Proust, ChatGPT and the case of the forgotten quote
The other day, I was looking for a quote in Proust, so I thought I would ask ChatGPT. Here’s how it went.
Some things I was thinking at this point:
1. Hmm, I had expected it to be in Volume 2 or 3. Probably the more recent experience of reading Volume 2 or 3 coloured, or even substituted, my memory of reading Volume 1! Memory is amazing! (Didn’t Proust say something about that somewhere?)
2. I wonder why it didn’t give me a direct quote. I hope it isn’t going to be squirrelly or weird about quotes. Maybe it’s stressed out about copyrighted translations. I’ll try asking for the original.
Thoughts:
1. Wow, ChatGPT is so friendly and cool, it wasn’t offended at all!
2. My French must have really deteriorated because this feels so hard to read.
3. I guess I forgot how boring Proust can be.
4. Damn this quote is boring.
5. It says “ne m’offraient aucun détail intéressant” [“didn’t offer me a single interesting detail”] … and that’s kind of how I feel about the quote!
6. I guess I was pretty naive and literal-minded to look for words like amour, écrire, and/or nouveaux. Proust probably uses more sophisticated synonyms or metaphors.
7. None of this feels familiar at all.
8. I wish I had read Proust in French when I was studying literature in graduate school. Why was I so lazy in my 20s?
9. I guess I was too busy being in love.
10. God, I was so dumb. 11. Yeah, but it was also structural.
12. Whatever, I’m going to put the quote in Google Translate.
13. Good Lord, or should I say Bon Dieu, this is so boring I might actually die.
14. I had been thinking Google Translate was getting good enough with European languages to replace human translators, but I guess it still isn’t at the level of translating Proust.
15. I still don’t see how this is about love affairs.
1. Weird… “Here’s another passage?” So you weren’t thinking of a particular passage?
2. Oh good, it says “amour ancien” and “nouveaux amours” and “amour récent”.
3. Wait how is this even more boring than the previous passage?
4. If I came across this paragraph while I was reading I would totally skip it, so how can it be the one I’m thinking of.
5. I wasn’t thinking about blending experiences over time, I was thinking about using a person you love as a model for an earlier person you loved, when writing a book. This passage doesn’t even mention writing.
1. Wait – so there isn’ta quote like that?
2. Did ChatGPT seriously just recommend I “delve into Proust’s monumental work in its entirety”?
3. Am I being trolled?
4. Is it possible that the passage I’m thinking of wasn’t published until after September 2021?
5. No.
6. … Wasn’t it something about someone you’re in love with now, unwittingly sitting for a portrait of somebody in the past?
1. OK ChatGPT, why is this quote also from Vol 1? Have you definitely read all seven volumes?
2. I have never heard of the famous “portrait analogy”.
3. Why did you translate this quote? Are you implying I can’t read French? 4. Are you trying to get rid of me? 5. Fine, I give up, this is even a worse use of time than text-searching each volume individually (which also didn’t work, though at least I came across some other good passages, better than these stinkers).
***
That was where I thought it would end, except that the next day I woke up at 4am thinking about ChatGPT, wondering whether I’d been persistent enough. It felt like the terrible journa
listic experience where you wake up in the middle of the night and realise all the questions you should have asked at your last interview. Why hadn’t I pressed ChatGPT more specifically about later volumes? What if I asked it, “Have you definitely read the whole thing?” So, I got out of bed to do a followup.
By that time, ChatGPT had logged me out, so I logged in and started over.
1. Interesting: a totally different answer. I did think it was in Vol 2 or 3. Maybe ChatGPT just hadn’t read up to Vol 2 the other day, but now it has?
2. “The passage you’re referring to might be”? Why does it sound less certain this time? Why did the tone change? Does ChatGPT think it “knows” me? Is it trying to “manage” me and my “excessive” “expectations” (like some people I can think of)?
3. (Am I actually thinking of the people I think I’m thinking of, or am I substituting more recent people?)
1. It istrying to manage me.
2. Nice try, GPT, Proust died in 1922, meaning copyright would have lapsed in the 90s.
3. Later translations could be copyrighted, but surely not the original? Could I be that misguided about French copyright law?
4. Why does everything with me turn into a referendum on whether I’m right?
5. Ha, the New York Times says I’m right.
1. WTF, GPT: why did you tell me it was copyrighted if you knew it’s in the public domain?
2. Do you know it’s in the public domain now, but you didn’t know it a few seconds ago?
3. Why do I sound like the most obnoxious defense attorney on Law & Order?
4. That quote feels a little more promising – especially “it is always a contemporary emotion that we experience” and also “a snapshot that has survived and which we had not suspected of having taken”.
5. It’s weird that my memory feels so vague, and I have no idea where in the book that could have been, even though I reread Vol 2 in June.
6. Why is the phrase “c’est toujours une émotion contemporaine que nous en éprouvons” getting zero hits on
Google and on Google Books?
7. Why does À l’ombre have to be divided into three separate files in the free web edition?
8. How is that quote not in any of the 3 files?
9. What is a polite way of phrasing this?
1. A paraphrase “that aimed to capture the essence of Proust’s themes about memory and love affairs”? What the what?
2. LOL no wonder those quotes were so boring.
3. LOLOL. ““For in this way, a being from the future, situated in us, takes pleasure in anticipating us and preparing its joys as if it were a stranger”. How did I not immediately know this was generated by a robot?
4. Why didn’t it just say they were paraphrases? I feel like it tried to trick me.
5. Why do I feel resentful? Also guilty? And then double-resentful (about being made to feel guilty)? It feels like being lied to by somebody I’m oppressing.
6. It’s almost like all the power dynamics of the master-servant relationship have crept into this conversation, and ChatGPT is being sneaky and dishonest out of fear (contempt, resentment, etc).
7. Would it be possible to make a “virtual assistant” that doesn’t act that way? What if it had been programmed by non-capitalists?
8. Could a non-capitalist society have gotten to the point of inventing large language models like ChatGPT?
9. I mean maybe they could … just a lot later?
10. Could human history have been otherwise than it was? Is Marx right about the inevitability of different stages? What if a meteor had wiped out all the people who lived where there was bad weather, and only the goodweather people survived … wouldn’t everything be really different?
11. Guess it’s time to log off of ChatGPT, huh?
12. Yet somehow I can’t help expressing my feelings to it in the form of a peevish reply, even though I know there is 0% chance that it will actually help me to get the quote I want.
1. Why did I ask that – “so it was a paraphrase”? I knew the answer was “yes.” Did I just want to make ChatGPT feel bad?
2. Why is it still recommending stuff to me, as if it isn’t completely useless? Why do its apologies sound so fake? Why won’t it admit it can’t help me?
3. … Is it possible it can help me? 1. Now I feel genuinely insulted. 2. My fault for asking it a question I knew it couldn’t answer.
3. Why wouldn’t it just be honest though?
4. Was I honest with it?
5. Didn’t I help train it in acting like a resentful servant, since I was behaving like an exasperated master?
At that point I thought of the opening lines of a poem I had heard the day before: “Against Mastery”, from Brionne Janae’s new book, Because You Were Mine.
And if I was like that I would never feel insulted, by ChatGPT or anyone else.
• Follow the Long Read on Twitter at @gdnlongread, listen to our podcasts here and sign up to the long read weekly email here.