4 ms·

I agree with your point about the dangers of chatbots being mistaken as sentient. And I also recognize the dangers of the self-improving systems like this in ge

I agree with your point about the dangers of chatbots being mistaken as sentient. And I also recognize the dangers of the self-improving systems like this in general -- that's why I'm not letting it run loose with a LangChainAI Agent script to self-execute a new language (though wouldn't that be fascinating?) While the language it generated needs refinement, the design process is iterative, and it takes time to develop a programming language that is worth using. I encourage you to consider how you would approach designing a programming language. Like you, I would likely start by incorporating my favorite features from languages I love and addressing problems I've encountered in the past, similar to the dialogue GPT and I have in the post. As I mention in the post -- creativity often involves remixing existing ideas, and GPT-4's attempt at generating a programming language is no exception. Notably, it also called out explicitly the constructs it was borrowing from other languages when designing the language. It's worth noting that this was GPT-4's initial attempt at something highly complex, and even if there are "plot holes" -- it's a noteworthy achievement in and of itself.

krupan 4y ago
Ok, but take a step back and really look at the whole situation here. So you've created a machine that solves problems in kind of the same way humans do. Yay? What overall problem is a machine like this solving? How is this better than just having a human do the job? It seems to have most of the same drawbacks as humans, and then some on top of that
- justinclift 4y ago
  > How is this better than just having a human do the job? The cynical take is quite clearly: "Because sales people can sell it" eg wrap some pretty UX around it, tell people it's "nearly human", then have your sales people try and get non-technical CxO types (or consultancies) to cut them a cheque
- marviel 4y ago
  It's a hard question to be honest. If you look at the incentives involved, people are very incentivized to create these machines to reduce the amount of toil humans have to endure. I don't disagree about the risks at scale of something like this; as I mention in the post -- I don't get pure warm fuzzies about this at all. But I also don't know how to stop AI progress, because the incentives to keep it going are so real. Government regulation has been brought up multiple times, most recently by Sama himself on twitter. I think it will have to come, but I don't know what flavor it should take.
- ummonk 4y ago
  I don’t have a massive pool of skilled slave labor I can delegate work to, but I can delegate work to ChatGPT.
- KyeRussell 4y ago
  What!? Look back at the last few hundred years, where the majority of great leaps have been the result of getting a machine to do something that humans did.
geraneum 4y ago
> As I mention in the post -- creativity often involves remixing existing ideas This is carrying the weight of your core argument. It’s not wrong but it’s not completely correct either. Creativity involves remixing but is not all remixing. Is it? Not even all humans are at the same level in this. Number of attempts means nothing if you don’t add features that the LLM desperately needs to be grounded in the reality, or perform logical reasoning beyond RLHF.
- marviel 4y ago
  You're totally right -- random recombination isn't good enough. not trying to make that case at all.

krupan 4y ago

Ok, but take a step back and really look at the whole situation here. So you've created a machine that solves problems in kind of the same way humans do. Yay? What overall problem is a machine like this solving? How is this better than just having a human do the job? It seems to have most of the same drawbacks as humans, and then some on top of that

justinclift 4y ago
> How is this better than just having a human do the job? The cynical take is quite clearly: "Because sales people can sell it" eg wrap some pretty UX around it, tell people it's "nearly human", then have your sales people try and get non-technical CxO types (or consultancies) to cut them a cheque
marviel 4y ago
It's a hard question to be honest. If you look at the incentives involved, people are very incentivized to create these machines to reduce the amount of toil humans have to endure. I don't disagree about the risks at scale of something like this; as I mention in the post -- I don't get pure warm fuzzies about this at all. But I also don't know how to stop AI progress, because the incentives to keep it going are so real. Government regulation has been brought up multiple times, most recently by Sama himself on twitter. I think it will have to come, but I don't know what flavor it should take.
ummonk 4y ago
I don’t have a massive pool of skilled slave labor I can delegate work to, but I can delegate work to ChatGPT.
KyeRussell 4y ago
What!? Look back at the last few hundred years, where the majority of great leaps have been the result of getting a machine to do something that humans did.

geraneum 4y ago

> As I mention in the post -- creativity often involves remixing existing ideas This is carrying the weight of your core argument. It’s not wrong but it’s not completely correct either. Creativity involves remixing but is not all remixing. Is it? Not even all humans are at the same level in this. Number of attempts means nothing if you don’t add features that the LLM desperately needs to be grounded in the reality, or perform logical reasoning beyond RLHF.

marviel 4y ago
You're totally right -- random recombination isn't good enough. not trying to make that case at all.