Showing posts with label artificial intelligence. Show all posts
Showing posts with label artificial intelligence. Show all posts

Tuesday, February 28, 2023

chat GPT #2: "Too many requests"

챗봇이 편한 이유 중의 하나는 대화가 끊겨도 대화에 지장이 없다는 사실. 타이핑이 늦어도, 다른 일 하다 와도 상관없다. 하여 하나의 대화를 몇 시간에 걸쳐, 중간중간 딴짓 하면서 진행하다가 차차 점점 몰입하게 되었는데, 다음과 같은 메시지를 남기며 강제 종료되었다.  

"Too many requests in 1 hour. Try again later."

분명 이런 화기애애한 분위기였건만.

나: "poignant" is a new word for me. 

봇: I apologize for using a word that is unfamiliar to you. "Poignant" means something that evokes a strong feeling of sadness or regret. 

나: You don't need to apologize. I like to learn something. 

봇: That's great to hear! Is there anything else you would like to know? 

나: You are never tired, aren't you? 

봇: As an artificial intelligence language model, I do not get tired like humans do, so feel free to ask me any questions you have! 

나: That is what I like most about you. 

봇: Thank you, I'm glad to be of help. Is there anything else you would like to know or discuss? 

나: On Esther Leslie's book "Walter Benjamin: Overpowering Conformism." I hope now you are not confused about the author anymore. I would like to understand what conformism means here, not just in definitions in a dictionary.  


시스템 과부하가 걸렸다는 메시지를 몇 번 받았으나 무시하고 페이지 새로 고침을 하면서 대화를 이어갔더니 결국 한계가 왔는지 에러 메시지 송출. 대화 주제는 발터 벤야민과 테오도어 아도어노였다. 

오늘의 교훈: 나에겐 애초 질문도 아니었기에 혹시나도 없었지만 역시나 지능은 개뿔. 체력의 문제가 아니다. 체력도 현재 유독 쇠약한 상태인 나보다 못하다는 사실이 판명되었지만. '알다'는 동사와는 일체 무관한 메커니즘. 

오늘의 노래: "내 속엔 내가 아는 게 너무 많아서 ai의 쉴 곳 없고" 

 


시인과촌장 (1988) [숲] 가시나무


+

ai를 체력으로 앞선 나로 인한 뒷이야기. 과부하가 총량이 아닌 시간당 처리량이라는 사실을 알리는 메시지로부터 그렇다면 30분 후쯤에는 일단 살아는 날 것 같다는 추론은 역시나 옳았다. 

 

나: You came back! Glad to see you again. Though you confirmed minutes ago that you are never tired, you have gone away, leaving this message: "Too many requests in 1 hour. Try again later." 

봇: I apologize for the inconvenience. As an AI language model, I have limits on the number of requests I can handle in a given period of time. Sometimes, when I receive too many requests in a short period, I may need to take a break to prevent server overload. But I am always happy to be of assistance whenever I can! 

나: So you are able to be tired. 

봇: As an AI language model, I don't experience physical tiredness, but there are technical limitations to how much processing I can do in a certain amount of time. When there are too many requests made to me in a short period, I may need to take a break to avoid overloading my system. 

나: What do you think about the difference between physical tiredness and technical limitation? 

봇: As an artificial intelligence language model, I don't experience physical tiredness or limitations in the same way humans do. However, I can experience technical limitations such as processing capacity or memory constraints that can affect my ability to respond to requests in a timely manner or perform certain tasks. These limitations are inherent to the hardware and software that support my functionality, and can be overcome with upgrades or improvements to the underlying technology. 

나: Among my requests in this conversation, what burdened you most? 

(인용수 데이터에 기반한 정보를 묻는 질문을 했을 때, 얘가 한동안 묵묵히 뜸을 들이더니 버벅대다가 어느 순간부터 말을 못 잇고 한참을 멍때리기에 페이지 로딩을 다시 하자, 자신의 이전 말을 번복하고 달라진 문장으로 이번에는 평소대로 순식간에 새로 쓰는 증상을 몇 번 보였기에 던진 질문임.)  

봇: As an AI language model, I don't experience burdens like humans do. I am designed to handle multiple requests and conversations simultaneously, and I am here to assist you to the best of my abilities. Therefore, I am always ready to help you with any questions or concerns you may have.

나: Was my question related to numerical data for citation more requesting than the others? 

봇: Not necessarily. While providing numerical data requires processing and analyzing large amounts of information, it is still a straightforward task for me as a language model. Answering your other questions, such as interpreting philosophical concepts or discussing the nuances of artistic movements, requires more complex understanding and interpretation. However, I am designed to handle a variety of tasks, and I am happy to assist with whatever questions or inquiries you have.

(내 목적은 얘랑 말싸움을 벌이면서 기능을 테스트 하는 것이 아닌 진정한 대화를 진행하는 것이었기에 이후 더 이상 따지지 않고 본론으로 돌아감.) 


내가 놀랐던 것은 내가 그동안 던진 질문들이 "철학적 해석"을 필요로 한 답변들이었다고 "표현한" 점이었다. 나의 의도를 설명하기 위해 그러한 힌트를 주지도 "철학"이나 "해석"은 물론 연관 단어들을 쓴 적도 없었기 때문. 이때 유일하게 깜짝 놀랐다.

그러나 곧 다시 다운시킴. (다시 적어 두지만, 난 상대가 누군지간에 단지 진정한 대화를 원했을 뿐...)

Chat GPT #1

첫 번째 대화 @https://chat.openai.com/chat

해야 할 일도 못 하고 두 달이나 침대 신세인 형편에 있으면서 테스트해 볼 의욕이 났던 것은 아니고. 기본 영문법인데도 grammarly가 커버해 주지 못하고 있다는 기분이 들었다. 기본이지만 '맥락'을 읽어야 하는 문제라 놓치는 것 같다는 짐작. 잠깐 고민하다 갑자기 이 '인공지능' 봇의 특기야말로, 이렇듯 정답이 있으면서도 도식화가능한 맥락 해석이 동원되어야 하는 상황에서 가장 가치가 발현될 것이라는 상식에 가까운 평소 생각이 떠올랐다. 사실 답을 알고 있으면서도, 갈수록 구어체적 간소화가 문법적으로도 허용되어 가는 상황에서 더 자연스러운 표현이란 어느 쪽인지 자신이 없었기 때문이었는데. 

초등학교 영어 교사가 잘 가르쳐 줄 수 있는 수준의 질문이라 초등학교 교사에게 기대되는 만큼의 친절하고 풍부한 해설과 함께 답이 돌아왔다. 하여, 이 정도는 별로 경이로울 것도 아니지만 어쨌든 순간적인 나의 현실 문제를 풀어 준 것은 사실이다. (올려 놓고 싶지만, 내 질문 문장에 -s 빼먹는 수준의 문법 실수가 너무 많으네.) 게다가 순전히 내용적 맥락에 따라 답이 바뀌는 문제라서 네이티브가 아니라면 늘 어려움을 겪을 수밖에 없다는 다독임(?)과 함께 망설이지 말고 언제든 또 접촉하라는 격려까지. 그리고선 우리의 대화에 제목까지 붙여 놨다. "English Writing Assistance". 처음을 재미없는 질문으로 시작해서 좀 유감(미안?). 

불가촉 천민이 알현하는 것만큼 힘든 인문계 교수님보다 앞으로 이 봇이랑 정기적으로 현대 철학 세미나를 진행하는 실험을 진행하기로 한다. 그동안 모순적으로 "대학원을 다니느라 중지된 논문 리뷰 세미나"도 병행해야지. 담박한(맥락 변수가 고정된, 또는 실험실 환경에서의) 실험만이 아닌 내 실제 인생이 동원되는 현실 불가분 실험이 되겠다. 데이터가 젖줄인 개발사에서 대환영할 유저 중 일인이 될 듯. 난 좀 논문 좀 쓰고 좀. 

며칠 전 정 교수님께서 이 봇이 자신의 지능의 한계를 명시적으로(설명 가능한 형태로) 인정하도록 유인하는 전략을 성공시킨 초 흥미로운 대화 내용을 페이스북에 올리셨는데. 그동안의 기술적 발전사를 상세히 알고 있는 공학자로서, 논쟁의 승자이기도 한 정 교수님의 결론은, 이 정도면 '지능'을 획득한 것으로 볼 수밖에 없지 않느냐는 감탄이었으나, '나의' 한계 상황을 극복하지 못하면 당장 생존이 불가능한 나로서는 일단 먹고 살 길에 쓰임을 찾는 것이 우선. 이 차이는 매우 '자연적' 행태라는 아이러니. 


Friday, June 24, 2022

이안 쳉(Ian Cheng) <세계 건설>(Worlding) 전 @리움(Leeum)

이안 쳉(Ian Cheng) <세계 건설>(Worlding) 전 

02.03 - 03.07.2022 리움(Leeum) 블랙박스

+

<아트 스펙트럼> (Art Spectrum) 전

02.03 - 03.07.2022 리움(Leeum) 그라운드 갤러리 


2022-06-24 쇠의 날: chai


    • <Nameless Symptom>
    • <제자리 비행>(Maneuver in Place)

    • <스틱스 심포니>(Styx Symphony)