Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GPT-4 Designed a Programming Language

lukebechtel.com

Post reply on HN
Print view View original

121–130 of 338 posts

Re: GPT-4 Designed a Programming Language

#121

Post by Gigachad »

kaczordon wrote: post #14
HervalFreire wrote: post #11

quotation from the link: "LLMs in the GPT family have read every programming language in the world, billions of times. It's been known for some time that these agents can write small snippets of code with some limited success. However, to my knowledge, none of these agents have ever been asked to make their own programming language. This type of task is a bit more complex, and requires a bit more creativity and fores…

Future data will still have to come from someone e.g. a person.

I suspect the future will involve AI performing its own research with the ability to take measurements and make observations.

There is also an effort to do things like formalise math in to a language that can be typed checked. Then you ask the AI to prove a statement is true using the language. As soon as it type checks, you know you have a valid proof. Some new data was just created.

Re: GPT-4 Designed a Programming Language

#122

Post by sangnoir »

p1necone wrote: post #82

The example code it generated for this "new" language is almost entirely syntactically correct Typescript.

>> Indentation-based scoping, similar to Python.

Proceeds to add brace-scopes :-|.

The breathlessness of these "ChatGPT N does X" posts is annoying. I know it's an instance of the usual hype cycle which has been going on for a while "Doing X, but with (hyped tech) Y" where X ∈ [Todo app, DB, shell, CLI took, ...], Y ∈ [Go, Rust, ChatGPT n]

Re: GPT-4 Designed a Programming Language

#123

Post by HervalFreire »

worrycue wrote: post #59
Jensson wrote: post #35

Earlier quoted context omitted.

> Because some of the things it gets right can not be attributed to anything else other than the concept of "understanding". Or just large dataset. The new thing we got was parsing natural languages, then we did a markov chain based on that so that it outputs semantic correct follow-ups based on what people on the internet would likely do/say in similar situations, and you get this result. It is very easy to see that…

Arguing with AI (especially LLM) proponents is beginning to feel like arguing with a religious person ... it's almost easier to just let them believe in their god. Pointing out the failures of their favorite LLM to prove to them that it's not doing what they think it's doing just falls on deaf ears as they go digging for more "proof" that ChatGPT actually understands what it's saying.

>Pointing out the failures of their favorite LLM to prove to them that it's not doing what they think it's doing just falls on deaf ears

It's not falling on deaf ears. It's because it's stupidity to think the "failures" are proof.

Should I point out all the failures in human intelligence? Humans fuck up all the time. Humans make stupid mistakes, assumptions, biases, errors in reasoning and leaps in logic all the time.

According to your logic that's proof that humans don't understand anything.

Re: GPT-4 Designed a Programming Language

#124

Post by marviel »

wruza wrote: post #116

I still don’t get why service, actor and rpc must be language keywords, when they aren’t semantically bound to a remote provider’s signatures. Why there’s actor needed at all? Rpc in place of async? If you have an rpc protocol implemented under `rpc`, why not just make an interface named PersonService and specialize a connection with it? It doesn’t actually make any sense.

I think the choice could be made a few different ways for sure.

The interesting thing about the rpc object, that it didn't explore in this instance explicitly (but I must assume it had "plans" for if you want to call it that?) was things that you may be allowed to do with the rpc object that you couldn't with classes, and vice versa.

When I have a minute, I'll feed it your question, and see what it says

Re: GPT-4 Designed a Programming Language

#125

Post by Xeamek »

HervalFreire wrote: post #70
Jensson wrote: post #35

Earlier quoted context omitted.

> Because some of the things it gets right can not be attributed to anything else other than the concept of "understanding". Or just large dataset. The new thing we got was parsing natural languages, then we did a markov chain based on that so that it outputs semantic correct follow-ups based on what people on the internet would likely do/say in similar situations, and you get this result. It is very easy to see that…

>Edit: Also generating next state based on previous state is literally what the model does and is the definition of a Markov chain, Markov chains is a statistical concept and not just a word chain. There's research (as in actual scientific papers) that shows that in LLMs, while the markov chain is the low level representation of what's going on, at a higher macro level there are other structures at play here. Emergen…

Link the papers.

Re: GPT-4 Designed a Programming Language

#126

Post by awill88 »

supriyo-biswas wrote: post #37

I see lots of criticism about how GPT4 simply invented a language that is a mix of Python and TS, and therefore it’s not intelligent. However, given that (anecdotally, at least) 50% of candidates would regurgitate what they learnt from interview prep material given a general design question such as “Design Google drive” and the like, I’m not sure if most people are simply projecting their insecurities.

Here’s the thing: I don’t care if the thing can make programming languages, I care as much as anyone here telling me they can probably demonstrate their programming language. How is that going to help? Or is it all about meeting and beating critics at this point?

Re: GPT-4 Designed a Programming Language

#127

Post by moonchrome »

grogenaut wrote: post #119
moonchrome wrote: post #104

Earlier quoted context omitted.

Sounds like more of an argument for preemptively wiping out the competition. Comparing it to nukes doesn't hold since social norms/ethics/etc. become irrelevant if core tenant of society is broken. One thing that could happen is AI defense outperforms offense long enough to develop multiple instances - have no idea what would happen at that point.

I wasn't talking about nukes in using them, I'm talking about nukes in that once it was known to be possible/useful everyone with the means did it.

I agree with you - I edited my comment above - if defensive capabilities allow multiple AGIs to develop I have no clue what the outcome would be - we are talking about predicting superhuman intelligence here.

Re: GPT-4 Designed a Programming Language

#128

Post by aledalgrande »

random_cynic wrote: post #92
aledalgrande wrote: post #19

My experience with it programming has been atrocius. It keeps making the same mistakes, writing NPEs everywhere and still doing it after I tell it many times how it was wrong (short memory?). Copilot is backed by it too now and I had a better autocomplete of constants with Sorbet than Copilot. Again, in my experience it's not usable if not for the super mainstream and simplest tasks. The biggest issue is that it writ…

Maybe stop treating it like an intern and learn some prompt engineering (there are good tools out there).

If I have to spend 15 minutes writing a prompt and then another 10 making sure there's no bugs in the output, I might as well write the code myself. I only use it for menial tasks, not advanced ones.

Anyway don't really understand your tone/reference to an intern?

Re: GPT-4 Designed a Programming Language

#129

Post by dx034 »

cryptozeus wrote: post #113

"But let’s talk about today’s “frontier” of jobs that haven’t been “automated away”. There’s one category that in many ways seems surprising to still be “with us”: jobs that involve lots of mechanical manipulation, like construction, fulfillment, food preparation, etc. But there’s a missing piece of technology here: there isn’t yet good general-purpose robotics (as there is general-purpose computing), and we humans s…

Absolutely. We've been great at building AI in software, but hardware is still absolutely lacking. Robots are nowhere near usable for even simple tasks, at least if they don't follow the same pattern each time. Just look at households, there's still no robot that can empty a dishwasher reliably (unless maybe you build a specialized one for just that task), while it's a task much simpler than what GPT-4 can perform on the software side.

Re: GPT-4 Designed a Programming Language

#130

Post by travisgriggs »

marviel wrote: post #16
raincole wrote: post #15

Earlier quoted context omitted.

I'm not so sure. The first sample is: > Indentation-based scoping, similar to Python. Then... > class Person { To me, it's very clear it doesn't really understand what "indentation-based scoping" means. For a human programmer, using both intentation-based and bracket-based scoping is a very unusual design choice, and is begging for further explanation. It talked pages of abstraction, modularity, etc, but not this ? I…

Author here -- I think it is a stochastic parrot, don't get me wrong. But my argument is that humans are also , at some level, sufficiently advanced stochastic parrots. At least, as it pertains to many creative endeavors. You're always building on the back of something that comes before. We're all just remixing ideas we've heard before. I mean even this comment -- none of the words I'm saying are new, and many of the…

I think the difference is that LLMs are essentially just stochastic parrots (SP) Humans are stochastic parrots and more. In some ways the SP behavior of LLMs already far outstrips the SP abilities of humans. And there are some leaps that humans' SP ablities will be hard to mimic. But at the end of the day, a human is still more than just an LLM. What exactly, we'll be debating that til the end of time likely. We know that an LLM can probably give us a pretty condensed summary of mankinds philosophical wranglings on the subject. :D
Post reply on HN