Hegel vs. GPT-3

post by Bezzi · 2021-10-27T05:55:18.296Z · LW · GW · 21 comments

Contents

  I
  II
  III
None
21 comments

The good is the idea, or unity of the conception of the will with the particular will. Abstract right, well-being, the subjectivity of consciousness, and the contingency of external reality, are in their independent and separate existences superseded in this unity, although in their real essence they are contained in it and preserved. This unity is realized freedom, the absolute final cause of the world. Every stage is properly the idea, but the earlier steps contain the idea only in more abstract form. The I, as person, is already the idea, although in its most abstract guise. The good is the idea more completely determined; it is the unity of the conception of will with the particular will. It is not something abstractly right, but has a real content, whose substance constitutes both right and well-being.

— Georg Wilhelm Friedrich Hegel

 

I bet this quote will sound vaguely familiar to most readers. I'll help you: it was used by Scott Alexander in his Nonfiction Writing Advice as an example of entirely unreadable abstract paragraph.

Like Scott, I won't advice any human to end up sounding like Hegel. But GPT-3 isn't human, so I prompted it with the above excerpt (using the Eleuther UI). I will remark that that there was no cherrypicking. I genuinely ran the prompt just twice with default parameters on the same input, and reported here the generated gibberish.

Two of the following paragraphs are GPT-3 output. The other one is the actual following paragraph from the Philosophy of Right. Now you can play this little game:

Easy mode: Carefully re-read the initial paragraph, check it against each one of the following, and guess which one is from actual Hegel.

Hard mode: You can't re-read the initial paragraph (if you didn't read it at all, read it now, then go for a 5-minutes walk and return). Trust your short-term memory, examine each of the following and guess which one is from Hegel.

I

It is a will, whose unity is the idea, which has become actual, but not as a power of some abstract thing in itself, but as the idea in itself. To be the good means to be the idea, that is, freedom. It means to be at home in one's right, and, in this sense, it is to be at home everywhere. And in this sense, it is a power.

II

In this idea well-being has value, not independently as the realization of the separate particular will, but only as universal well-being, as universal, that is, in its essence, intrinsically or in accordance with freedom. Hence, well-being is not a good, if separated from right; nor is right a good, if separated from well-being.

III

What is properly the idea here is not the universal but the particular; but the latter in turn is already contained in the former as a more abstract expression, since the concept of will is already contained in the concept of being able to will. The Idea is the absolute final cause of the world. It is an abstract possibility, which has no determinate existence in the world and can be realized only in an individual.

 

Did you spot the original? If not, you can read it directly from here.

 

Update after 3 days: This exercise turned out to be simpler than I expected. On one hand, this community seems to include a disproportionately large number of people who've actually read some Hegel. On the other hand, it was possible to spot the correct paragraph without even taking into account the meaning of the text. Nice catch, Richard_Kennaway!

Until now, we [LW(p) · GW(p)] have [LW(p) · GW(p)] four [LW(p) · GW(p)] comments [LW(p) · GW(p)] explaining their methodology, and I am happy to see that the correct paragraph was found through many different considerations I hadn't thought of. I am now wondering if the absence of wrong guess is due to real ease of this exercise for the average LW user, or at least partially to selection bias (eg people who guessed wrong didn't leave a comment).

21 comments

Comments sorted by top scores.

comment by alexgieg · 2021-10-27T12:34:49.497Z · LW(p) · GW(p)

... this quote ... was used by Scott Alexander in his Nonfiction Writing Advice as an example of entirely unreadable abstract paragraph.

It isn't unreadable. Hegel is arguing with concepts from previous philosophies which he presumes the reader already knows and understands well. If one begins reading him possessing the prerequisite knowledge one can understand him just fine. Besides, this is a point in the middle of a long discussion, so he already presumes the reader understood the previous points and is connecting the dots.

Great philosophers are great because they notice something no one has noticed before and are thus the very first person in History to try and express that. They have no tool for doing so other than everything that was said before, which, by definition, doesn't include what they're trying to say. So, on top of trying to say something utterly, absolutely novel, they must invent the language and semantics with which to say it by repurposing words and concepts that aren't appropriate for the task. Eventually (measured in decades to centuries) students of that philosopher figure out better ways to express the same novel notions he pioneered, and cause the learning curve to become less and less steep. In the extreme this is so well done, and that philosopher's ideas and terminology gain such widespread adoption, that language itself adapts to the way the philosopher used it. And then everyone is talking from within that philosopher's terminology, and wondering, when they read the original work, what was the big deal with someone who was all about stating, and badly at that, mere truisms.

If philosophers wrote presuming their readers have no philosophical knowledge at all, and under the requirement that all words they use must retain their current, commonsensical meaning, every sentence of theirs would balloon into an entire book. The philosopher would die of old age before having presented 1% of what they wanted to say.

Replies from: ChristianKl, BossSleepy
comment by ChristianKl · 2021-10-27T21:55:22.771Z · LW(p) · GW(p)

It isn't unreadable. Hegel is arguing with concepts from previous philosophies which he presumes the reader already knows and understands well. If one begins reading him possessing the prerequisite knowledge one can understand him just fine.

Did you actually try reading Hegel in the German orignal or spoke with someone who tried that instead of reading an English interpretation that's more like an interpretation of Hegel by doing things like splitting one sentence into multiple one's as in the opening quote?

comment by Randomized, Controlled (BossSleepy) · 2021-10-27T17:11:39.751Z · LW(p) · GW(p)

Which texts is Hegel responding too? Is it ultimately rooted in Aristotle/Plato/Socretes? How much work does one have to do to get up to speed?

I can see the argument for the language style as being fundamentally rooted in the difficultly of trying to say something fundamentally new, but being oracular can also give something memetic fitness -- perhaps at some point philosophers realized they could get status by moving more towards poetry or riddles? I've also read some contemporary philosophers who seem to take pains to write in extremely clear ways -- when I read Bostrom, Parfait, or Focault or listen to Amanda Askill or Agnes Callard or Amia Srinivasan I don't get the sense that they're necessarily trying to bring fundamentally new objects into our ontology or metaphysics, but rather that they're trying to clarify and tease apart distinctions and think through implications; if so, is that a project that tends to lend itself to a really different, "clearer" way of using language?

Replies from: alexgieg
comment by alexgieg · 2021-10-27T17:53:10.954Z · LW(p) · GW(p)

Which texts is Hegel responding too? Is it ultimately rooted in Aristotle/Plato/Socretes? How much work does one have to do to get up to speed?

I'm not well versed in Hegel's philosophy, but I know he does three things (and probably more).

First, he builds upon Kant, who himself is moving against all philosophy that came before him and refunding the entire thing so as to be compatible with modern scientific inquiry.

Second, he changes the concept of truth, from static to dynamic, not in the sense that what we think is true may be wrong and so we fix our knowledge until it becomes actually true, but in the sense that the very notion of "truth" itself changes over time, and hence a knowledge that was true once becomes false not because it was incorrect, but because it's aligned with a notion of truth that isn't valid anymore. This comes on the heels of a new analysis methodology he invented for this purpose, and that you need to master before seeing it in use.

Third, he tries to integrate notions of justice, rights etc. that are still grounded on pre-Kantian notion with all the above.

That paragraph quoted touches on all of the above, so it takes a knowledge of classic metaphysics, plus Kantian anti-metaphysics, plus classic political philosophy, plus Hegel's own take on words such as "truth", "rights" etc. actually refer to.

It's an extremely ambitious project, and on top of that he has to deal with the potential censorship of rulers and church, so even in parts in which he could be clearer he has to deliberately obfuscate things so that censors don't catch up with what he's actually trying to say (this was a usual procedure for many philosophers, and continues being among some).

(...) when I read Bostrom, Parfait, or Focault or listen to Amanda Askill or Agnes Callard or Amia Srinivasan I don't get the sense that they're necessarily trying to bring fundamentally new objects into our ontology or metaphysics, but rather that they're trying to clarify and tease apart distinctions and think through implications;

I don't know the last three, but the first two basically go in the opposite direction. They take all these complex novel notions of the genius philosophers and distillate them down into useable bits by applying them to specific problems, with some small insights of theirs sprinkled here and there. Foucault in particular also did some of the "big insight" thing, but on a more limited fashion and with a narrower focus, so it isn't as earth-shattering as what the major philosophers did.

Besides, there are movements among professional academic philosophers that propose developing philosophy in small bits, one tiny problem at a time worked to exhaustion. Much of what they do is in fact this. But how that's seen varies. When I was majoring in Philosophy in the 2000's, for example, there was an opinion shared by all professors and teachers in the Philosophy Department that from all of them who worked there since it was founded in the 1930's until that date, only one single professor has been seen as a real philosopher. Everyone else were historians of philosophy, which indeed was how they described what we were learning how to do. :-)

is that a project that tends to lend itself to a really different, "clearer" way of using language?

Yes, undoubtedly. On the flip side, it doesn't lend itself to noticing large scale structural issues. For instance, from working tiny problem by tiny problem, one after the other, one would never do as Hegel did, stop, look at things from a distance, and perceive the very concept of truth everyone was using is itself full of assumptions that need unpacking and criticizing, in particular the assumption of the atemporality of truth. Rather, they will all tend to keep working from within that very concept of truth, assumed wholesale, doing their 9-to-5 job, accumulating their quotations so as to get a higher pay, and not really looking outside any of it.

A rule of thumb is that major philosophers make you feel ill. They destroy your certainties by showing what you used to consider solid ground were mirages. Minor philosophers and professional philosophers, in contrast, feel safe. At most a little inconvenient here and there, but still safe, since with them the ground is still the same, and still mostly as firm as before.

Replies from: Chris_Leong, avturchin, BossSleepy
comment by Chris_Leong · 2021-10-28T03:16:12.809Z · LW(p) · GW(p)

"Second, he changes the concept of truth, from static to dynamic, not in the sense that what we think is true may be wrong and so we fix our knowledge until it becomes actually true, but in the sense that the very notion of "truth" itself changes over time, and hence a knowledge that was true once becomes false not because it was incorrect, but because it's aligned with a notion of truth that isn't valid anymore."

I don't suppose you could give a rough idea of why he sees the notion of truth as changing?

comment by avturchin · 2021-10-27T23:46:49.022Z · LW(p) · GW(p)

I have an impression that Hegel was the first who predicted that humanity will eventually evolve in what we now call "superintelligence". He suggested that human evolution represents stages of the evolution of the spirit, which eventually will become God.

He also described something which reminds me Big Bang in the beginning of his "Science of logic", where "being" and "nothingness" are combined into evolving becoming. 

comment by Randomized, Controlled (BossSleepy) · 2021-10-27T19:21:15.473Z · LW(p) · GW(p)

Thanks!

First, he builds upon Kant,

Do you have any recommendations for how to start with Kant?

Replies from: jessica.liu.taylor
comment by jessicata (jessica.liu.taylor) · 2021-10-28T02:37:39.621Z · LW(p) · GW(p)
  1. Read Descartes's Meditations (representative of Enlightenment rationalism).
  2. Read Hume's A Treatise on Human Nature (representative of Enlightenment empiricism, responded to directly by Kant).
  3. Read Kant's Critique of Pure Reason (synthesizing rationalism and empiricism).

Here's a comment I wrote [LW(p) · GW(p)] relating some of Kant's basic arguments to Solomonoff induction, it might help as a starting point.

comment by ChristianKl · 2021-10-28T10:14:49.832Z · LW(p) · GW(p)

Did you spot the original? If not, you can read it directly from here.

The original is not in English. If you take the first three sentences of the opening quote, they match to one sentence in the the German orignal that Hegel wrote. It removes all the fun of reading Hegel where you have to think about how the subclauses of a sentence relate to each other.

The original for it is:

Das Gute ist die Idee, als Einheit des Begriffs des Willens und des besonderen Willens, in welcher das abstrakte Recht, wie das Wohl und die Subjektivität des Wissens und die Zufälligkeit des äußerlichen Daseins, als für sich selbständig aufgehoben, damit aber ihrem Wesen nach darin enthalten und erhalten sind, - die realisierte Freiheit, der absolute Endzweck der Welt.

Get's turned in English into:

The good is the idea, or unity of the conception of the will with the particular will. Abstract right, well-being, the subjectivity of consciousness, and the contingency of external reality, are in their independent and separate existences superseded in this unity, although in their real essence they are contained in it and preserved. This unity is realized freedom, the absolute final cause of the world.

In German, you have do decide whether to read that as:

Das Gute ist die Idee, ((als Einheit des Begriffs des Willens und des besonderen Willens), in welcher das abstrakte Recht, (wie das Wohl und die Subjektivität des Wissens und die Zufälligkeit des äußerlichen Daseins), als für sich selbständig aufgehoben, damit aber ihrem Wesen nach darin enthalten und erhalten sind,) - die realisierte Freiheit, der absolute Endzweck der Welt.

Or:

Das Gute ist die Idee, ((als Einheit des Begriffs des Willens und des besonderen Willens), in welcher das abstrakte Recht, (wie das Wohl und die Subjektivität des Wissens und die Zufälligkeit des äußerlichen Daseins, als für sich selbständig aufgehoben), damit aber ihrem Wesen nach darin enthalten und erhalten sind,) - die realisierte Freiheit, der absolute Endzweck der Welt.

Reading Hegel in English is essentially reading him in easy mode.

comment by Chris_Leong · 2021-10-27T15:28:49.842Z · LW(p) · GW(p)

I was able to correctly identify it without even reading the initial paragraph. I have some brief exposure to Hegel's philosophy having read A Very Short Introduction to Hegel and having read a few philosophical articles about him. Nonetheless, I haven't read him directly and I don't have any knowledge of his philosophy of right and wrong. One of the paragraphs had a claim the wrong way round and the other paragraph seemed kind of fluffy.

However, if I'd been shown the original paragraph, then I might have had more trouble differentiating it. That's not surprising - presumably Scott chose that paragraph for a reason.

comment by Richard_Kennaway · 2021-10-27T07:30:57.360Z · LW(p) · GW(p)

The giveaway for me was nothing to do with the meaning (if any) of the text, but the antimetabole in the second sentence of II. I have not observed GPT-3 to make such exact structures.

Replies from: Chris_Leong
comment by Chris_Leong · 2021-10-27T15:24:24.964Z · LW(p) · GW(p)

How did you hide it?

Replies from: KaynanK, Pattern
comment by Multicore (KaynanK) · 2021-10-27T18:47:50.032Z · LW(p) · GW(p)

https://www.lesswrong.com/posts/2rWKkWuPrgTMpLRbp/lesswrong-faq-1#How_do_I_insert_spoiler_protections_

Replies from: Chris_Leong
comment by Chris_Leong · 2021-10-28T03:07:42.207Z · LW(p) · GW(p)

Thanks

comment by Pattern · 2021-10-28T17:04:01.303Z · LW(p) · GW(p)

>!

Except instead of hitting enter after >! at the beginning of a line, like I did, hit space.

comment by Mitchell_Porter · 2021-10-27T10:14:39.935Z · LW(p) · GW(p)

Hegel seems to be saying that one can become an incarnation of a moral ideal. I imagine quite a few people in the rationalist and effective altruist communities have attempted this... 

comment by Pattern · 2021-10-28T17:01:13.666Z · LW(p) · GW(p)
Like Scott, I won't advice any human to end up sounding like Hegel.

Unless this is regional variation in English, it's advise or advize. (The second one may trigger 'autocorrects'.)


Two of the following paragraphs are GPT-3 output. The other one is the actual following paragraph from the Philosophy of Right. Now you can play this little game:
Easy mode: Carefully re-read the initial paragraph, check it against each one of the following, and guess which one is from actual Hegel.

It'd be an amusing trick if GPT coughs up (actual) Hegel when prompted with Hegel.

Replies from: Bezzi
comment by Bezzi · 2021-10-30T08:59:55.264Z · LW(p) · GW(p)

Unless this is regional variation in English, it's advise or advize. (The second one may trigger 'autocorrects'.)

Uh, no regional variation, it was just my first typo on LW.

Fixed. I've also included a small update with some final thoughts.

Replies from: Pattern
comment by Pattern · 2021-11-01T18:45:57.040Z · LW(p) · GW(p)
Update after 3 days: This exercise turned out to be simpler than I expected. On one hand, this community seems to include a disproportionately large number of people who've actually read some Hegel. On the other hand, it was possible to spot the correct paragraph without even taking into account the meaning of the text. Nice catch, Richard_Kennaway!
Until now, we have four comments explaining their methodology, and I am happy to see that the correct paragraph was found through many different considerations I hadn't thought of. I am now wondering if the absence of wrong guess is due to real ease of this exercise for the average LW user, or at least partially to selection bias (eg people who guessed wrong didn't leave a comment).

I got it wrong.

I don't remember what my wrong guess was though, or why. Just that it looked like Hegel (based on what I read of the paragraph, which wasn't a lot) - i.e. nonsense. (Which could be because it's translated.)

comment by jessicata (jessica.liu.taylor) · 2021-10-28T02:31:48.286Z · LW(p) · GW(p)

I. "The idea is freedom": weird. "At home in one's right": doesn't parse. "To be at home everywhere": why would one's right be everywhere? "It is a power": overly trivial. Probably not Hegel.

II. It seems to be saying: Well-being as a property of isolated individuals is not valuable; well-being as part of an integrated whole is valuable. "Right" is a universal good, "well-being" is an individual good, in line with normal moral philosophy. It's a coherent thought, the different sentences relate to each other and are compatible with each other. Probably Hegel.

III. Particulars are more abstract expressions? That's backwards. Saying the Idea can only be realized in an individual contradicts Hegelian monism, as well. Probably not Hegel.

(I've read some amount of Hegel but not closely)

comment by Multicore (KaynanK) · 2021-10-27T18:46:47.787Z · LW(p) · GW(p)

I quickly disqualified III because of its inconsistent capitalization of "idea", which doesn't seems like something Hegel would do. From there I noticed that I is a completion that continues about vaguely the same sort of things as the first paragraph, while II specifically focuses on the things mentioned right at the end of the first paragraph. I wasn't sure whether GPT or Hegel was more likely to choose the vague completion, or the specific completion. I ended up guessing correctly after scrutinizing the first paragraph, but I was uncertain about my guess.