Discussion of Blog Post: The Continuing Vitality of Epicurean Physics

This thread is for discussion of the blog article:

Blog Article

The Continuing Vitality Of Epicurean Physics

[legend='Superseded:','#b8860b']

This entry has been superseded. Please read the current version here.

[/legend]

Introduction: A Story About the Sun That Is Really About Everything

In a recent scholarly essay, T.H.M. Gellar-Goad takes up one of the most ridiculed positions in the history of ancient philosophy: the Epicurean claim that the sun is the size it appears to be. Cicero mocked it. Stoic philosophers used it as evidence that Epicureans were intellectually unserious. Modern critics have…

Cassius

April 27, 2026 at 9:52 AM

1 - Do you input those texts and papers and "query" that as it's training model? Yes that's exactly how I am starting. I am telling it specific papers to review (mainly Sedley but also others) and I am also building a list of "instructions" on things to avoid and things to prefer, and I have also fed it many of the papers/pages/blog articles which we've prepared in the past and told it to use these as the base for its writing (rather than just letting it make its own way as to what positions to favor.

2 - Do you ask it specific queries section by section? Pretty much yes here too, but I am more providing it an outline of the structure of the document and what I expect it to say. Now of course it would be useless if it just told me what i wanted to hear, so I often give it particular citations and directions for each section.

3 - Do you simply use it for grammar and syntax or to come up with "turns of phrase"? Many of the "turns of phrase" are come up with by it. It seems like it has a style of writing (which I also tell it to be firm or not-so-firm) that does seem like an advanced form of grammar-checking. It doesn't provide "grammar" so much as it provides "manners of arguing"

4 - In a way it's just going for and compiling citations, but it's far beyond that. It's very hard for me to distinguish between what it is doing and having a "real person" compose the argument.

Of course one of the keys here is that at least in most cases I know the material, know what to expect, and can recognize if it botches a quotation, so I have pretty firm quality control over it for that reason. Of course if i were asking it about brain surgery I would have no clue how to watch it for errors.

I have another one close to ready to go which will go specifically through the opposition to geometry. It's even more specialized than this one, but the argument is very logical and understandable and even if some detail of a person gets botched, it's still very possible to understand the logical force of the general argument and know that it's solid even if a detail is botched.

I've come near a line today however. I decided I would take advantage of it to go back and study this issue of what Aristotle said against Democritus, and how Epicurus adjusted his views of atomism to account for that. Years ago I had read some articles on that by Farley but it didn't make impression on me at the time and I've never had time to dig into Aristotle far enough to even find the references, much less understand them.

So the third article that will come will address Epicurus' advocary for "minlmal parts" of the atom. After reading what Claude has compiled I think I finally see the issue and how indeed Epicurus did make significant advances beyond Democritus on that part of atomism. The key seems to be that Aristotle strung together four or five highly logic-based objections to Democritus that - if not dealt with - were held to be persuasive against atomism. After all, Aristotle was a pretty smart guy and he rejected atomism.

Claude was able to find the citations for each of those Aristiotle objections, explain their meaning in pretty plain language, and then explain how Epicurus answered them. The key to the answer is that they all require reference to "parts" of an atom and not consider atoms to be a monolithic blob, as Democritus "apparently" had done.

So this excursion has proven to be very helpful to me in addition to producing a useful article. (And that's the point of the output right now - to get some reasonable articles out there where we've never been able to get anything for ten years.) But on the "minimal parts of the atom" in particular, i think we're going to find that this will prove to be an illustration of Epicurus really getting head-to-head on the logic games, which is making me more confident that when Sedley says (inferential basis of epicurean ethics article) that the division of feeling into pleasure and pain is much influenced by dividing nature into matter and void.

Anyway now I am rambling but the basic point is that if you feed in the authorities you want it to summarize, and you also issue specific instructions as to the outline of the article, and you have enough command of the subject to know that the article is what you want it to say, it can produce very good output. Not ideal - the ideal in my case would be to find a time machine, go back to college and spend a career specializing in classical philosophy and physics, and write every word myself. That's not going to happen, so we have another back and forth about "the perfect" fighting with "the good" and how to reconcile them.

Closing on this point Don: As you yourself have said, there would be very little benefit in flooding the world with AI generated pamphlets. But here we are using it to pull together research on matters that we largely understand, and we can take that and use it on the podcast and in discussions to advance real live conversations.

Don here is an example - something to which you can personally relate given that you are the star of the result. I will probably polish this further and come up with a graphic for the heading and move this to substack eventually, but while we work through what all this means, take a look at this when you have time and let me know what you think.

Great points Dave. I am struggling with some of the same questions.

I think it could EASILY trim the length and tune the grade level to the intended audience.

Already I have run into the question in regard to what to link from the Epicurustoday.com home page. for example the buttons on the very front page point to Canonics, Ethics, and Physics sections, but I don't really want to dump people straight to the full versions, so currently they point to a condensed version of the longer article, which exists in the "analysis" section of the site.

Now that's not ideal either, because I don't really want people who go to the full analysis thinking that they are rereading what they've already read, but then again I don't want to offer only the condensed version. Same thing would apply in similar fashion to crafting particular articles for particular reading levels.

Thanks Martin!

Quote from Godfrey

VS26: Understand that a long discourse and a short one both achieve the same result.

That's one that always has seemed to me to be in need of clarification. i think I've seen some analysis of it by commentators to the effect that the real meaning is that the purpose of both is to obtain the same result. It seems obvious that the meaning cannot always be: "use few words" or Epicurus would not have written 37 books on nature of so many articles. There are times when one or the other is appropriate, and the long will definitely not do the job when short is appropriate, not the short when the long is needed.

Patrikios while I occasionally use Grok and ChatGPT, these artilcles I have been working on have all been Claude. I "suspect" but cannot confirm that the reason I am finding Claude's output so useful is that I have now spent several months pointing it to, asking about, and uploading many of our past articles, plus much material from Dewitt and Sedley and others which i think are the most perceptive. I am combining that with a lengthy series of instructions about which perspectives on issues seem to me to make the most sense. For example, just feeding it the list of 15 principles on the front page here, plus some number of academic articles with which I agree, etc. is what is prompting it to produce such good drafts. I definitely have to read each word, however, because even against my strict instructions it will still fall back into views that I consider to be wide-accepted but wrong.

For example I am sure that it would be completely happy to produce an article saying that Epicurus merits little more than a footnote in history and that the real genius was Democritus, and than goodness we have Aristotle so Ayn Rand could base her arguments on him and we could all be Objectivists. It appears to me that Claude or any other AI engine is going to give you what you ask for - within limits - especially if there is a body of work out there which agrees with that opinion.

So while there are obvious very great dangers with AI, I don't think in the end that it can ever replace a strong human editor who has and end-goal in mind. No doubt the AI programmers code their own preferences into the system, but AI doesn't "care" about what it is producing unless it violates one of those hard-coded rules.

We as the authors taking responsibility for the output have to guarantee that it is worthwhile. AI doesn't care about us, and I continue to agree with those who criticize AI that in the end - I don't care what AI's opinion is either. It's a tool but it is no substitute for a human direction.

Getting back to your initial question, I suspect it's entirely possible that a similar investment in Grok or chatgpt or any other could produce similar results. And I am sure that at some point there's a wall to hit as to storage and other costs involved in particular platforms. I'm not paying for anything more than a "basic" tier of service at this point. And no matter how much storage and how many data points you load into the system, there are always going to be conflicts between the sources, so ultimately you have to be sure it follows reasoning that you yourself are willing to stand behind.

All this is fascinating and really is a brave new world.

Darn I have not had time to read that article. I just glanced at it and it appears pretty substantive. Maybe needs a thread of its own? If needed I'll separate it out.