we also show the differences. We start with the general view that function has had a lot of debate, functioning has been overlooked, often conflated with function itself. But from an aesthetic perspective, at least, functioning seems as important as function — or even more important. ENRICO: And there's also an analogy with agency. For instance, in that case, it's activities or processes versus objects. Function is more a property of objects. Functioning is a process. NICK: I think we can develop that. ENRICO: And the aesthetics of causality, which is similar. And then we can also draw a distinction between functioning when agency is involved versus when it's not. Because function — we already have some of this. It brings in the Montero stuff. NICK: Yeah. Exactly — Montero. But in that case, it seems like the same thing. It's very interesting because we have cases where no agency at all is involved, as in nature; then cases where agency is just the design of the functioning but not the subject — like a machine; and then cases with full conscious agency. ENRICO: There's probably some nice contrast cases where you have the same job being done by a person with a utensil versus some sort of machine, where you can say there's the beauty of it being done by this person with grace and skill, but then there's another type of beauty when it's a machine. And then you have a third case where it's just a natural phenomenon. NICK: Yeah. In one case you have conscious human activity which is constitutive of the process. In the technological case, the process doesn't have agency, but you still have the intentional source that designed the project. And in the natural case, you just have the process itself. ENRICO: Yeah, so I think that can be a very good paper. And it's not the aesthetics of engineering — it's the aesthetics of functioning, of which engineering is one instance, but there are other instances. And it's worth considering different instances, comparing them, and showing how this notion of aesthetic functioning works. Because agency, in a sense, is the functioning of intentional, conscious subjects. NICK: So that's — I think it can be very good. ENRICO: And then for the conclusion, I used the examples of the clock, the bridge, and then I raised the black box problem — which is also about whether we can perceive functioning. We can also frame the aesthetics of functioning as appreciation of functioning through perception. And we can perceive functioning even in a computer, without looking inside the computer, because we can imagine it — we can have cognitive phenomenology. NICK: I think this doesn't seem like too expansive a paper because we could take the existing paper, reframe it, and give it the right scope. ENRICO: I could plug the transcripts into an LLM to see what it makes of it. NICK: Yeah. ### Practical matters ENRICO: So maybe — I wonder whether this is also a good reason to drop the Hong Kong conference, in the sense that it's less pressure. It's a pity because it's a nice venue and it would have been a stepping stone to my honeymoon — we still want to go to Japan. We need to work out some dates. NICK: When is it okay for you to leave and when is it okay for her to leave? ENRICO: I think Natalie would like to go before the summer. But we'll work something out. In this way, you don't have to rush. NICK: Yeah, exactly. It wouldn't be — maybe doing that conference too early. ENRICO: Yeah, and you're having problems with your back. So probably it's the wisest choice to just do that online. NICK: I wouldn't go to Finland though. Did you get into Finland? ENRICO: No, I'm not sure yet. But there's also another consideration — we're almost at the end of the funding, so it's hard. But in this case, if we drop Hong Kong, we surely still have funding for Finland. NICK: Yeah. Also having that time — then let's see what happens. We'll talk soon. But yeah, I think — and moreover, there'll be more time to work on these two papers. I'm not a huge fan of conferences. I'm much happier to be at home. ENRICO: I care about them because they work really well for meeting people, and it would be nice to have more contacts. But there will be other occasions. And the online system works well. It actually works. NICK: Is anyone going in person? ENRICO: Every — well, Maria Silvia told me. I've already seen the list. Yeah. It was nice. NICK: I told my friend that he's a designer and he was furious with me because I'd caused — designers have a job to do. ENRICO: And then we got this annoying email about the ethics review — every time. I always get cut off. I can do that myself, but maybe if you do that it's better. I always complain. So if somebody else is complaining, it's better — just saying, please take care of that because it's crucial for our work. It shouldn't have to interrupt our research. NICK: So I don't mind complaining. I spoke with somebody from recruitment a few weeks ago as well. Basically, everything was organised with my contract extension last year, and then when I was in the hospital, this guy wrote to me and said "you have to sign this contract urgently, urgently, urgently." And I'm thinking — why have we waited two and a half months and why is it suddenly so urgent? ENRICO: Sorry for complaining. Do you have anything fun to do this weekend? NICK: Work, I think. ENRICO: Yeah. Because next week I'm going to Turin. NICK: The engineers? ENRICO: Yes, the engineers. Verity — you remember her? She was the chair, I think, at the SIFA in Turin. She's short, black glasses. Ethics of technology at the engineering school in Turin. She invited me. Let's see. --- _Una trascrizione fedele restituisce alla conversazione la struttura che il parlato le aveva negato._