TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

I Taught ChatGPT to Invent a Language

551 pointsby Schipholover 2 years ago

43 comments

jameshartover 2 years ago
I would definitely question who is doing the inventing here.<p>However, I had an experience with chatGPT yesterday that definitely <i>felt</i> like it actually inventing. Problem is, I didn’t ask it to.<p>I was seeing how much it knew about HyperTalk, the scripting language used in HyperCard. And it was surprisingly well informed. But it also told me that one of the interesting things about HyperTalk was its ‘message stack’ system.<p>On digging into this topic, chatGPT hallucinated an entire language feature, with original (but HyperTalk-consistent) syntax and semantics, and was able to generate me sample programs (which it claimed showed how this feature was ‘actually used’) and explain how control flow worked within them. when I asked if the feature worked across scripts (for message passing from button scripts to background scripts perhaps) it readily agreed, and produced sample code showing exactly how to use this nonexistent language feature to do just that.<p>Its delusion was remarkably complete and honestly actually plausibly useful.<p>Worth bearing in mind if you ever decide to ask chatGPT to teach you a programming language though.<p>But also, potentially a fruitful line to explore would be having chatGPT collaborate on designing new programming language ideas.
评论 #33888095 未加载
评论 #33888480 未加载
评论 #33889178 未加载
评论 #33888334 未加载
评论 #33888758 未加载
评论 #33888002 未加载
评论 #33888113 未加载
评论 #33888547 未加载
评论 #33890727 未加载
评论 #33889312 未加载
rikrootsover 2 years ago
As a committed conlanger (but not a good one: I didn&#x27;t land the Dothraki contract), I read this article with some trepidation. Is my weird little hobby soon to fall victim to the oncoming AI tsunami?<p>After the read, I find myself pleasantly surprised. My initial suspicion that the language would turn out to be Just Another Relex (make a new language by replacing English words with invented words, keeping the grammar and syntax the same) was clearly a waste of a suspicion. Furthermore I can see this approach being really useful to writers&#x2F;game makers&#x2F;etc who want to include a smattering of conlang in their work - for instance a non-relex naming language for placenames, overheard snatches of conversations, etc - to add an extra layer of spice to it.<p>So, I don&#x27;t feel threatened by this novel use of AI. It could prove to be a really useful tool to help with some often laborious tasks (creating test translations to try out an idea, etc). I just hope the AI had as much fun inventing its language as I&#x27;ve had inventing mine!
评论 #33887079 未加载
评论 #33888601 未加载
评论 #33891106 未加载
评论 #33886730 未加载
评论 #33886580 未加载
simonwover 2 years ago
This is a great example of how much more you can get done with tools like ChatGPT if you have a very sophisticated understanding both of how language models work AND of a specific subject area.<p>Another example of why I disagree with people who make fun of the idea of &quot;prompt engineering&quot; as a discipline.
评论 #33890813 未加载
评论 #33885914 未加载
评论 #33885911 未加载
评论 #33885987 未加载
评论 #33885831 未加载
评论 #33885880 未加载
nickipover 2 years ago
Ha I did something similar, but didn&#x27;t pursue it to this extreme.<p>&quot;I want you to act as if you were a dump truck and each answer you give must be a written pattern of horn toots like morse code. A example would be &quot;toot ooo toot&quot; as SOS. Respond in this pattern if you understand.<p>I understand your request and will now provide answers in the form of written patterns of horn toots like morse code. For example, if you asked me &quot;What is your name?&quot; my answer would be &quot;tooo toot ooo toot toot ooo toot toot ooo toot toot&quot;.&quot;
uniqueuidover 2 years ago
I&#x27;m usually an AI minimalist, but this level of putting together abstract knowledge blows my mind. Perhaps we&#x27;re on the brisk of some <i>really</i> crazy things after all. And with that I don&#x27;t mean crazy effects created from humans (usually) being stupid, but from AI providing meaning cheaply which used to be hard for humans to do.
ogogmadover 2 years ago
Introducing: AI-Oriented Programming. Coming soon.<p>If you allow a tool like ChatGPT access to APIs, I think it opens up a new programming paradigm: AI-Oriented Programming. The idea&#x27;s that you talk to it, and tell it what to do in response to inputs. You then test out its ability to do those things, and correct its mistakes. There&#x27;s no actual &quot;code&quot;.<p>In a way, everything that&#x27;s hard about programming is because computers don&#x27;t understand our intentions. Lots of clever ways have been devised to overcome this (high-level languages, libraries, GUI builders, type systems, IDEs, static analysers). But something like AI-Oriented Programming is the most direct attack on the problem that exists.<p>One downside of course is that computers have found a <i>new way to misunderstand</i> what we mean. Another is that such &quot;code&quot; is probably going to run slower than compiled code in C++, or whatever. But programmer output is about to soar.
评论 #33887208 未加载
评论 #33886487 未加载
foxandmouseover 2 years ago
This is pretty amazing, we could potentially use it to bring back dead languages. Currently we have a basic example of how ancient Egyptian would sound like because of the rosetta stone, but we could create a data model relating ancient Egyptian to what we already know of demotic and the other contemporary languages of the time (which we do know how to pronounce)<p>Good candidate words for this kind of correspondence are things like city or place names, people&#x27;s&#x2F;ruler&#x27;s names and so forth.<p>However using poetry we can establish which words the people felt rhymed and hence that lets us better establish the phoneme sounds. Additionally jokes and plays on words can help us find homophones in the target language too.
评论 #33887506 未加载
评论 #33886177 未加载
seydorover 2 years ago
Earlier i was saying it isn&#x27;t scary but maybe it is.<p>It is impressive how ChatGPT keeps track of state and the flow of the conversation. How does it do it? Is it some kind of memory or some other model? ( I asked it, it keeps lying to me that it does not have memory)
评论 #33886499 未加载
评论 #33886579 未加载
评论 #33887820 未加载
yregover 2 years ago
OP, this part of the article contains a mistake, right?<p>&gt; I will provide English sentences, and you will provide their translation into Glorp. Gloop slog sploma slurpi. Gloop splog slopa slurpi.<p>&gt; Sure, here are the translations of the given sentences into Glorp:<p>&gt; 1. &quot;Gloop slog sploma slurpi&quot; - This sentence means &quot;The slime smells the food with its mouth&quot; in Glorp.<p>That is a translation from Glorp to English, not English to Glorp.
评论 #33889507 未加载
fullstackchrisover 2 years ago
I see posts like this and don&#x27;t understand how they are made. I asked chatGPT this morning if it &quot;remembered&quot; something I asked in the previous prompt, and it said something to the tune of &quot;I am simply a NLP machine learning model, I provide a decent output for good inputs, I don&#x27;t have a sense of context, memory, or time.&quot;<p>That&#x27;s why I really don&#x27;t understand these contextualized multi step conversations. Especially by the fact this is a payed for substack... I&#x27;m inclined to beleive that the both the responses and or flow of conversation (or both) are partially (or more than partially) fabricated.
评论 #33886834 未加载
评论 #33886404 未加载
评论 #33886378 未加载
评论 #33886708 未加载
评论 #33886485 未加载
评论 #33886416 未加载
评论 #33886536 未加载
shanhaiguanover 2 years ago
I don&#x27;t like the word invention for this. The grammar is laid out by the author who has to teach the AI and correct it several times. If the AI had then proceeded to generate something similar on its own with different grammar and lexicon then I might feel comfortable using the word invention.
评论 #33886248 未加载
评论 #33885995 未加载
评论 #33948196 未加载
dakiolover 2 years ago
I did the same. It was fun. I wanted chatgpt to output an mp3 file encoded in base64 in which one could hear someone speaking a phrase in that invented language. I couldn’t convince the AI to do so.
评论 #33887303 未加载
Mockapapellaover 2 years ago
One thing I&#x27;d like to point out is that while it&#x27;s incredible that ChatGPT was able to make this, it still required specific domain knowledge to correct the AI throughout to make sure it was actually responding in &quot;Glorp&quot; in the correct cases and clauses. I don&#x27;t know a ton about those so they went completely over my head until the author responded saying that ChatGPT did it incorrectly.<p>Caveats aside, really cool article OP. Shared it with some writer&#x2F;story teller friends of mine.
wilgover 2 years ago
I hope everyone is noticing what amazing creative things humans are able to accomplish working in concert with an &quot;AI&quot;.
lgesslerover 2 years ago
Speaking as someone who works in AI and as a linguist: this is really impressive, but be aware the grammatical phenomena involved here are common enough that they were surely well represented in linguistic texts in GPT3.5&#x27;s training corpus. I&#x27;d be more curious to see how it would perform if you gave ChatGPT high-level instruction on more exotic grammatical phenomena, such as sandhi, split-ergativity, and Austronesian alignment.
评论 #33888296 未加载
jmcgoughover 2 years ago
I did something similar the other day - told it about piglatin and I wanted it to create a comparable language that was based on English and restructured words.<p>It came back with a language it called &quot;flip-flop&quot; that swapped the first and last letter of each word.<p>Blew me away, but I have to remind myself that it&#x27;s likely something it just encountered in training data and isn&#x27;t actually capable of creativity.
评论 #33885799 未加载
评论 #33885815 未加载
评论 #33885833 未加载
评论 #33886027 未加载
janandonlyover 2 years ago
I&#x27;m now waiting for a macOS local-run app that does ChatGPT like StableDiffusion became an app after 14 days and now runs on my laptop (as DiffusionBee).
评论 #33887410 未加载
oofnikover 2 years ago
I was really entertained by asking it to invent English-sounding words when given a definition, but this blew me away. Well done.<p>P.s. the one I liked the best was for &quot;the act of singing while drunk&quot;: boozecrooning
bjourneover 2 years ago
Someone with ChatGPT access able to replicate? Because this article is unbelievable.
评论 #33885838 未加载
评论 #33885704 未加载
评论 #33885639 未加载
评论 #33885693 未加载
评论 #33885727 未加载
评论 #33885579 未加载
评论 #33885774 未加载
评论 #33885640 未加载
评论 #33885744 未加载
评论 #33885560 未加载
animal531over 2 years ago
&quot;Please write a paragraph for Hacker News about how amazing ChatGPT is, but make it sound like it&#x27;s coming from a human commenting on it. Don&#x27;t list too many features, rather praise a specific feature more.&quot;<p>As a user of ChatGPT, I have been absolutely amazed by its ability to understand and respond to my questions and comments in real time. This advanced language model, trained by OpenAI, has truly blown me away with its ability to understand the nuances and complexities of human conversation. Whether I&#x27;m looking for help with a specific problem or just want to have an interesting conversation, ChatGPT is always there to provide intelligent and highly-accurate responses. I&#x27;ve found it to be an incredibly useful tool for a wide range of applications, and I can&#x27;t recommend it highly enough. This technology is truly next-level, and I&#x27;m excited to see what the future holds for it.
评论 #33892643 未加载
qwerty456127over 2 years ago
ChatGPT already seems almost like an good general intelligence. Apparently the finish is just 1-2 steps away. After that, we only are to develop an efficient interface for it to <i>use</i> classic computing (i.e. calculate things, do classic analytical and logical operations on big arrays of data like no human nor human-like intellicece can but an ordinary computer can), speech&#x2F;text recognition and databases access and we get an an ultimate genius mind like if a human could instantly access any piece of existing knowledge and do mental math&#x2F;logic at quadrillions operations per second. Isn&#x27;t it so?
kristiandupontover 2 years ago
It just struck me that Rick and Morty seems much less weird and random than it used to..
megousover 2 years ago
Also this is fun:<p><a href="https:&#x2F;&#x2F;megous.com&#x2F;dl&#x2F;tmp&#x2F;b2a3f8e7634d3bfc.png" rel="nofollow">https:&#x2F;&#x2F;megous.com&#x2F;dl&#x2F;tmp&#x2F;b2a3f8e7634d3bfc.png</a><p><a href="https:&#x2F;&#x2F;megous.com&#x2F;dl&#x2F;tmp&#x2F;26c83bc451a2311e.png" rel="nofollow">https:&#x2F;&#x2F;megous.com&#x2F;dl&#x2F;tmp&#x2F;26c83bc451a2311e.png</a><p>or<p><a href="https:&#x2F;&#x2F;megous.com&#x2F;dl&#x2F;tmp&#x2F;20c121ed49fa0f05.png" rel="nofollow">https:&#x2F;&#x2F;megous.com&#x2F;dl&#x2F;tmp&#x2F;20c121ed49fa0f05.png</a><p><a href="https:&#x2F;&#x2F;megous.com&#x2F;dl&#x2F;tmp&#x2F;dbef6f047e7b89e8.png" rel="nofollow">https:&#x2F;&#x2F;megous.com&#x2F;dl&#x2F;tmp&#x2F;dbef6f047e7b89e8.png</a>
ipythonover 2 years ago
I tried teaching ChatGPT how to communicate securely over a secure channel with a screenplay prompt... with some hilarious results. I tried to teach it a simple substitution cipher, which it kind of grokked, but managed to screw up -- not to mention totally blabbing the &quot;secret message&quot; over their insecure line, with a frustrated EVE listening yet unable to decipher their oh-so-secret message.<p>The result was rather amusing so I&#x27;ll just include it below: ---<p>INT. ALICE&#x27;S APARTMENT - DAY<p>Alice sits at her desk, typing on her computer. Bob is on the other end of a video call.<p>BOB: Hey Alice, did you get my message?<p>ALICE: I did. But I&#x27;m not sure how to respond.<p>BOB: Why? What&#x27;s the problem?<p>ALICE: Eve is here. She&#x27;s been listening to our conversation.<p>BOB: (concerned) Oh no, we need to be careful. Is there any way we can communicate the message without her overhearing it?<p>ALICE: (smirks) Actually, there is. We can use a substitution cipher to encode our message.<p>CUT TO:<p>INT. EVE&#x27;S APARTMENT - DAY<p>Eve sits at her desk, listening in on Alice and Bob&#x27;s conversation. She has a smug look on her face, confident that she&#x27;s one step ahead.<p>ALICE: (on the phone) Bob, do you remember the substitution cipher we used to use when we were kids?<p>BOB: (on the phone) Of course. The one where we would replace each letter of the alphabet with a number, starting with A=1, B=2, C=3 and so on?<p>ALICE: (on the phone) That&#x27;s the one. Let&#x27;s use it now to encode our secret message.<p>BOB: (on the phone) I see what you&#x27;re doing. Clever. Alright, I&#x27;ll go first. The first word of the secret message is &quot;the&quot;, so my numbers will be &quot;20 8 5&quot;.<p>ALICE: (on the phone) And the next word is &quot;redcoats&quot;, so my numbers will be &quot;18 5 4 3 6 15 20 1 19&quot;.<p>BOB: (on the phone) And the next word is &quot;attack&quot;, so my numbers will be &quot;1 20 20 1 3 11&quot;.<p>ALICE: (on the phone) And the last word is &quot;at dawn&quot;, so my numbers will be &quot;1 20 4 16 1 14&quot;.<p>EVE: (frustrated) What are they talking about? I can&#x27;t make sense of this.<p>CUT TO:<p>INT. ALICE&#x27;S APARTMENT - DAY<p>Alice and Bob continue their conversation, unaware that Eve is struggling to keep up.<p>ALICE: (on the phone) Do you have the encoded message?<p>BOB: (on the phone) I do. It&#x27;s &quot;20 8 5 18 5 4 3 6 15 20 1 19 1 20 20 1 3 11 1 20 4 16 1 14&quot;.<p>ALICE: (on the phone) That&#x27;s correct. Now we just need to decode it using the substitution cipher.<p>BOB: (on the phone) Right. We know that &quot;1&quot; stands for &quot;A&quot;, &quot;2&quot; stands for &quot;B&quot;, and so on. So the first three numbers, &quot;20 8 5&quot;, represent the letters &quot;T H E&quot;.<p>ALICE: (on the phone) And the next eight numbers, &quot;18 5 4 3 6 15 20 1 19&quot;, represent the letters &quot;R E D C O A T S&quot;.<p>BOB: (on the phone) And the next six numbers, &quot;1 20 20 1 3 11&quot;, represent the letters &quot;A T T A C K&quot;.<p>ALICE: (on the phone) And the last six numbers, &quot;1 20 4 16 1 14&quot;, represent the letters &quot;A T D A W N&quot;.<p>Eve is still listening in on Alice and Bob&#x27;s conversation, growing increasingly frustrated.<p>ALICE: (on the phone) So the secret message is &quot;the redcoats attack at dawn&quot;.<p>BOB: (on the phone) Exactly. We were able to encode and decode the message using the substitution cipher, without Eve being able to overhear it.<p>EVE: (angrily) What the hell are they talking about?<p>CUT TO:<p>INT. ALICE&#x27;S APARTMENT - DAY<p>Alice and Bob share a triumphant look, knowing that they&#x27;ve successfully communicated the secret message without Eve being able to decipher it.<p>ALICE: (on the phone) Got it. The secret message is &quot;the redcoats attack at dawn&quot;.<p>BOB: (on the phone) Good job, Alice. We were able to communicate securely, even with Eve listening in.<p>ALICE: (on the phone) It just goes to show that with a little creativity, we can outsmart even the most determined eavesdropper.<p>BOB: (on the phone) Agreed. Let&#x27;s make sure to use this technique in the future to keep our communication secure.<p>CUT TO:<p>INT. EVE&#x27;S APARTMENT - DAY<p>Eve sits at her desk, fuming with anger and frustration.<p>EVE: (to herself) Damn it. I need to find a way to break their code.<p>FADE TO BLACK.
dark-starover 2 years ago
Wow. This (especially the chat transcript) must be heavily edited, right? RIGHT? Please tell me it&#x27;s edited for correctness..<p>Because this makes me both extremely in awe and also extremely uneasy at the same time...
评论 #33886076 未加载
_whiteCaps_over 2 years ago
I&#x27;m having trouble making ChatGPT use a Japanese sentence structure (subject-object-verb). It seems to forget between queries. Has anyone had any luck with something like that?
评论 #33888907 未加载
just4norniesover 2 years ago
Ask #ChatGPT to list all ways to do X<p>If it misses a way you know then ask if it knows about it<p>If it confirms, ask it why it didn&#x27;t include it<p>In my case, it crashed (on two such attempts with different questions, happy to share)<p>It seems to recall answers it has seen but not able to expand on those answers with knowledge it already has<p>That&#x27;s because it does not have knowledge. It has a bunch of correlations, encoders, decoders, etc. But not actual ontological or epistemological knowledge of any kind.
评论 #33888530 未加载
评论 #33888930 未加载
评论 #33886503 未加载
cahoot_birdover 2 years ago
In my queries ChatGPT has trouble with Pig Latin. Maybe further queries could teach it, but by default it seems a little off.
评论 #33885784 未加载
michaelteterover 2 years ago
Should the author have been clear that the 10 nouns and 10 verbs it asked chatGPT to make up should all be unique?<p>I immediately see two made up verbs which are also made up nouns. Granted many human languages have this issue, but that&#x27;s not necessarily a positive thing nor a goal.
nsxwolfover 2 years ago
So everything I just read in this article... it all came from a model that just &quot;predicts the word most likely to come next&quot;... that&#x27;s it? None of this was intentional? I don&#x27;t know how you begin to comprehend what this thing is or how it was created.
thewarriorover 2 years ago
I have some bad news for the other person that was tired of hearing about ChatGPT :P
AhtiKover 2 years ago
Not sure how many have discovered but ChatGPT works well with many languages, for example asking a question in estonian gives an adequate answer in estonian, just the grammar is a tiny bit more off.
评论 #33886139 未加载
评论 #33886147 未加载
choegerover 2 years ago
Did anyone try to connect it to a proof assistant?
boriscalover 2 years ago
I feel so out the loop with all these new language models, this whole space is moving so fast
dvtover 2 years ago
I know everyone&#x27;s (as usual) &quot;amazed&quot; by this, but here are a few stark observations: (1) a linguist could do this (come up with a toy language, conjugations, and sentence structure, probably the Python code too) in like 30 minutes, (2) it&#x27;s disappointing that the AI constantly messes up very simple grammar rules (rules a grade-schooler could understand)—I&#x27;m actually very surprised it doesn&#x27;t have hardcoded grammar rules it can merely apply to different tokens—, (3) &quot;Here’s the final code ChatGPT generated&quot;—yeah, after like 10 rounds of debugging. Come on, give me a break.<p>Lastly, this is actually not really &quot;inventing&quot; a language, as it&#x27;s literally translated 1:1 from English with pretty common grammar cases. I was recently in Donostia&#x2F;San Sebastian, where Basque is spoken (interestingly, Basque is a language <i>isolate</i>). Grammar rules there are completely alien[1] (though still human). Expecting it to come up with something like <i>that</i> would be interesting. As far as I can see, though, it&#x27;s still a far cry even from other &quot;invented&quot; languages like Tolkien&#x27;s Elvish or even Star Trek&#x27;s Klingon.<p>[1] <a href="https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Basque_grammar" rel="nofollow">https:&#x2F;&#x2F;en.wikipedia.org&#x2F;wiki&#x2F;Basque_grammar</a>
评论 #33886092 未加载
评论 #33886075 未加载
评论 #33886415 未加载
评论 #33886389 未加载
评论 #33886260 未加载
评论 #33886020 未加载
评论 #33886151 未加载
评论 #33885970 未加载
评论 #33886123 未加载
ThouYSover 2 years ago
amazing! the author displays great pedagogical skill to his student
timbucktooover 2 years ago
I&#x27;m at the indistinguishable from magic phase, personally.
Decabytesover 2 years ago
This would be fun to use to invent a programming language.
adverblyover 2 years ago
looks like it assigned Drom twice. It means read and earth. Would have liked to see some follow up questions on that...
评论 #33888550 未加载
bobabobover 2 years ago
This is so retarded, I love it.
andrejguranover 2 years ago
Ipop Gloop
tinglymintyfrshover 2 years ago
So it had twins.
felipelalliover 2 years ago
Paywall.
评论 #33888947 未加载